2023-04-27 12:19:53,970 INFO [train.py:976] (5/8) Training started 2023-04-27 12:19:53,970 INFO [train.py:986] (5/8) Device: cuda:5 2023-04-27 12:19:53,971 INFO [train.py:995] (5/8) {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, 'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.23.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'a23383c5a381713b51e9014f3f05d096f8aceec3', 'k2-git-date': 'Wed Apr 26 15:33:33 2023', 'lhotse-version': '1.14.0.dev+git.b61b917.dirty', 'torch-version': '1.13.1', 'torch-cuda-available': True, 'torch-cuda-version': '11.6', 'python-version': '3.1', 'icefall-git-branch': 'master', 'icefall-git-sha1': '45c13e9-dirty', 'icefall-git-date': 'Mon Apr 24 15:00:02 2023', 'icefall-path': '/k2-dev/yangyifan/icefall-master', 'k2-path': '/k2-dev/yangyifan/anaconda3/envs/icefall/lib/python3.10/site-packages/k2-1.23.4.dev20230427+cuda11.6.torch1.13.1-py3.10-linux-x86_64.egg/k2/__init__.py', 'lhotse-path': '/k2-dev/yangyifan/anaconda3/envs/icefall/lib/python3.10/site-packages/lhotse-1.14.0.dev0+git.b61b917.dirty-py3.10.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-3-0423201227-84b4557756-8lx4n', 'IP address': '10.177.6.147'}, 'world_size': 8, 'master_port': 12354, 'tensorboard': True, 'num_epochs': 30, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('pruned_transducer_stateless7/exp_multidataset'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'base_lr': 0.05, 'lr_batches': 5000, 'lr_epochs': 3.5, 'context_size': 2, 'prune_range': 5, 'lm_scale': 0.25, 'am_scale': 0.0, 'simple_loss_scale': 0.5, 'seed': 42, 'print_diagnostics': False, 'inf_check': False, 'save_every_n': 2000, 'keep_last_k': 30, 'average_period': 200, 'use_fp16': True, 'use_multidataset': True, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'cv_manifest_dir': PosixPath('data/en/fbank'), 'max_duration': 700, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'blank_id': 0, 'vocab_size': 500} 2023-04-27 12:19:53,972 INFO [train.py:997] (5/8) About to create model 2023-04-27 12:19:54,651 INFO [zipformer.py:178] (5/8) At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8. 2023-04-27 12:19:54,667 INFO [train.py:1001] (5/8) Number of model parameters: 70369391 2023-04-27 12:19:57,169 INFO [train.py:1016] (5/8) Using DDP 2023-04-27 12:19:58,263 INFO [multidataset.py:46] (5/8) About to get multidataset train cuts 2023-04-27 12:19:58,264 INFO [multidataset.py:49] (5/8) Loading LibriSpeech in lazy mode 2023-04-27 12:19:58,281 INFO [multidataset.py:65] (5/8) Loading GigaSpeech 1998 splits in lazy mode 2023-04-27 12:20:00,752 INFO [multidataset.py:72] (5/8) Loading CommonVoice in lazy mode 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:230] (5/8) Enable MUSAN 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:231] (5/8) About to get Musan cuts 2023-04-27 12:20:03,010 INFO [asr_datamodule.py:255] (5/8) Enable SpecAugment 2023-04-27 12:20:03,010 INFO [asr_datamodule.py:256] (5/8) Time warp factor: 80 2023-04-27 12:20:03,011 INFO [asr_datamodule.py:266] (5/8) Num frame mask: 10 2023-04-27 12:20:03,011 INFO [asr_datamodule.py:279] (5/8) About to create train dataset 2023-04-27 12:20:03,011 INFO [asr_datamodule.py:306] (5/8) Using DynamicBucketingSampler. 2023-04-27 12:20:07,524 INFO [asr_datamodule.py:321] (5/8) About to create train dataloader 2023-04-27 12:20:07,525 INFO [asr_datamodule.py:435] (5/8) About to get dev-clean cuts 2023-04-27 12:20:07,526 INFO [asr_datamodule.py:442] (5/8) About to get dev-other cuts 2023-04-27 12:20:07,527 INFO [asr_datamodule.py:352] (5/8) About to create dev dataset 2023-04-27 12:20:07,761 INFO [asr_datamodule.py:369] (5/8) About to create dev dataloader 2023-04-27 12:20:25,619 INFO [train.py:904] (5/8) Epoch 1, batch 0, loss[loss=7.58, simple_loss=6.886, pruned_loss=6.931, over 15409.00 frames. ], tot_loss[loss=7.58, simple_loss=6.886, pruned_loss=6.931, over 15409.00 frames. ], batch size: 190, lr: 2.50e-02, grad_scale: 2.0 2023-04-27 12:20:25,620 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 12:20:32,881 INFO [train.py:938] (5/8) Epoch 1, validation: loss=6.911, simple_loss=6.238, pruned_loss=6.721, over 944034.00 frames. 2023-04-27 12:20:32,882 INFO [train.py:939] (5/8) Maximum memory allocated so far is 11043MB 2023-04-27 12:20:36,292 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:20:52,225 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:20:55,596 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=91.16 vs. limit=5.0 2023-04-27 12:21:03,481 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=8.43 vs. limit=2.0 2023-04-27 12:21:17,135 INFO [train.py:904] (5/8) Epoch 1, batch 50, loss[loss=1.349, simple_loss=1.188, pruned_loss=1.433, over 17045.00 frames. ], tot_loss[loss=2.164, simple_loss=1.961, pruned_loss=1.949, over 749106.22 frames. ], batch size: 50, lr: 2.75e-02, grad_scale: 2.0 2023-04-27 12:21:21,685 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=18.17 vs. limit=2.0 2023-04-27 12:21:41,184 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=18.57 vs. limit=2.0 2023-04-27 12:21:46,460 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:22:02,802 WARNING [train.py:894] (5/8) Grad scale is small: 0.001953125 2023-04-27 12:22:02,803 INFO [train.py:904] (5/8) Epoch 1, batch 100, loss[loss=1.177, simple_loss=1.004, pruned_loss=1.361, over 16771.00 frames. ], tot_loss[loss=1.658, simple_loss=1.475, pruned_loss=1.636, over 1312930.14 frames. ], batch size: 124, lr: 3.00e-02, grad_scale: 0.00390625 2023-04-27 12:22:13,669 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 5.700e+01 2.326e+02 5.095e+02 1.135e+03 3.099e+06, threshold=1.019e+03, percent-clipped=0.0 2023-04-27 12:22:18,331 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=9.69 vs. limit=2.0 2023-04-27 12:22:20,962 WARNING [optim.py:388] (5/8) Scaling gradients by 0.0112030990421772, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:21,068 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.24, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.974e+09, grad_sumsq = 5.165e+10, orig_rms_sq=3.822e-02 2023-04-27 12:22:38,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=3.19 vs. limit=2.0 2023-04-27 12:22:43,835 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:22:46,131 WARNING [optim.py:388] (5/8) Scaling gradients by 0.0022801109589636326, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:46,241 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.68, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.358e+11, grad_sumsq = 3.286e+12, orig_rms_sq=4.131e-02 2023-04-27 12:22:49,653 WARNING [optim.py:388] (5/8) Scaling gradients by 0.04246773198246956, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:49,794 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.92, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=5.309e+08, grad_sumsq = 1.285e+10, orig_rms_sq=4.131e-02 2023-04-27 12:22:51,352 WARNING [optim.py:388] (5/8) Scaling gradients by 0.000716241542249918, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:51,480 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.93, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.876e+12, grad_sumsq = 4.542e+13, orig_rms_sq=4.131e-02 2023-04-27 12:22:52,994 INFO [train.py:904] (5/8) Epoch 1, batch 150, loss[loss=1.045, simple_loss=0.8777, pruned_loss=1.194, over 15894.00 frames. ], tot_loss[loss=1.413, simple_loss=1.238, pruned_loss=1.462, over 1756477.12 frames. ], batch size: 35, lr: 3.25e-02, grad_scale: 0.00390625 2023-04-27 12:22:53,728 WARNING [optim.py:388] (5/8) Scaling gradients by 0.049951765686273575, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:53,833 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.92, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=3.845e+08, grad_sumsq = 8.968e+09, orig_rms_sq=4.287e-02 2023-04-27 12:22:58,545 WARNING [optim.py:388] (5/8) Scaling gradients by 0.00609818659722805, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:58,655 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.49, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.379e+10, grad_sumsq = 3.140e+11, orig_rms_sq=4.392e-02 2023-04-27 12:23:16,870 WARNING [optim.py:388] (5/8) Scaling gradients by 0.059935860335826874, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:16,971 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.63, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.808e+08, grad_sumsq = 3.916e+09, orig_rms_sq=4.617e-02 2023-04-27 12:23:28,341 WARNING [optim.py:388] (5/8) Scaling gradients by 0.060559310019016266, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:28,447 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.90, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=2.556e+08, grad_sumsq = 6.221e+09, orig_rms_sq=4.108e-02 2023-04-27 12:23:42,816 WARNING [train.py:894] (5/8) Grad scale is small: 0.00390625 2023-04-27 12:23:42,816 INFO [train.py:904] (5/8) Epoch 1, batch 200, loss[loss=0.9926, simple_loss=0.8373, pruned_loss=1.036, over 16504.00 frames. ], tot_loss[loss=1.268, simple_loss=1.1, pruned_loss=1.328, over 2106493.08 frames. ], batch size: 75, lr: 3.50e-02, grad_scale: 0.0078125 2023-04-27 12:23:47,488 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=10.71 vs. limit=2.0 2023-04-27 12:23:51,004 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 2.076e+02 2.707e+02 4.762e+02 1.423e+06, threshold=5.415e+02, percent-clipped=11.0 2023-04-27 12:23:51,004 WARNING [optim.py:388] (5/8) Scaling gradients by 0.002041660714894533, model_norm_threshold=541.4743041992188 2023-04-27 12:23:51,146 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.86, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=6.060e+10, grad_sumsq = 1.544e+12, orig_rms_sq=3.924e-02 2023-04-27 12:23:54,719 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=11.02 vs. limit=2.0 2023-04-27 12:24:00,579 WARNING [optim.py:388] (5/8) Scaling gradients by 0.02974529005587101, model_norm_threshold=541.4743041992188 2023-04-27 12:24:00,684 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.85, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=2.829e+08, grad_sumsq = 7.259e+09, orig_rms_sq=3.897e-02 2023-04-27 12:24:01,477 WARNING [optim.py:388] (5/8) Scaling gradients by 0.01955481991171837, model_norm_threshold=541.4743041992188 2023-04-27 12:24:01,585 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.84, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=6.426e+08, grad_sumsq = 1.649e+10, orig_rms_sq=3.897e-02 2023-04-27 12:24:30,807 INFO [train.py:904] (5/8) Epoch 1, batch 250, loss[loss=0.9848, simple_loss=0.8242, pruned_loss=0.9972, over 17240.00 frames. ], tot_loss[loss=1.168, simple_loss=1.006, pruned_loss=1.214, over 2366290.90 frames. ], batch size: 52, lr: 3.75e-02, grad_scale: 0.0078125 2023-04-27 12:24:33,597 WARNING [optim.py:388] (5/8) Scaling gradients by 0.057925041764974594, model_norm_threshold=541.4743041992188 2023-04-27 12:24:33,740 INFO [optim.py:450] (5/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.59, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=5.145e+07, grad_sumsq = 1.327e+09, orig_rms_sq=3.876e-02 2023-04-27 12:24:50,605 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=15.55 vs. limit=2.0 2023-04-27 12:24:52,141 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=3.96 vs. limit=2.0 2023-04-27 12:25:06,827 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=64.12 vs. limit=5.0 2023-04-27 12:25:16,083 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:25:20,569 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:25:21,114 WARNING [train.py:894] (5/8) Grad scale is small: 0.0078125 2023-04-27 12:25:21,114 INFO [train.py:904] (5/8) Epoch 1, batch 300, loss[loss=0.9749, simple_loss=0.8118, pruned_loss=0.9522, over 16774.00 frames. ], tot_loss[loss=1.095, simple_loss=0.9373, pruned_loss=1.122, over 2578926.75 frames. ], batch size: 57, lr: 4.00e-02, grad_scale: 0.015625 2023-04-27 12:25:30,078 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 9.163e+01 1.282e+02 1.731e+02 2.635e+02 2.769e+04, threshold=3.463e+02, percent-clipped=7.0 2023-04-27 12:25:32,234 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=51.12 vs. limit=5.0 2023-04-27 12:25:45,058 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=34.73 vs. limit=5.0 2023-04-27 12:26:12,615 INFO [train.py:904] (5/8) Epoch 1, batch 350, loss[loss=0.8597, simple_loss=0.7169, pruned_loss=0.7975, over 16868.00 frames. ], tot_loss[loss=1.042, simple_loss=0.8864, pruned_loss=1.051, over 2740007.83 frames. ], batch size: 116, lr: 4.25e-02, grad_scale: 0.015625 2023-04-27 12:26:18,704 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:26:52,626 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:27:06,014 INFO [train.py:904] (5/8) Epoch 1, batch 400, loss[loss=0.911, simple_loss=0.7562, pruned_loss=0.8222, over 15511.00 frames. ], tot_loss[loss=1.008, simple_loss=0.8509, pruned_loss=0.9994, over 2872430.33 frames. ], batch size: 190, lr: 4.50e-02, grad_scale: 0.03125 2023-04-27 12:27:17,882 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 7.937e+01 1.110e+02 1.413e+02 1.837e+02 3.136e+02, threshold=2.826e+02, percent-clipped=0.0 2023-04-27 12:27:43,835 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=34.45 vs. limit=5.0 2023-04-27 12:27:43,890 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=46.29 vs. limit=5.0 2023-04-27 12:27:46,079 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:27:56,252 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:27:59,555 INFO [train.py:904] (5/8) Epoch 1, batch 450, loss[loss=0.9525, simple_loss=0.7728, pruned_loss=0.8808, over 17139.00 frames. ], tot_loss[loss=0.9826, simple_loss=0.8236, pruned_loss=0.9593, over 2976550.13 frames. ], batch size: 48, lr: 4.75e-02, grad_scale: 0.03125 2023-04-27 12:28:14,301 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=2.90 vs. limit=2.0 2023-04-27 12:28:37,204 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5315, 4.3772, 4.5621, 4.5234, 4.5576, 4.5690, 4.5646, 4.5424], device='cuda:5'), covar=tensor([0.0194, 0.0189, 0.0105, 0.0192, 0.0147, 0.0142, 0.0146, 0.0144], device='cuda:5'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:5'), out_proj_covar=tensor([8.9770e-06, 9.0696e-06, 8.9892e-06, 9.1408e-06, 9.0167e-06, 8.8186e-06, 9.0867e-06, 8.9497e-06], device='cuda:5') 2023-04-27 12:28:51,194 INFO [train.py:904] (5/8) Epoch 1, batch 500, loss[loss=0.867, simple_loss=0.6963, pruned_loss=0.7925, over 16769.00 frames. ], tot_loss[loss=0.963, simple_loss=0.8014, pruned_loss=0.9248, over 3051559.00 frames. ], batch size: 39, lr: 4.99e-02, grad_scale: 0.0625 2023-04-27 12:29:01,369 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 7.996e+01 1.277e+02 1.626e+02 2.121e+02 5.144e+02, threshold=3.251e+02, percent-clipped=17.0 2023-04-27 12:29:15,021 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=11.30 vs. limit=2.0 2023-04-27 12:29:25,676 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=17.36 vs. limit=5.0 2023-04-27 12:29:34,623 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8504, 4.8663, 4.8529, 4.6896, 4.8784, 4.7782, 4.8937, 4.7963], device='cuda:5'), covar=tensor([0.0119, 0.0120, 0.0161, 0.0386, 0.0110, 0.0154, 0.0068, 0.0236], device='cuda:5'), in_proj_covar=tensor([0.0010, 0.0010, 0.0011, 0.0010, 0.0011, 0.0010, 0.0010, 0.0011], device='cuda:5'), out_proj_covar=tensor([1.0717e-05, 1.0961e-05, 1.0652e-05, 1.1190e-05, 1.0832e-05, 1.0778e-05, 1.0452e-05, 1.0881e-05], device='cuda:5') 2023-04-27 12:29:36,588 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=2.52 vs. limit=2.0 2023-04-27 12:29:44,937 INFO [train.py:904] (5/8) Epoch 1, batch 550, loss[loss=0.944, simple_loss=0.7572, pruned_loss=0.8367, over 17137.00 frames. ], tot_loss[loss=0.9498, simple_loss=0.7851, pruned_loss=0.8964, over 3110512.88 frames. ], batch size: 47, lr: 4.98e-02, grad_scale: 0.0625 2023-04-27 12:29:58,110 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:30:03,893 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:30:15,123 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=2.09 vs. limit=2.0 2023-04-27 12:30:17,191 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 12:30:26,961 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:30:38,234 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:30:38,756 INFO [train.py:904] (5/8) Epoch 1, batch 600, loss[loss=1.015, simple_loss=0.8176, pruned_loss=0.8638, over 17049.00 frames. ], tot_loss[loss=0.9365, simple_loss=0.7694, pruned_loss=0.8676, over 3162612.03 frames. ], batch size: 55, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:30:48,391 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 8.862e+01 1.447e+02 1.955e+02 2.870e+02 6.309e+02, threshold=3.911e+02, percent-clipped=20.0 2023-04-27 12:31:02,550 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:31:02,669 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=9.89 vs. limit=5.0 2023-04-27 12:31:07,499 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:31:28,003 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:31:30,999 INFO [train.py:904] (5/8) Epoch 1, batch 650, loss[loss=0.8472, simple_loss=0.6853, pruned_loss=0.6959, over 16786.00 frames. ], tot_loss[loss=0.9262, simple_loss=0.7576, pruned_loss=0.8404, over 3200286.76 frames. ], batch size: 39, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:31:31,349 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:31:32,151 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:31:41,012 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=5.02 vs. limit=2.0 2023-04-27 12:32:22,396 INFO [train.py:904] (5/8) Epoch 1, batch 700, loss[loss=0.8036, simple_loss=0.6524, pruned_loss=0.6391, over 16838.00 frames. ], tot_loss[loss=0.9116, simple_loss=0.7443, pruned_loss=0.8074, over 3222315.65 frames. ], batch size: 42, lr: 4.98e-02, grad_scale: 0.25 2023-04-27 12:32:31,781 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 2.100e+02 3.098e+02 3.988e+02 9.500e+02, threshold=6.196e+02, percent-clipped=26.0 2023-04-27 12:32:45,963 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=2.39 vs. limit=2.0 2023-04-27 12:33:00,861 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:33:05,955 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:33:14,017 INFO [train.py:904] (5/8) Epoch 1, batch 750, loss[loss=0.8144, simple_loss=0.6696, pruned_loss=0.6179, over 16438.00 frames. ], tot_loss[loss=0.8967, simple_loss=0.733, pruned_loss=0.7723, over 3219235.72 frames. ], batch size: 146, lr: 4.97e-02, grad_scale: 0.25 2023-04-27 12:33:18,018 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.95 vs. limit=5.0 2023-04-27 12:33:51,952 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:34:06,267 INFO [train.py:904] (5/8) Epoch 1, batch 800, loss[loss=0.7711, simple_loss=0.6571, pruned_loss=0.5333, over 17221.00 frames. ], tot_loss[loss=0.8709, simple_loss=0.7153, pruned_loss=0.7258, over 3236925.38 frames. ], batch size: 45, lr: 4.97e-02, grad_scale: 0.5 2023-04-27 12:34:17,532 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.855e+02 4.131e+02 5.648e+02 8.889e+02, threshold=8.261e+02, percent-clipped=19.0 2023-04-27 12:34:54,128 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:34:58,589 INFO [train.py:904] (5/8) Epoch 1, batch 850, loss[loss=0.7229, simple_loss=0.6258, pruned_loss=0.4767, over 17140.00 frames. ], tot_loss[loss=0.8402, simple_loss=0.6951, pruned_loss=0.6761, over 3254785.26 frames. ], batch size: 47, lr: 4.96e-02, grad_scale: 0.5 2023-04-27 12:35:02,311 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-27 12:35:03,258 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=2.14 vs. limit=2.0 2023-04-27 12:35:46,345 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.15 vs. limit=2.0 2023-04-27 12:35:50,491 INFO [train.py:904] (5/8) Epoch 1, batch 900, loss[loss=0.6818, simple_loss=0.5951, pruned_loss=0.4366, over 17148.00 frames. ], tot_loss[loss=0.8043, simple_loss=0.6713, pruned_loss=0.6244, over 3267234.91 frames. ], batch size: 46, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:35:57,667 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:36:00,424 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.021e+02 3.478e+02 3.989e+02 5.953e+02 1.563e+03, threshold=7.979e+02, percent-clipped=10.0 2023-04-27 12:36:05,772 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3906, 5.1689, 5.1484, 5.3374, 4.9308, 5.3466, 4.9758, 5.3215], device='cuda:5'), covar=tensor([0.0698, 0.0868, 0.1061, 0.0770, 0.1625, 0.0929, 0.1205, 0.0964], device='cuda:5'), in_proj_covar=tensor([0.0048, 0.0044, 0.0052, 0.0048, 0.0049, 0.0053, 0.0040, 0.0052], device='cuda:5'), out_proj_covar=tensor([4.2709e-05, 3.9918e-05, 4.4559e-05, 4.2798e-05, 4.5533e-05, 4.4404e-05, 3.7804e-05, 5.0941e-05], device='cuda:5') 2023-04-27 12:36:08,847 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:36:14,426 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:36:37,793 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:42,352 INFO [train.py:904] (5/8) Epoch 1, batch 950, loss[loss=0.6573, simple_loss=0.5838, pruned_loss=0.4021, over 17209.00 frames. ], tot_loss[loss=0.7741, simple_loss=0.6518, pruned_loss=0.5799, over 3285835.76 frames. ], batch size: 45, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:36:44,147 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:50,857 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.43 vs. limit=5.0 2023-04-27 12:37:18,315 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9119, 3.9157, 3.5176, 3.8997, 3.7892, 4.0341, 3.7123, 3.8139], device='cuda:5'), covar=tensor([0.3483, 0.3925, 0.4304, 0.3275, 0.3688, 0.2694, 0.4063, 0.3679], device='cuda:5'), in_proj_covar=tensor([0.0068, 0.0065, 0.0076, 0.0056, 0.0068, 0.0054, 0.0065, 0.0063], device='cuda:5'), out_proj_covar=tensor([5.7957e-05, 5.9361e-05, 7.2402e-05, 4.8171e-05, 5.9821e-05, 5.2540e-05, 6.0366e-05, 5.7110e-05], device='cuda:5') 2023-04-27 12:37:35,471 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:37:35,999 INFO [train.py:904] (5/8) Epoch 1, batch 1000, loss[loss=0.5795, simple_loss=0.5091, pruned_loss=0.3594, over 16771.00 frames. ], tot_loss[loss=0.7378, simple_loss=0.6266, pruned_loss=0.5348, over 3293422.49 frames. ], batch size: 83, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:37:46,293 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.400e+02 3.913e+02 4.932e+02 6.154e+02 1.349e+03, threshold=9.864e+02, percent-clipped=6.0 2023-04-27 12:37:58,576 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1088, 3.0448, 2.3638, 2.4155, 2.1119, 2.5799, 2.4815, 2.6240], device='cuda:5'), covar=tensor([0.9757, 0.7249, 0.8020, 0.7034, 0.9914, 0.8520, 0.7818, 0.5355], device='cuda:5'), in_proj_covar=tensor([0.0075, 0.0074, 0.0075, 0.0077, 0.0076, 0.0080, 0.0076, 0.0067], device='cuda:5'), out_proj_covar=tensor([6.3850e-05, 7.0958e-05, 6.5193e-05, 6.7648e-05, 6.6125e-05, 7.1663e-05, 6.2986e-05, 5.9169e-05], device='cuda:5') 2023-04-27 12:38:21,653 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:38:29,497 INFO [train.py:904] (5/8) Epoch 1, batch 1050, loss[loss=0.5631, simple_loss=0.512, pruned_loss=0.3241, over 17194.00 frames. ], tot_loss[loss=0.7097, simple_loss=0.6088, pruned_loss=0.4971, over 3297675.33 frames. ], batch size: 44, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:38:47,530 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=2.23 vs. limit=2.0 2023-04-27 12:39:12,771 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:39:21,805 INFO [train.py:904] (5/8) Epoch 1, batch 1100, loss[loss=0.5271, simple_loss=0.4766, pruned_loss=0.3052, over 16951.00 frames. ], tot_loss[loss=0.6824, simple_loss=0.5906, pruned_loss=0.4633, over 3300619.35 frames. ], batch size: 90, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:39:33,206 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.538e+02 4.173e+02 5.189e+02 6.744e+02 1.137e+03, threshold=1.038e+03, percent-clipped=1.0 2023-04-27 12:40:17,546 INFO [train.py:904] (5/8) Epoch 1, batch 1150, loss[loss=0.5681, simple_loss=0.5094, pruned_loss=0.3321, over 12015.00 frames. ], tot_loss[loss=0.6564, simple_loss=0.5741, pruned_loss=0.4316, over 3299348.97 frames. ], batch size: 246, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:40:19,546 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:41:10,411 INFO [train.py:904] (5/8) Epoch 1, batch 1200, loss[loss=0.5223, simple_loss=0.4846, pruned_loss=0.286, over 17106.00 frames. ], tot_loss[loss=0.63, simple_loss=0.5565, pruned_loss=0.4022, over 3299503.88 frames. ], batch size: 53, lr: 4.93e-02, grad_scale: 2.0 2023-04-27 12:41:12,652 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:41:21,566 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.465e+02 4.824e+02 5.887e+02 7.169e+02 1.569e+03, threshold=1.177e+03, percent-clipped=1.0 2023-04-27 12:41:25,794 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:41:29,975 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:41:34,768 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:41:57,210 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:42:03,011 INFO [train.py:904] (5/8) Epoch 1, batch 1250, loss[loss=0.5395, simple_loss=0.5138, pruned_loss=0.2811, over 17128.00 frames. ], tot_loss[loss=0.6105, simple_loss=0.5441, pruned_loss=0.3793, over 3304865.28 frames. ], batch size: 47, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:42:17,894 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:42:25,410 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:49,932 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:56,589 INFO [train.py:904] (5/8) Epoch 1, batch 1300, loss[loss=0.5367, simple_loss=0.4951, pruned_loss=0.2956, over 16782.00 frames. ], tot_loss[loss=0.591, simple_loss=0.531, pruned_loss=0.3584, over 3310746.37 frames. ], batch size: 102, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:43:07,780 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.547e+02 4.841e+02 5.972e+02 7.377e+02 1.990e+03, threshold=1.194e+03, percent-clipped=4.0 2023-04-27 12:43:51,774 INFO [train.py:904] (5/8) Epoch 1, batch 1350, loss[loss=0.5139, simple_loss=0.4777, pruned_loss=0.279, over 15617.00 frames. ], tot_loss[loss=0.5739, simple_loss=0.5209, pruned_loss=0.3392, over 3315920.35 frames. ], batch size: 191, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:44:49,671 INFO [train.py:904] (5/8) Epoch 1, batch 1400, loss[loss=0.5158, simple_loss=0.4992, pruned_loss=0.2622, over 16751.00 frames. ], tot_loss[loss=0.5591, simple_loss=0.5122, pruned_loss=0.3229, over 3315841.57 frames. ], batch size: 62, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:45:00,104 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.619e+02 5.087e+02 6.130e+02 7.329e+02 1.559e+03, threshold=1.226e+03, percent-clipped=4.0 2023-04-27 12:45:20,341 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-27 12:45:27,728 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1434.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:45:46,457 INFO [train.py:904] (5/8) Epoch 1, batch 1450, loss[loss=0.4773, simple_loss=0.4585, pruned_loss=0.2459, over 16528.00 frames. ], tot_loss[loss=0.5445, simple_loss=0.5027, pruned_loss=0.3085, over 3311160.64 frames. ], batch size: 68, lr: 4.90e-02, grad_scale: 2.0 2023-04-27 12:46:34,505 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:46:41,360 INFO [train.py:904] (5/8) Epoch 1, batch 1500, loss[loss=0.4939, simple_loss=0.483, pruned_loss=0.2478, over 17147.00 frames. ], tot_loss[loss=0.5327, simple_loss=0.4958, pruned_loss=0.2963, over 3313548.55 frames. ], batch size: 49, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:46:44,226 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:46:51,045 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:46:52,707 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.422e+02 5.181e+02 6.478e+02 8.944e+02 1.260e+03, threshold=1.296e+03, percent-clipped=1.0 2023-04-27 12:47:23,146 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 2023-04-27 12:47:38,781 INFO [train.py:904] (5/8) Epoch 1, batch 1550, loss[loss=0.584, simple_loss=0.5349, pruned_loss=0.3215, over 12133.00 frames. ], tot_loss[loss=0.5254, simple_loss=0.492, pruned_loss=0.288, over 3314732.86 frames. ], batch size: 247, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:47:39,062 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:47:56,158 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:47:59,231 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 2023-04-27 12:48:35,338 INFO [train.py:904] (5/8) Epoch 1, batch 1600, loss[loss=0.496, simple_loss=0.4931, pruned_loss=0.2441, over 17070.00 frames. ], tot_loss[loss=0.5221, simple_loss=0.4903, pruned_loss=0.2836, over 3314510.38 frames. ], batch size: 50, lr: 4.88e-02, grad_scale: 4.0 2023-04-27 12:48:47,168 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.141e+02 5.212e+02 6.871e+02 8.557e+02 2.137e+03, threshold=1.374e+03, percent-clipped=9.0 2023-04-27 12:48:47,573 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:49:05,681 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:49:32,292 INFO [train.py:904] (5/8) Epoch 1, batch 1650, loss[loss=0.4606, simple_loss=0.4419, pruned_loss=0.2386, over 16876.00 frames. ], tot_loss[loss=0.5147, simple_loss=0.4867, pruned_loss=0.276, over 3324910.02 frames. ], batch size: 96, lr: 4.87e-02, grad_scale: 4.0 2023-04-27 12:49:57,826 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:50:29,821 INFO [train.py:904] (5/8) Epoch 1, batch 1700, loss[loss=0.4881, simple_loss=0.4943, pruned_loss=0.2357, over 16768.00 frames. ], tot_loss[loss=0.509, simple_loss=0.4847, pruned_loss=0.2697, over 3307885.39 frames. ], batch size: 57, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:50:41,831 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.189e+02 5.218e+02 6.714e+02 8.049e+02 1.427e+03, threshold=1.343e+03, percent-clipped=1.0 2023-04-27 12:51:12,701 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9073, 4.8885, 4.6501, 4.7731, 4.6896, 4.9623, 4.9571, 4.5926], device='cuda:5'), covar=tensor([0.0451, 0.0656, 0.0797, 0.0841, 0.0938, 0.0534, 0.0634, 0.1138], device='cuda:5'), in_proj_covar=tensor([0.0104, 0.0125, 0.0120, 0.0131, 0.0136, 0.0106, 0.0097, 0.0131], device='cuda:5'), out_proj_covar=tensor([9.0956e-05, 1.1464e-04, 1.0406e-04, 1.1460e-04, 1.2820e-04, 9.4287e-05, 8.7169e-05, 1.2278e-04], device='cuda:5') 2023-04-27 12:51:20,322 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 2023-04-27 12:51:28,101 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5311, 4.8299, 4.6439, 4.9934, 4.7548, 4.8513, 4.7279, 4.7074], device='cuda:5'), covar=tensor([0.0542, 0.0596, 0.0520, 0.0214, 0.0425, 0.0364, 0.0389, 0.0470], device='cuda:5'), in_proj_covar=tensor([0.0084, 0.0094, 0.0092, 0.0074, 0.0088, 0.0080, 0.0081, 0.0074], device='cuda:5'), out_proj_covar=tensor([6.4842e-05, 8.3184e-05, 7.6028e-05, 5.0246e-05, 6.5218e-05, 5.9369e-05, 6.1652e-05, 5.7764e-05], device='cuda:5') 2023-04-27 12:51:28,769 INFO [train.py:904] (5/8) Epoch 1, batch 1750, loss[loss=0.5043, simple_loss=0.4844, pruned_loss=0.2612, over 16443.00 frames. ], tot_loss[loss=0.501, simple_loss=0.4809, pruned_loss=0.2623, over 3311768.96 frames. ], batch size: 146, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:52:15,513 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:52:27,935 INFO [train.py:904] (5/8) Epoch 1, batch 1800, loss[loss=0.5025, simple_loss=0.4928, pruned_loss=0.2544, over 16438.00 frames. ], tot_loss[loss=0.4948, simple_loss=0.4785, pruned_loss=0.2563, over 3293345.05 frames. ], batch size: 146, lr: 4.85e-02, grad_scale: 4.0 2023-04-27 12:52:37,336 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1809.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:52:38,901 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.321e+02 5.485e+02 6.556e+02 7.752e+02 2.000e+03, threshold=1.311e+03, percent-clipped=5.0 2023-04-27 12:53:04,850 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2597, 3.1774, 2.9030, 2.9268, 3.2723, 2.7798, 3.0327, 2.9646], device='cuda:5'), covar=tensor([0.0559, 0.0550, 0.0846, 0.0644, 0.0523, 0.1121, 0.1149, 0.0691], device='cuda:5'), in_proj_covar=tensor([0.0068, 0.0066, 0.0063, 0.0065, 0.0068, 0.0071, 0.0074, 0.0067], device='cuda:5'), out_proj_covar=tensor([6.6190e-05, 6.1787e-05, 6.2516e-05, 6.1402e-05, 6.3971e-05, 6.8622e-05, 7.0241e-05, 6.2619e-05], device='cuda:5') 2023-04-27 12:53:11,048 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1838.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:53:25,100 INFO [train.py:904] (5/8) Epoch 1, batch 1850, loss[loss=0.4528, simple_loss=0.4513, pruned_loss=0.2255, over 16721.00 frames. ], tot_loss[loss=0.4871, simple_loss=0.4747, pruned_loss=0.2498, over 3298990.66 frames. ], batch size: 124, lr: 4.84e-02, grad_scale: 4.0 2023-04-27 12:53:32,841 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:53:32,949 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:53:48,642 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:54:22,034 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:54:23,728 INFO [train.py:904] (5/8) Epoch 1, batch 1900, loss[loss=0.4595, simple_loss=0.4552, pruned_loss=0.2309, over 16663.00 frames. ], tot_loss[loss=0.4753, simple_loss=0.4683, pruned_loss=0.2408, over 3298730.61 frames. ], batch size: 134, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:54:24,649 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2344, 5.4747, 5.2777, 5.5827, 5.3324, 5.4603, 5.2644, 5.3263], device='cuda:5'), covar=tensor([0.0258, 0.0322, 0.0311, 0.0129, 0.0238, 0.0214, 0.0222, 0.0207], device='cuda:5'), in_proj_covar=tensor([0.0093, 0.0101, 0.0099, 0.0081, 0.0096, 0.0084, 0.0088, 0.0079], device='cuda:5'), out_proj_covar=tensor([7.2668e-05, 8.9065e-05, 8.2192e-05, 5.5693e-05, 7.2530e-05, 6.4894e-05, 6.7213e-05, 6.2886e-05], device='cuda:5') 2023-04-27 12:54:28,419 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 12:54:36,244 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.249e+02 5.423e+02 6.685e+02 8.506e+02 1.861e+03, threshold=1.337e+03, percent-clipped=8.0 2023-04-27 12:54:44,562 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:54:49,137 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:55:00,056 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:55:22,826 INFO [train.py:904] (5/8) Epoch 1, batch 1950, loss[loss=0.3806, simple_loss=0.4149, pruned_loss=0.1719, over 16814.00 frames. ], tot_loss[loss=0.467, simple_loss=0.4645, pruned_loss=0.2341, over 3303384.04 frames. ], batch size: 42, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:55:42,115 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:56:23,574 INFO [train.py:904] (5/8) Epoch 1, batch 2000, loss[loss=0.5127, simple_loss=0.4997, pruned_loss=0.2628, over 15440.00 frames. ], tot_loss[loss=0.4606, simple_loss=0.4613, pruned_loss=0.2294, over 3301198.93 frames. ], batch size: 191, lr: 4.82e-02, grad_scale: 8.0 2023-04-27 12:56:36,807 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.491e+02 5.646e+02 7.242e+02 8.659e+02 1.834e+03, threshold=1.448e+03, percent-clipped=2.0 2023-04-27 12:57:21,762 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6360, 3.9665, 3.8868, 4.1950, 3.9390, 1.8106, 3.9197, 4.0424], device='cuda:5'), covar=tensor([0.3640, 0.1524, 0.2224, 0.0758, 0.2998, 0.9162, 0.0940, 0.0416], device='cuda:5'), in_proj_covar=tensor([0.0047, 0.0031, 0.0048, 0.0035, 0.0028, 0.0061, 0.0033, 0.0023], device='cuda:5'), out_proj_covar=tensor([4.4353e-05, 2.8237e-05, 4.3037e-05, 2.6833e-05, 2.7034e-05, 5.3311e-05, 2.6143e-05, 2.2071e-05], device='cuda:5') 2023-04-27 12:57:27,607 INFO [train.py:904] (5/8) Epoch 1, batch 2050, loss[loss=0.4213, simple_loss=0.4485, pruned_loss=0.197, over 16427.00 frames. ], tot_loss[loss=0.4487, simple_loss=0.4544, pruned_loss=0.2211, over 3309767.49 frames. ], batch size: 68, lr: 4.81e-02, grad_scale: 8.0 2023-04-27 12:58:18,983 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:58:32,234 INFO [train.py:904] (5/8) Epoch 1, batch 2100, loss[loss=0.3425, simple_loss=0.3867, pruned_loss=0.1492, over 16959.00 frames. ], tot_loss[loss=0.4401, simple_loss=0.4499, pruned_loss=0.2148, over 3309633.10 frames. ], batch size: 41, lr: 4.80e-02, grad_scale: 16.0 2023-04-27 12:58:45,935 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.641e+02 4.108e+02 4.910e+02 5.858e+02 1.001e+03, threshold=9.819e+02, percent-clipped=0.0 2023-04-27 12:59:01,407 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1785, 2.9050, 2.4876, 2.5566, 2.8267, 1.8717, 2.5857, 3.0619], device='cuda:5'), covar=tensor([0.2976, 0.1043, 0.1959, 0.0589, 0.1255, 0.4546, 0.0814, 0.0456], device='cuda:5'), in_proj_covar=tensor([0.0053, 0.0032, 0.0051, 0.0036, 0.0030, 0.0066, 0.0033, 0.0024], device='cuda:5'), out_proj_covar=tensor([5.0682e-05, 3.1274e-05, 4.6555e-05, 2.7408e-05, 2.9471e-05, 5.7410e-05, 2.6972e-05, 2.2366e-05], device='cuda:5') 2023-04-27 12:59:02,449 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3764, 4.3923, 4.2466, 4.6225, 4.4569, 4.6738, 4.5220, 4.4876], device='cuda:5'), covar=tensor([0.0246, 0.0264, 0.0522, 0.0310, 0.0343, 0.0202, 0.0288, 0.0299], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0096, 0.0130, 0.0104, 0.0097, 0.0100, 0.0089, 0.0101], device='cuda:5'), out_proj_covar=tensor([9.5537e-05, 9.6530e-05, 1.3840e-04, 1.1263e-04, 1.0224e-04, 1.0051e-04, 8.9061e-05, 1.0555e-04], device='cuda:5') 2023-04-27 12:59:20,452 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2138.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:59:35,893 INFO [train.py:904] (5/8) Epoch 1, batch 2150, loss[loss=0.4329, simple_loss=0.4401, pruned_loss=0.2128, over 16458.00 frames. ], tot_loss[loss=0.4361, simple_loss=0.448, pruned_loss=0.2119, over 3312652.05 frames. ], batch size: 146, lr: 4.79e-02, grad_scale: 16.0 2023-04-27 12:59:39,074 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0763, 3.9629, 3.8877, 4.3164, 3.9812, 4.2239, 4.2381, 4.0540], device='cuda:5'), covar=tensor([0.0502, 0.0578, 0.0364, 0.0266, 0.0682, 0.0228, 0.0295, 0.0411], device='cuda:5'), in_proj_covar=tensor([0.0061, 0.0056, 0.0072, 0.0060, 0.0065, 0.0059, 0.0064, 0.0064], device='cuda:5'), out_proj_covar=tensor([5.2543e-05, 4.8368e-05, 6.1574e-05, 4.9852e-05, 5.9225e-05, 4.8208e-05, 5.8736e-05, 5.6010e-05], device='cuda:5') 2023-04-27 13:00:32,070 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:00:40,262 INFO [train.py:904] (5/8) Epoch 1, batch 2200, loss[loss=0.4103, simple_loss=0.4202, pruned_loss=0.2002, over 16871.00 frames. ], tot_loss[loss=0.4301, simple_loss=0.4444, pruned_loss=0.2077, over 3301837.20 frames. ], batch size: 116, lr: 4.78e-02, grad_scale: 16.0 2023-04-27 13:00:53,199 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.805e+02 5.075e+02 6.122e+02 8.359e+02 1.658e+03, threshold=1.224e+03, percent-clipped=12.0 2023-04-27 13:00:56,712 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:01,845 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1936, 4.7453, 5.1953, 5.2621, 4.5171, 5.0393, 5.3140, 4.6677], device='cuda:5'), covar=tensor([0.0353, 0.0475, 0.0233, 0.0118, 0.0815, 0.0343, 0.0207, 0.0370], device='cuda:5'), in_proj_covar=tensor([0.0078, 0.0071, 0.0094, 0.0072, 0.0098, 0.0081, 0.0072, 0.0078], device='cuda:5'), out_proj_covar=tensor([7.8708e-05, 6.5008e-05, 9.8145e-05, 6.8414e-05, 1.0662e-04, 7.8995e-05, 6.9341e-05, 7.9130e-05], device='cuda:5') 2023-04-27 13:01:07,619 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:01:11,883 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:44,914 INFO [train.py:904] (5/8) Epoch 1, batch 2250, loss[loss=0.3508, simple_loss=0.3889, pruned_loss=0.1564, over 16991.00 frames. ], tot_loss[loss=0.4231, simple_loss=0.4405, pruned_loss=0.2027, over 3303928.59 frames. ], batch size: 41, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:01:49,123 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3219, 4.5560, 3.9600, 4.0909, 4.2035, 4.2996, 4.2927, 4.2078], device='cuda:5'), covar=tensor([0.0246, 0.0231, 0.0517, 0.0400, 0.0376, 0.0218, 0.0615, 0.0245], device='cuda:5'), in_proj_covar=tensor([0.0053, 0.0052, 0.0049, 0.0053, 0.0054, 0.0055, 0.0060, 0.0053], device='cuda:5'), out_proj_covar=tensor([4.9783e-05, 5.0046e-05, 4.7027e-05, 5.0077e-05, 4.8559e-05, 5.4369e-05, 5.9271e-05, 5.1035e-05], device='cuda:5') 2023-04-27 13:02:05,710 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:02:08,864 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:02:49,157 INFO [train.py:904] (5/8) Epoch 1, batch 2300, loss[loss=0.4166, simple_loss=0.4364, pruned_loss=0.1984, over 16773.00 frames. ], tot_loss[loss=0.4168, simple_loss=0.4368, pruned_loss=0.1983, over 3305697.08 frames. ], batch size: 134, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:03:01,867 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.985e+02 4.330e+02 5.680e+02 7.295e+02 1.284e+03, threshold=1.136e+03, percent-clipped=1.0 2023-04-27 13:03:03,726 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 13:03:07,886 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:39,730 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:03:53,139 INFO [train.py:904] (5/8) Epoch 1, batch 2350, loss[loss=0.4979, simple_loss=0.4986, pruned_loss=0.2486, over 12199.00 frames. ], tot_loss[loss=0.4126, simple_loss=0.4345, pruned_loss=0.1952, over 3307090.13 frames. ], batch size: 246, lr: 4.76e-02, grad_scale: 16.0 2023-04-27 13:04:54,415 INFO [train.py:904] (5/8) Epoch 1, batch 2400, loss[loss=0.3878, simple_loss=0.4373, pruned_loss=0.1692, over 16699.00 frames. ], tot_loss[loss=0.4087, simple_loss=0.4328, pruned_loss=0.1922, over 3316122.68 frames. ], batch size: 62, lr: 4.75e-02, grad_scale: 16.0 2023-04-27 13:04:55,602 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:05:07,298 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.161e+02 4.808e+02 6.214e+02 8.409e+02 1.581e+03, threshold=1.243e+03, percent-clipped=3.0 2023-04-27 13:05:36,372 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7239, 4.0805, 3.7422, 3.6809, 3.9481, 3.9360, 3.8206, 3.7886], device='cuda:5'), covar=tensor([0.0301, 0.0320, 0.0370, 0.0270, 0.0392, 0.0300, 0.0543, 0.0328], device='cuda:5'), in_proj_covar=tensor([0.0051, 0.0050, 0.0045, 0.0048, 0.0050, 0.0054, 0.0058, 0.0052], device='cuda:5'), out_proj_covar=tensor([5.0204e-05, 5.0136e-05, 4.4587e-05, 4.7608e-05, 4.6613e-05, 5.7007e-05, 5.9021e-05, 5.1320e-05], device='cuda:5') 2023-04-27 13:05:55,901 INFO [train.py:904] (5/8) Epoch 1, batch 2450, loss[loss=0.3567, simple_loss=0.401, pruned_loss=0.1562, over 16849.00 frames. ], tot_loss[loss=0.4046, simple_loss=0.431, pruned_loss=0.189, over 3316276.19 frames. ], batch size: 42, lr: 4.74e-02, grad_scale: 16.0 2023-04-27 13:06:51,203 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:06:59,007 INFO [train.py:904] (5/8) Epoch 1, batch 2500, loss[loss=0.4423, simple_loss=0.4571, pruned_loss=0.2137, over 16250.00 frames. ], tot_loss[loss=0.3992, simple_loss=0.4282, pruned_loss=0.1851, over 3321818.20 frames. ], batch size: 165, lr: 4.73e-02, grad_scale: 16.0 2023-04-27 13:07:11,503 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.680e+02 4.365e+02 5.111e+02 7.039e+02 1.163e+03, threshold=1.022e+03, percent-clipped=0.0 2023-04-27 13:07:14,788 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:07:29,212 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:07:50,931 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:01,510 INFO [train.py:904] (5/8) Epoch 1, batch 2550, loss[loss=0.4176, simple_loss=0.4322, pruned_loss=0.2015, over 16436.00 frames. ], tot_loss[loss=0.3956, simple_loss=0.4259, pruned_loss=0.1826, over 3327764.37 frames. ], batch size: 146, lr: 4.72e-02, grad_scale: 16.0 2023-04-27 13:08:15,981 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:32,319 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:09:07,697 INFO [train.py:904] (5/8) Epoch 1, batch 2600, loss[loss=0.3432, simple_loss=0.3954, pruned_loss=0.1455, over 17209.00 frames. ], tot_loss[loss=0.3896, simple_loss=0.4224, pruned_loss=0.1784, over 3332479.17 frames. ], batch size: 45, lr: 4.71e-02, grad_scale: 16.0 2023-04-27 13:09:20,092 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.935e+02 4.794e+02 6.148e+02 7.560e+02 1.171e+03, threshold=1.230e+03, percent-clipped=4.0 2023-04-27 13:10:11,897 INFO [train.py:904] (5/8) Epoch 1, batch 2650, loss[loss=0.3724, simple_loss=0.4251, pruned_loss=0.1599, over 17070.00 frames. ], tot_loss[loss=0.3855, simple_loss=0.4208, pruned_loss=0.1751, over 3334688.81 frames. ], batch size: 50, lr: 4.70e-02, grad_scale: 16.0 2023-04-27 13:10:38,148 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7754, 3.8027, 3.8203, 3.8683, 4.1810, 3.8477, 3.6243, 4.0616], device='cuda:5'), covar=tensor([0.0522, 0.0314, 0.0698, 0.0569, 0.0367, 0.0454, 0.0617, 0.0321], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0097, 0.0123, 0.0119, 0.0111, 0.0107, 0.0110, 0.0092], device='cuda:5'), out_proj_covar=tensor([1.1381e-04, 1.1668e-04, 1.3723e-04, 1.2427e-04, 1.2282e-04, 1.2010e-04, 1.1487e-04, 9.2200e-05], device='cuda:5') 2023-04-27 13:10:55,288 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:11:09,982 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:11:11,304 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 13:11:15,054 INFO [train.py:904] (5/8) Epoch 1, batch 2700, loss[loss=0.3698, simple_loss=0.42, pruned_loss=0.1599, over 17032.00 frames. ], tot_loss[loss=0.383, simple_loss=0.4193, pruned_loss=0.1733, over 3330403.09 frames. ], batch size: 53, lr: 4.69e-02, grad_scale: 16.0 2023-04-27 13:11:29,646 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.771e+02 4.333e+02 5.264e+02 6.238e+02 1.242e+03, threshold=1.053e+03, percent-clipped=1.0 2023-04-27 13:12:14,253 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:12:20,234 INFO [train.py:904] (5/8) Epoch 1, batch 2750, loss[loss=0.3786, simple_loss=0.4087, pruned_loss=0.1742, over 16447.00 frames. ], tot_loss[loss=0.379, simple_loss=0.4167, pruned_loss=0.1706, over 3334863.57 frames. ], batch size: 146, lr: 4.68e-02, grad_scale: 16.0 2023-04-27 13:13:23,786 INFO [train.py:904] (5/8) Epoch 1, batch 2800, loss[loss=0.383, simple_loss=0.4081, pruned_loss=0.179, over 16858.00 frames. ], tot_loss[loss=0.3783, simple_loss=0.4163, pruned_loss=0.1702, over 3328661.71 frames. ], batch size: 102, lr: 4.67e-02, grad_scale: 16.0 2023-04-27 13:13:35,362 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.742e+02 4.329e+02 5.433e+02 6.254e+02 2.106e+03, threshold=1.087e+03, percent-clipped=5.0 2023-04-27 13:13:40,848 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0972, 2.3848, 2.1173, 1.4553, 2.1688, 2.3276, 2.4046, 2.1027], device='cuda:5'), covar=tensor([0.0212, 0.0209, 0.0278, 0.0399, 0.0212, 0.0148, 0.0170, 0.0264], device='cuda:5'), in_proj_covar=tensor([0.0030, 0.0031, 0.0036, 0.0036, 0.0033, 0.0035, 0.0036, 0.0032], device='cuda:5'), out_proj_covar=tensor([3.6519e-05, 3.3258e-05, 3.9615e-05, 3.6735e-05, 3.5748e-05, 3.7403e-05, 3.6191e-05, 3.6608e-05], device='cuda:5') 2023-04-27 13:13:54,304 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:14:26,001 INFO [train.py:904] (5/8) Epoch 1, batch 2850, loss[loss=0.3761, simple_loss=0.3967, pruned_loss=0.1778, over 16842.00 frames. ], tot_loss[loss=0.3752, simple_loss=0.4135, pruned_loss=0.1685, over 3318458.78 frames. ], batch size: 90, lr: 4.66e-02, grad_scale: 16.0 2023-04-27 13:15:09,614 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:15:27,296 INFO [train.py:904] (5/8) Epoch 1, batch 2900, loss[loss=0.3523, simple_loss=0.3975, pruned_loss=0.1536, over 17136.00 frames. ], tot_loss[loss=0.3739, simple_loss=0.411, pruned_loss=0.1684, over 3323337.91 frames. ], batch size: 47, lr: 4.65e-02, grad_scale: 16.0 2023-04-27 13:15:40,742 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 5.015e+02 6.760e+02 8.832e+02 1.641e+03, threshold=1.352e+03, percent-clipped=13.0 2023-04-27 13:16:13,952 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:16:19,903 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8755, 4.8332, 5.1349, 5.2014, 5.4701, 5.1141, 4.8985, 5.2528], device='cuda:5'), covar=tensor([0.0320, 0.0210, 0.0447, 0.0355, 0.0297, 0.0234, 0.0444, 0.0221], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0100, 0.0129, 0.0126, 0.0123, 0.0111, 0.0117, 0.0097], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:5') 2023-04-27 13:16:32,283 INFO [train.py:904] (5/8) Epoch 1, batch 2950, loss[loss=0.3855, simple_loss=0.4076, pruned_loss=0.1817, over 16877.00 frames. ], tot_loss[loss=0.3707, simple_loss=0.4077, pruned_loss=0.1669, over 3319585.71 frames. ], batch size: 90, lr: 4.64e-02, grad_scale: 16.0 2023-04-27 13:17:29,924 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:17:32,235 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:17:35,319 INFO [train.py:904] (5/8) Epoch 1, batch 3000, loss[loss=0.3367, simple_loss=0.3852, pruned_loss=0.1441, over 15925.00 frames. ], tot_loss[loss=0.3679, simple_loss=0.4059, pruned_loss=0.165, over 3323771.88 frames. ], batch size: 35, lr: 4.63e-02, grad_scale: 16.0 2023-04-27 13:17:35,319 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 13:17:45,053 INFO [train.py:938] (5/8) Epoch 1, validation: loss=0.2847, simple_loss=0.3895, pruned_loss=0.08992, over 944034.00 frames. 2023-04-27 13:17:45,054 INFO [train.py:939] (5/8) Maximum memory allocated so far is 15524MB 2023-04-27 13:17:50,021 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7279, 4.6977, 4.0920, 4.2852, 4.6468, 4.7562, 4.2953, 4.6982], device='cuda:5'), covar=tensor([0.0174, 0.0141, 0.0188, 0.0304, 0.0108, 0.0133, 0.0144, 0.0146], device='cuda:5'), in_proj_covar=tensor([0.0052, 0.0047, 0.0067, 0.0066, 0.0047, 0.0054, 0.0063, 0.0058], device='cuda:5'), out_proj_covar=tensor([6.2764e-05, 5.4799e-05, 9.4898e-05, 8.2595e-05, 5.0255e-05, 6.0856e-05, 8.1823e-05, 7.5019e-05], device='cuda:5') 2023-04-27 13:17:59,843 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.008e+02 4.355e+02 5.212e+02 6.501e+02 1.071e+03, threshold=1.042e+03, percent-clipped=0.0 2023-04-27 13:18:09,307 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:18:09,577 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 2023-04-27 13:18:36,922 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:18:41,575 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:18:50,100 INFO [train.py:904] (5/8) Epoch 1, batch 3050, loss[loss=0.3665, simple_loss=0.409, pruned_loss=0.162, over 16520.00 frames. ], tot_loss[loss=0.3655, simple_loss=0.4042, pruned_loss=0.1634, over 3323232.63 frames. ], batch size: 75, lr: 4.62e-02, grad_scale: 16.0 2023-04-27 13:19:21,341 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 13:19:27,668 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 13:19:53,991 INFO [train.py:904] (5/8) Epoch 1, batch 3100, loss[loss=0.3932, simple_loss=0.4307, pruned_loss=0.1779, over 16646.00 frames. ], tot_loss[loss=0.3636, simple_loss=0.4027, pruned_loss=0.1622, over 3320147.84 frames. ], batch size: 62, lr: 4.61e-02, grad_scale: 16.0 2023-04-27 13:20:07,654 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.721e+02 4.392e+02 5.238e+02 7.570e+02 1.450e+03, threshold=1.048e+03, percent-clipped=8.0 2023-04-27 13:20:30,227 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3605, 3.5852, 3.4576, 3.2779, 3.4529, 3.5090, 3.6465, 3.7510], device='cuda:5'), covar=tensor([0.0256, 0.0211, 0.0244, 0.0241, 0.0235, 0.0279, 0.0237, 0.0148], device='cuda:5'), in_proj_covar=tensor([0.0044, 0.0036, 0.0036, 0.0041, 0.0038, 0.0041, 0.0044, 0.0039], device='cuda:5'), out_proj_covar=tensor([5.2229e-05, 4.5634e-05, 4.4609e-05, 4.7064e-05, 4.5514e-05, 5.6920e-05, 5.2673e-05, 4.6432e-05], device='cuda:5') 2023-04-27 13:21:00,210 INFO [train.py:904] (5/8) Epoch 1, batch 3150, loss[loss=0.4336, simple_loss=0.4426, pruned_loss=0.2123, over 12325.00 frames. ], tot_loss[loss=0.3613, simple_loss=0.4006, pruned_loss=0.161, over 3302877.09 frames. ], batch size: 246, lr: 4.60e-02, grad_scale: 16.0 2023-04-27 13:21:28,238 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 13:21:38,543 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3180.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:22:04,623 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9970, 4.8165, 5.2275, 5.2476, 5.4599, 5.1900, 5.0409, 5.2742], device='cuda:5'), covar=tensor([0.0270, 0.0221, 0.0445, 0.0354, 0.0241, 0.0237, 0.0355, 0.0157], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0106, 0.0136, 0.0135, 0.0137, 0.0119, 0.0129, 0.0101], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:5') 2023-04-27 13:22:05,306 INFO [train.py:904] (5/8) Epoch 1, batch 3200, loss[loss=0.3028, simple_loss=0.3531, pruned_loss=0.1263, over 17210.00 frames. ], tot_loss[loss=0.3582, simple_loss=0.3985, pruned_loss=0.1589, over 3310839.50 frames. ], batch size: 44, lr: 4.59e-02, grad_scale: 16.0 2023-04-27 13:22:12,998 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.40 vs. limit=5.0 2023-04-27 13:22:17,349 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.273e+02 4.612e+02 6.158e+02 7.733e+02 1.150e+03, threshold=1.232e+03, percent-clipped=3.0 2023-04-27 13:23:07,322 INFO [train.py:904] (5/8) Epoch 1, batch 3250, loss[loss=0.3342, simple_loss=0.3892, pruned_loss=0.1396, over 17206.00 frames. ], tot_loss[loss=0.3569, simple_loss=0.3972, pruned_loss=0.1583, over 3321471.80 frames. ], batch size: 46, lr: 4.58e-02, grad_scale: 16.0 2023-04-27 13:23:56,345 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:24:02,466 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:12,259 INFO [train.py:904] (5/8) Epoch 1, batch 3300, loss[loss=0.3453, simple_loss=0.3889, pruned_loss=0.1509, over 16874.00 frames. ], tot_loss[loss=0.3562, simple_loss=0.3978, pruned_loss=0.1573, over 3316037.33 frames. ], batch size: 102, lr: 4.57e-02, grad_scale: 16.0 2023-04-27 13:24:25,126 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.531e+02 4.292e+02 5.268e+02 6.867e+02 1.392e+03, threshold=1.054e+03, percent-clipped=2.0 2023-04-27 13:25:03,548 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:25:06,932 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.50 vs. limit=5.0 2023-04-27 13:25:16,158 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:25:17,769 INFO [train.py:904] (5/8) Epoch 1, batch 3350, loss[loss=0.3242, simple_loss=0.3809, pruned_loss=0.1338, over 17209.00 frames. ], tot_loss[loss=0.3527, simple_loss=0.3963, pruned_loss=0.1545, over 3324718.92 frames. ], batch size: 45, lr: 4.56e-02, grad_scale: 16.0 2023-04-27 13:25:38,597 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3967, 3.9417, 3.4034, 4.5878, 3.7310, 2.0750, 4.4620, 4.0584], device='cuda:5'), covar=tensor([0.3271, 0.0801, 0.1770, 0.0249, 0.2062, 0.3520, 0.0260, 0.0140], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0065, 0.0099, 0.0053, 0.0061, 0.0100, 0.0053, 0.0030], device='cuda:5'), out_proj_covar=tensor([1.1713e-04, 7.2789e-05, 9.7687e-05, 4.9791e-05, 7.4291e-05, 9.6395e-05, 5.3114e-05, 3.4017e-05], device='cuda:5') 2023-04-27 13:25:50,437 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:25:58,437 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 13:26:09,261 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:26:24,827 INFO [train.py:904] (5/8) Epoch 1, batch 3400, loss[loss=0.3846, simple_loss=0.4165, pruned_loss=0.1764, over 15550.00 frames. ], tot_loss[loss=0.3505, simple_loss=0.395, pruned_loss=0.153, over 3331595.98 frames. ], batch size: 191, lr: 4.55e-02, grad_scale: 16.0 2023-04-27 13:26:39,221 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.681e+02 4.195e+02 5.225e+02 6.801e+02 1.040e+03, threshold=1.045e+03, percent-clipped=0.0 2023-04-27 13:26:43,289 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-04-27 13:27:32,317 INFO [train.py:904] (5/8) Epoch 1, batch 3450, loss[loss=0.3059, simple_loss=0.3545, pruned_loss=0.1286, over 16969.00 frames. ], tot_loss[loss=0.3469, simple_loss=0.392, pruned_loss=0.1509, over 3325779.24 frames. ], batch size: 41, lr: 4.54e-02, grad_scale: 16.0 2023-04-27 13:27:33,542 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5223, 4.4271, 4.2434, 2.3335, 3.7844, 4.2948, 4.2441, 4.1961], device='cuda:5'), covar=tensor([0.0104, 0.0131, 0.0169, 0.1314, 0.0401, 0.0150, 0.0076, 0.0148], device='cuda:5'), in_proj_covar=tensor([0.0039, 0.0043, 0.0048, 0.0077, 0.0042, 0.0041, 0.0040, 0.0046], device='cuda:5'), out_proj_covar=tensor([4.5070e-05, 4.9114e-05, 5.6009e-05, 8.7971e-05, 5.4764e-05, 4.7509e-05, 5.3685e-05, 5.1968e-05], device='cuda:5') 2023-04-27 13:28:12,596 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:28:33,992 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5554, 3.7192, 3.6388, 3.5263, 3.6708, 3.9226, 3.9140, 4.0426], device='cuda:5'), covar=tensor([0.0252, 0.0222, 0.0177, 0.0187, 0.0232, 0.0179, 0.0213, 0.0144], device='cuda:5'), in_proj_covar=tensor([0.0045, 0.0034, 0.0033, 0.0043, 0.0036, 0.0039, 0.0044, 0.0039], device='cuda:5'), out_proj_covar=tensor([5.9832e-05, 4.7256e-05, 4.5581e-05, 5.5068e-05, 4.7966e-05, 6.1994e-05, 5.7875e-05, 5.1722e-05], device='cuda:5') 2023-04-27 13:28:39,716 INFO [train.py:904] (5/8) Epoch 1, batch 3500, loss[loss=0.3291, simple_loss=0.3941, pruned_loss=0.132, over 17095.00 frames. ], tot_loss[loss=0.3428, simple_loss=0.3893, pruned_loss=0.1482, over 3315669.95 frames. ], batch size: 53, lr: 4.53e-02, grad_scale: 16.0 2023-04-27 13:28:53,817 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 4.599e+02 5.593e+02 7.423e+02 2.273e+03, threshold=1.119e+03, percent-clipped=10.0 2023-04-27 13:29:16,290 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:29:29,658 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-27 13:29:39,461 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-27 13:29:40,147 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:29:46,556 INFO [train.py:904] (5/8) Epoch 1, batch 3550, loss[loss=0.3483, simple_loss=0.4102, pruned_loss=0.1432, over 16740.00 frames. ], tot_loss[loss=0.3414, simple_loss=0.388, pruned_loss=0.1474, over 3324484.71 frames. ], batch size: 57, lr: 4.51e-02, grad_scale: 16.0 2023-04-27 13:30:44,882 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:30:54,499 INFO [train.py:904] (5/8) Epoch 1, batch 3600, loss[loss=0.2862, simple_loss=0.3491, pruned_loss=0.1117, over 17232.00 frames. ], tot_loss[loss=0.3371, simple_loss=0.3842, pruned_loss=0.145, over 3319038.18 frames. ], batch size: 45, lr: 4.50e-02, grad_scale: 16.0 2023-04-27 13:31:04,197 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:31:08,709 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.446e+02 5.845e+02 7.592e+02 1.096e+03 2.095e+03, threshold=1.518e+03, percent-clipped=22.0 2023-04-27 13:31:52,476 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:31:56,494 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:32:06,055 INFO [train.py:904] (5/8) Epoch 1, batch 3650, loss[loss=0.3276, simple_loss=0.367, pruned_loss=0.1441, over 16473.00 frames. ], tot_loss[loss=0.3363, simple_loss=0.3826, pruned_loss=0.145, over 3309774.21 frames. ], batch size: 68, lr: 4.49e-02, grad_scale: 16.0 2023-04-27 13:32:28,401 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0129, 5.1172, 5.2779, 5.4360, 5.6410, 5.2701, 5.1283, 5.3121], device='cuda:5'), covar=tensor([0.0276, 0.0156, 0.0497, 0.0413, 0.0250, 0.0197, 0.0465, 0.0192], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0103, 0.0131, 0.0133, 0.0138, 0.0114, 0.0126, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:5') 2023-04-27 13:32:42,329 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:32:43,612 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:33:04,276 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0400, 2.3862, 2.3052, 1.9997, 2.4221, 2.4103, 2.4171, 1.9081], device='cuda:5'), covar=tensor([0.0231, 0.0323, 0.0288, 0.0269, 0.0170, 0.0232, 0.0183, 0.0209], device='cuda:5'), in_proj_covar=tensor([0.0027, 0.0039, 0.0036, 0.0035, 0.0031, 0.0033, 0.0034, 0.0032], device='cuda:5'), out_proj_covar=tensor([3.9932e-05, 5.3504e-05, 5.0218e-05, 4.1679e-05, 4.0716e-05, 4.2980e-05, 4.0887e-05, 4.1862e-05], device='cuda:5') 2023-04-27 13:33:20,378 INFO [train.py:904] (5/8) Epoch 1, batch 3700, loss[loss=0.3621, simple_loss=0.3884, pruned_loss=0.1679, over 11350.00 frames. ], tot_loss[loss=0.3351, simple_loss=0.3792, pruned_loss=0.1455, over 3278794.71 frames. ], batch size: 246, lr: 4.48e-02, grad_scale: 16.0 2023-04-27 13:33:35,074 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.010e+02 4.933e+02 6.416e+02 8.029e+02 1.327e+03, threshold=1.283e+03, percent-clipped=0.0 2023-04-27 13:33:54,092 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:34:15,097 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:34:17,715 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5015, 3.0001, 2.5214, 2.9076, 2.5511, 1.8038, 3.0400, 3.0089], device='cuda:5'), covar=tensor([0.2373, 0.1014, 0.1834, 0.0522, 0.1708, 0.2386, 0.0522, 0.0510], device='cuda:5'), in_proj_covar=tensor([0.0105, 0.0062, 0.0092, 0.0051, 0.0060, 0.0089, 0.0052, 0.0029], device='cuda:5'), out_proj_covar=tensor([1.0890e-04, 6.8778e-05, 8.9951e-05, 4.8245e-05, 7.3442e-05, 8.5674e-05, 5.1883e-05, 3.2088e-05], device='cuda:5') 2023-04-27 13:34:33,712 INFO [train.py:904] (5/8) Epoch 1, batch 3750, loss[loss=0.3334, simple_loss=0.3867, pruned_loss=0.14, over 17107.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3783, pruned_loss=0.1463, over 3272749.70 frames. ], batch size: 49, lr: 4.47e-02, grad_scale: 16.0 2023-04-27 13:35:15,353 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8621, 2.8112, 2.2475, 2.1505, 2.2431, 1.7862, 2.8007, 3.0080], device='cuda:5'), covar=tensor([0.0435, 0.0275, 0.0473, 0.1806, 0.1134, 0.1605, 0.0392, 0.0325], device='cuda:5'), in_proj_covar=tensor([0.0041, 0.0038, 0.0058, 0.0101, 0.0086, 0.0091, 0.0047, 0.0035], device='cuda:5'), out_proj_covar=tensor([5.6286e-05, 5.0163e-05, 6.2682e-05, 1.0643e-04, 9.1133e-05, 9.3577e-05, 5.4222e-05, 4.6682e-05], device='cuda:5') 2023-04-27 13:35:24,011 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5998, 4.7917, 4.4480, 4.6605, 4.7118, 4.9755, 4.8842, 4.5893], device='cuda:5'), covar=tensor([0.0603, 0.0672, 0.0753, 0.0905, 0.1342, 0.0499, 0.0657, 0.1435], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0177, 0.0142, 0.0154, 0.0183, 0.0133, 0.0137, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:5') 2023-04-27 13:35:34,006 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1470, 3.5749, 3.6215, 3.7409, 3.5021, 2.6336, 3.7651, 4.2776], device='cuda:5'), covar=tensor([0.2237, 0.0778, 0.1102, 0.0437, 0.0962, 0.1754, 0.0335, 0.0096], device='cuda:5'), in_proj_covar=tensor([0.0103, 0.0060, 0.0090, 0.0051, 0.0060, 0.0083, 0.0051, 0.0029], device='cuda:5'), out_proj_covar=tensor([1.0648e-04, 6.7043e-05, 8.7311e-05, 4.8022e-05, 7.2804e-05, 8.0577e-05, 5.0881e-05, 3.1675e-05], device='cuda:5') 2023-04-27 13:35:46,291 INFO [train.py:904] (5/8) Epoch 1, batch 3800, loss[loss=0.3451, simple_loss=0.3788, pruned_loss=0.1556, over 16911.00 frames. ], tot_loss[loss=0.3369, simple_loss=0.3783, pruned_loss=0.1477, over 3270445.27 frames. ], batch size: 109, lr: 4.46e-02, grad_scale: 16.0 2023-04-27 13:35:55,747 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8073, 5.2241, 4.9455, 5.0543, 5.1955, 5.5343, 5.4516, 5.1007], device='cuda:5'), covar=tensor([0.0604, 0.0793, 0.0678, 0.1233, 0.1574, 0.0552, 0.0569, 0.1241], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0175, 0.0139, 0.0152, 0.0182, 0.0132, 0.0135, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:5') 2023-04-27 13:36:00,506 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.355e+02 5.070e+02 6.208e+02 8.135e+02 1.675e+03, threshold=1.242e+03, percent-clipped=4.0 2023-04-27 13:36:57,575 INFO [train.py:904] (5/8) Epoch 1, batch 3850, loss[loss=0.2851, simple_loss=0.3477, pruned_loss=0.1112, over 16477.00 frames. ], tot_loss[loss=0.3332, simple_loss=0.3749, pruned_loss=0.1458, over 3283826.02 frames. ], batch size: 68, lr: 4.45e-02, grad_scale: 16.0 2023-04-27 13:38:09,807 INFO [train.py:904] (5/8) Epoch 1, batch 3900, loss[loss=0.3188, simple_loss=0.3583, pruned_loss=0.1396, over 15602.00 frames. ], tot_loss[loss=0.3316, simple_loss=0.3727, pruned_loss=0.1452, over 3274630.50 frames. ], batch size: 191, lr: 4.44e-02, grad_scale: 16.0 2023-04-27 13:38:11,846 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:38:24,739 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.787e+02 4.914e+02 5.725e+02 7.504e+02 1.784e+03, threshold=1.145e+03, percent-clipped=3.0 2023-04-27 13:38:46,443 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:58,308 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 13:39:11,924 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:39:21,379 INFO [train.py:904] (5/8) Epoch 1, batch 3950, loss[loss=0.3344, simple_loss=0.3702, pruned_loss=0.1493, over 16726.00 frames. ], tot_loss[loss=0.3306, simple_loss=0.3713, pruned_loss=0.145, over 3269732.87 frames. ], batch size: 134, lr: 4.43e-02, grad_scale: 16.0 2023-04-27 13:39:50,322 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5065, 3.0340, 2.5003, 2.9043, 2.3738, 1.8207, 3.0505, 3.2902], device='cuda:5'), covar=tensor([0.1961, 0.0939, 0.1455, 0.0553, 0.1842, 0.2046, 0.0402, 0.0342], device='cuda:5'), in_proj_covar=tensor([0.0121, 0.0072, 0.0106, 0.0059, 0.0076, 0.0095, 0.0058, 0.0033], device='cuda:5'), out_proj_covar=tensor([1.2956e-04, 8.1612e-05, 1.0438e-04, 5.7147e-05, 9.0935e-05, 9.4201e-05, 5.8121e-05, 3.6864e-05], device='cuda:5') 2023-04-27 13:40:13,993 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:20,466 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:40:37,502 INFO [train.py:904] (5/8) Epoch 1, batch 4000, loss[loss=0.3514, simple_loss=0.3941, pruned_loss=0.1544, over 15425.00 frames. ], tot_loss[loss=0.3271, simple_loss=0.3684, pruned_loss=0.1429, over 3278285.98 frames. ], batch size: 191, lr: 4.42e-02, grad_scale: 16.0 2023-04-27 13:40:49,694 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4578, 3.6780, 2.0585, 3.8143, 3.3736, 3.9139, 2.1060, 3.1017], device='cuda:5'), covar=tensor([0.0114, 0.0092, 0.0734, 0.0101, 0.0166, 0.0083, 0.0687, 0.0194], device='cuda:5'), in_proj_covar=tensor([0.0030, 0.0027, 0.0046, 0.0030, 0.0037, 0.0026, 0.0051, 0.0031], device='cuda:5'), out_proj_covar=tensor([3.5699e-05, 3.7052e-05, 5.9165e-05, 3.5728e-05, 4.4093e-05, 3.5651e-05, 5.9868e-05, 3.8928e-05], device='cuda:5') 2023-04-27 13:40:52,173 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.081e+02 4.378e+02 5.733e+02 7.371e+02 1.613e+03, threshold=1.147e+03, percent-clipped=6.0 2023-04-27 13:41:06,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8892, 3.5027, 3.6515, 3.3825, 3.4345, 3.7077, 3.6198, 3.6253], device='cuda:5'), covar=tensor([0.0342, 0.0147, 0.0091, 0.0127, 0.0122, 0.0087, 0.0146, 0.0138], device='cuda:5'), in_proj_covar=tensor([0.0050, 0.0030, 0.0031, 0.0041, 0.0032, 0.0034, 0.0039, 0.0037], device='cuda:5'), out_proj_covar=tensor([7.4580e-05, 4.8327e-05, 4.7559e-05, 5.7848e-05, 4.8241e-05, 5.8903e-05, 5.7093e-05, 5.5062e-05], device='cuda:5') 2023-04-27 13:41:24,010 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:41:39,024 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-27 13:41:50,491 INFO [train.py:904] (5/8) Epoch 1, batch 4050, loss[loss=0.2778, simple_loss=0.3414, pruned_loss=0.1071, over 16959.00 frames. ], tot_loss[loss=0.318, simple_loss=0.3634, pruned_loss=0.1363, over 3277740.92 frames. ], batch size: 55, lr: 4.41e-02, grad_scale: 16.0 2023-04-27 13:43:04,582 INFO [train.py:904] (5/8) Epoch 1, batch 4100, loss[loss=0.3216, simple_loss=0.3797, pruned_loss=0.1318, over 16721.00 frames. ], tot_loss[loss=0.3126, simple_loss=0.3612, pruned_loss=0.132, over 3285357.00 frames. ], batch size: 89, lr: 4.40e-02, grad_scale: 32.0 2023-04-27 13:43:18,965 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.292e+02 4.086e+02 5.427e+02 7.425e+02 1.337e+03, threshold=1.085e+03, percent-clipped=4.0 2023-04-27 13:43:40,492 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5341, 3.5499, 2.9697, 2.2286, 2.7557, 1.9962, 3.1870, 3.8584], device='cuda:5'), covar=tensor([0.0298, 0.0341, 0.0435, 0.1770, 0.0952, 0.1336, 0.0516, 0.0119], device='cuda:5'), in_proj_covar=tensor([0.0047, 0.0046, 0.0069, 0.0112, 0.0100, 0.0098, 0.0067, 0.0037], device='cuda:5'), out_proj_covar=tensor([6.8087e-05, 6.4450e-05, 8.0545e-05, 1.2293e-04, 1.1230e-04, 1.0650e-04, 8.5796e-05, 5.1489e-05], device='cuda:5') 2023-04-27 13:44:04,989 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:44:20,303 INFO [train.py:904] (5/8) Epoch 1, batch 4150, loss[loss=0.4075, simple_loss=0.4421, pruned_loss=0.1864, over 15182.00 frames. ], tot_loss[loss=0.3253, simple_loss=0.3724, pruned_loss=0.1391, over 3223897.81 frames. ], batch size: 190, lr: 4.39e-02, grad_scale: 32.0 2023-04-27 13:45:22,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9174, 1.3951, 1.4013, 1.6278, 1.7017, 1.6986, 1.4575, 1.3376], device='cuda:5'), covar=tensor([0.0086, 0.0284, 0.0166, 0.0123, 0.0107, 0.0095, 0.0192, 0.0198], device='cuda:5'), in_proj_covar=tensor([0.0022, 0.0036, 0.0027, 0.0027, 0.0022, 0.0026, 0.0026, 0.0029], device='cuda:5'), out_proj_covar=tensor([2.3933e-05, 3.9486e-05, 2.8424e-05, 2.7897e-05, 1.9523e-05, 2.2694e-05, 2.5030e-05, 2.7207e-05], device='cuda:5') 2023-04-27 13:45:37,094 INFO [train.py:904] (5/8) Epoch 1, batch 4200, loss[loss=0.3455, simple_loss=0.4034, pruned_loss=0.1438, over 16818.00 frames. ], tot_loss[loss=0.3339, simple_loss=0.3815, pruned_loss=0.1431, over 3189828.64 frames. ], batch size: 83, lr: 4.38e-02, grad_scale: 16.0 2023-04-27 13:45:39,661 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:45:39,739 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:45:44,377 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-04-27 13:45:52,580 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.534e+02 4.783e+02 6.571e+02 8.274e+02 1.863e+03, threshold=1.314e+03, percent-clipped=9.0 2023-04-27 13:46:16,409 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3246, 3.7899, 3.5876, 3.6081, 2.8711, 2.2054, 3.8031, 4.0958], device='cuda:5'), covar=tensor([0.1537, 0.0629, 0.0830, 0.0416, 0.1686, 0.1692, 0.0264, 0.0070], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0080, 0.0114, 0.0064, 0.0086, 0.0100, 0.0062, 0.0033], device='cuda:5'), out_proj_covar=tensor([1.4164e-04, 9.4276e-05, 1.1294e-04, 6.5024e-05, 1.0360e-04, 1.0198e-04, 6.5239e-05, 3.7696e-05], device='cuda:5') 2023-04-27 13:46:37,012 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0858, 4.0214, 4.3177, 4.4577, 4.6077, 4.0667, 4.2773, 4.3961], device='cuda:5'), covar=tensor([0.0238, 0.0198, 0.0332, 0.0293, 0.0212, 0.0221, 0.0325, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0088, 0.0108, 0.0107, 0.0116, 0.0099, 0.0109, 0.0087], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:5') 2023-04-27 13:46:49,190 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 13:46:50,186 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:46:50,871 INFO [train.py:904] (5/8) Epoch 1, batch 4250, loss[loss=0.3506, simple_loss=0.3843, pruned_loss=0.1584, over 12044.00 frames. ], tot_loss[loss=0.3329, simple_loss=0.3823, pruned_loss=0.1417, over 3178282.88 frames. ], batch size: 247, lr: 4.36e-02, grad_scale: 16.0 2023-04-27 13:47:37,205 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:44,052 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0339, 3.9865, 4.2982, 4.2320, 4.5200, 4.0382, 4.1103, 4.2534], device='cuda:5'), covar=tensor([0.0229, 0.0225, 0.0375, 0.0384, 0.0261, 0.0268, 0.0411, 0.0184], device='cuda:5'), in_proj_covar=tensor([0.0103, 0.0092, 0.0114, 0.0111, 0.0120, 0.0103, 0.0115, 0.0091], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:5') 2023-04-27 13:47:45,351 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:48:04,736 INFO [train.py:904] (5/8) Epoch 1, batch 4300, loss[loss=0.3908, simple_loss=0.4367, pruned_loss=0.1725, over 15370.00 frames. ], tot_loss[loss=0.3313, simple_loss=0.3832, pruned_loss=0.1397, over 3185882.40 frames. ], batch size: 190, lr: 4.35e-02, grad_scale: 16.0 2023-04-27 13:48:21,412 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.968e+02 4.766e+02 5.895e+02 7.865e+02 1.445e+03, threshold=1.179e+03, percent-clipped=2.0 2023-04-27 13:48:38,485 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 2023-04-27 13:48:52,485 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:49:17,724 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:20,292 INFO [train.py:904] (5/8) Epoch 1, batch 4350, loss[loss=0.3826, simple_loss=0.4198, pruned_loss=0.1727, over 11349.00 frames. ], tot_loss[loss=0.3353, simple_loss=0.3873, pruned_loss=0.1417, over 3180751.70 frames. ], batch size: 247, lr: 4.34e-02, grad_scale: 16.0 2023-04-27 13:49:36,027 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 2023-04-27 13:50:01,648 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 13:50:04,951 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:50:36,412 INFO [train.py:904] (5/8) Epoch 1, batch 4400, loss[loss=0.345, simple_loss=0.4028, pruned_loss=0.1436, over 16177.00 frames. ], tot_loss[loss=0.3349, simple_loss=0.3878, pruned_loss=0.141, over 3184546.19 frames. ], batch size: 165, lr: 4.33e-02, grad_scale: 16.0 2023-04-27 13:50:51,962 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.736e+02 5.030e+02 6.587e+02 8.143e+02 1.430e+03, threshold=1.317e+03, percent-clipped=9.0 2023-04-27 13:51:48,426 INFO [train.py:904] (5/8) Epoch 1, batch 4450, loss[loss=0.2944, simple_loss=0.3632, pruned_loss=0.1128, over 17028.00 frames. ], tot_loss[loss=0.3341, simple_loss=0.3895, pruned_loss=0.1394, over 3208087.57 frames. ], batch size: 50, lr: 4.32e-02, grad_scale: 16.0 2023-04-27 13:52:00,128 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9196, 4.0950, 3.8139, 4.0081, 3.5787, 3.9665, 3.8377, 3.9598], device='cuda:5'), covar=tensor([0.0356, 0.0570, 0.0535, 0.0309, 0.0595, 0.0434, 0.0416, 0.0477], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0162, 0.0143, 0.0106, 0.0139, 0.0117, 0.0145, 0.0099], device='cuda:5'), out_proj_covar=tensor([1.3536e-04, 1.5799e-04, 1.2869e-04, 9.7618e-05, 1.3010e-04, 1.0821e-04, 1.4604e-04, 1.0025e-04], device='cuda:5') 2023-04-27 13:52:57,216 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:53:02,602 INFO [train.py:904] (5/8) Epoch 1, batch 4500, loss[loss=0.298, simple_loss=0.3698, pruned_loss=0.1131, over 16794.00 frames. ], tot_loss[loss=0.3311, simple_loss=0.3875, pruned_loss=0.1373, over 3219898.59 frames. ], batch size: 83, lr: 4.31e-02, grad_scale: 8.0 2023-04-27 13:53:20,031 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.477e+02 4.034e+02 4.961e+02 6.637e+02 1.457e+03, threshold=9.923e+02, percent-clipped=1.0 2023-04-27 13:54:14,094 INFO [train.py:904] (5/8) Epoch 1, batch 4550, loss[loss=0.3617, simple_loss=0.4119, pruned_loss=0.1557, over 16323.00 frames. ], tot_loss[loss=0.3305, simple_loss=0.3873, pruned_loss=0.1369, over 3230537.98 frames. ], batch size: 35, lr: 4.30e-02, grad_scale: 8.0 2023-04-27 13:54:58,055 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:55:25,726 INFO [train.py:904] (5/8) Epoch 1, batch 4600, loss[loss=0.3319, simple_loss=0.3989, pruned_loss=0.1324, over 16732.00 frames. ], tot_loss[loss=0.3272, simple_loss=0.3859, pruned_loss=0.1342, over 3238731.32 frames. ], batch size: 83, lr: 4.29e-02, grad_scale: 8.0 2023-04-27 13:55:26,889 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7677, 3.7116, 3.6747, 3.6610, 3.8105, 4.0918, 4.0159, 3.6761], device='cuda:5'), covar=tensor([0.0973, 0.1201, 0.0711, 0.1495, 0.1627, 0.0654, 0.0600, 0.1511], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0180, 0.0140, 0.0152, 0.0188, 0.0131, 0.0137, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:5') 2023-04-27 13:55:43,322 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.332e+02 3.980e+02 4.876e+02 6.501e+02 1.584e+03, threshold=9.751e+02, percent-clipped=6.0 2023-04-27 13:56:07,465 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:28,174 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4644.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:36,575 INFO [train.py:904] (5/8) Epoch 1, batch 4650, loss[loss=0.3635, simple_loss=0.3957, pruned_loss=0.1656, over 11458.00 frames. ], tot_loss[loss=0.3234, simple_loss=0.3828, pruned_loss=0.132, over 3235976.56 frames. ], batch size: 247, lr: 4.28e-02, grad_scale: 8.0 2023-04-27 13:57:29,092 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7543, 3.7889, 1.5615, 3.7897, 2.7786, 3.5742, 2.0025, 3.1211], device='cuda:5'), covar=tensor([0.0113, 0.0084, 0.1165, 0.0113, 0.0326, 0.0110, 0.0818, 0.0231], device='cuda:5'), in_proj_covar=tensor([0.0041, 0.0038, 0.0078, 0.0040, 0.0062, 0.0036, 0.0082, 0.0050], device='cuda:5'), out_proj_covar=tensor([5.0403e-05, 5.1150e-05, 1.0945e-04, 4.8734e-05, 7.7566e-05, 5.3646e-05, 1.0774e-04, 6.9071e-05], device='cuda:5') 2023-04-27 13:57:50,180 INFO [train.py:904] (5/8) Epoch 1, batch 4700, loss[loss=0.2983, simple_loss=0.3603, pruned_loss=0.1181, over 16660.00 frames. ], tot_loss[loss=0.3197, simple_loss=0.3795, pruned_loss=0.13, over 3235437.26 frames. ], batch size: 62, lr: 4.27e-02, grad_scale: 8.0 2023-04-27 13:58:07,987 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.460e+02 4.210e+02 5.632e+02 6.697e+02 1.082e+03, threshold=1.126e+03, percent-clipped=2.0 2023-04-27 13:58:26,259 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-04-27 13:58:45,585 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3891, 4.6567, 4.5205, 4.4512, 4.6277, 4.9235, 4.8720, 4.5139], device='cuda:5'), covar=tensor([0.0634, 0.0804, 0.0614, 0.1119, 0.1354, 0.0509, 0.0507, 0.1217], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0174, 0.0139, 0.0150, 0.0181, 0.0132, 0.0137, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:5') 2023-04-27 13:59:02,517 INFO [train.py:904] (5/8) Epoch 1, batch 4750, loss[loss=0.2623, simple_loss=0.3344, pruned_loss=0.09509, over 16511.00 frames. ], tot_loss[loss=0.3161, simple_loss=0.3756, pruned_loss=0.1283, over 3232374.59 frames. ], batch size: 68, lr: 4.26e-02, grad_scale: 8.0 2023-04-27 14:00:11,443 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:00:16,569 INFO [train.py:904] (5/8) Epoch 1, batch 4800, loss[loss=0.3567, simple_loss=0.3999, pruned_loss=0.1568, over 11642.00 frames. ], tot_loss[loss=0.3129, simple_loss=0.3723, pruned_loss=0.1268, over 3212715.31 frames. ], batch size: 248, lr: 4.25e-02, grad_scale: 8.0 2023-04-27 14:00:34,036 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.251e+02 4.322e+02 5.200e+02 6.651e+02 1.076e+03, threshold=1.040e+03, percent-clipped=0.0 2023-04-27 14:01:22,948 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:31,756 INFO [train.py:904] (5/8) Epoch 1, batch 4850, loss[loss=0.3043, simple_loss=0.3723, pruned_loss=0.1182, over 16822.00 frames. ], tot_loss[loss=0.3133, simple_loss=0.3736, pruned_loss=0.1265, over 3203627.16 frames. ], batch size: 96, lr: 4.24e-02, grad_scale: 8.0 2023-04-27 14:02:03,907 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 14:02:13,592 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.60 vs. limit=5.0 2023-04-27 14:02:49,104 INFO [train.py:904] (5/8) Epoch 1, batch 4900, loss[loss=0.3402, simple_loss=0.4003, pruned_loss=0.1401, over 16647.00 frames. ], tot_loss[loss=0.3107, simple_loss=0.3724, pruned_loss=0.1245, over 3199545.70 frames. ], batch size: 134, lr: 4.23e-02, grad_scale: 8.0 2023-04-27 14:03:07,683 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 3.780e+02 4.613e+02 5.959e+02 1.211e+03, threshold=9.227e+02, percent-clipped=3.0 2023-04-27 14:03:21,066 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 14:03:47,088 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3201, 4.7015, 4.4744, 4.4641, 4.5760, 5.0454, 4.8638, 4.4463], device='cuda:5'), covar=tensor([0.0774, 0.0791, 0.0716, 0.1217, 0.1502, 0.0482, 0.0648, 0.1545], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0186, 0.0146, 0.0157, 0.0191, 0.0137, 0.0142, 0.0206], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:5') 2023-04-27 14:03:54,983 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:04:04,725 INFO [train.py:904] (5/8) Epoch 1, batch 4950, loss[loss=0.3184, simple_loss=0.3785, pruned_loss=0.1291, over 16522.00 frames. ], tot_loss[loss=0.3114, simple_loss=0.3726, pruned_loss=0.1251, over 3209930.33 frames. ], batch size: 68, lr: 4.21e-02, grad_scale: 8.0 2023-04-27 14:05:04,815 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:05:17,377 INFO [train.py:904] (5/8) Epoch 1, batch 5000, loss[loss=0.3405, simple_loss=0.3985, pruned_loss=0.1413, over 15458.00 frames. ], tot_loss[loss=0.3135, simple_loss=0.375, pruned_loss=0.126, over 3203199.78 frames. ], batch size: 191, lr: 4.20e-02, grad_scale: 8.0 2023-04-27 14:05:24,237 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3611, 5.5592, 5.2253, 5.5979, 5.0155, 5.0690, 5.1060, 5.6979], device='cuda:5'), covar=tensor([0.0255, 0.0416, 0.0381, 0.0209, 0.0459, 0.0249, 0.0385, 0.0189], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0167, 0.0152, 0.0110, 0.0141, 0.0114, 0.0150, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 14:05:35,394 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.744e+02 4.820e+02 5.822e+02 7.344e+02 1.526e+03, threshold=1.164e+03, percent-clipped=12.0 2023-04-27 14:06:31,106 INFO [train.py:904] (5/8) Epoch 1, batch 5050, loss[loss=0.2594, simple_loss=0.3315, pruned_loss=0.09361, over 17124.00 frames. ], tot_loss[loss=0.3109, simple_loss=0.3736, pruned_loss=0.1241, over 3204229.86 frames. ], batch size: 48, lr: 4.19e-02, grad_scale: 8.0 2023-04-27 14:06:45,454 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0487, 1.1595, 1.4041, 1.6563, 1.3747, 1.5568, 1.6245, 1.7870], device='cuda:5'), covar=tensor([0.0114, 0.0535, 0.0242, 0.0231, 0.0183, 0.0189, 0.0222, 0.0205], device='cuda:5'), in_proj_covar=tensor([0.0027, 0.0049, 0.0032, 0.0030, 0.0030, 0.0033, 0.0029, 0.0031], device='cuda:5'), out_proj_covar=tensor([2.9183e-05, 6.3884e-05, 3.5784e-05, 3.5127e-05, 3.1163e-05, 3.3771e-05, 3.2949e-05, 3.4229e-05], device='cuda:5') 2023-04-27 14:07:10,213 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2023-04-27 14:07:21,194 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-27 14:07:23,351 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=2.12 vs. limit=2.0 2023-04-27 14:07:42,603 INFO [train.py:904] (5/8) Epoch 1, batch 5100, loss[loss=0.228, simple_loss=0.3147, pruned_loss=0.07062, over 16885.00 frames. ], tot_loss[loss=0.3073, simple_loss=0.3707, pruned_loss=0.122, over 3211270.80 frames. ], batch size: 96, lr: 4.18e-02, grad_scale: 8.0 2023-04-27 14:07:59,832 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.617e+02 4.246e+02 5.535e+02 6.678e+02 1.197e+03, threshold=1.107e+03, percent-clipped=1.0 2023-04-27 14:08:58,116 INFO [train.py:904] (5/8) Epoch 1, batch 5150, loss[loss=0.2759, simple_loss=0.3439, pruned_loss=0.104, over 16807.00 frames. ], tot_loss[loss=0.3077, simple_loss=0.3718, pruned_loss=0.1218, over 3199260.36 frames. ], batch size: 39, lr: 4.17e-02, grad_scale: 8.0 2023-04-27 14:10:11,736 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6882, 3.6379, 1.5183, 3.7393, 2.3322, 3.6796, 1.9166, 2.6968], device='cuda:5'), covar=tensor([0.0066, 0.0078, 0.1270, 0.0096, 0.0527, 0.0095, 0.1129, 0.0384], device='cuda:5'), in_proj_covar=tensor([0.0047, 0.0045, 0.0094, 0.0049, 0.0076, 0.0042, 0.0102, 0.0067], device='cuda:5'), out_proj_covar=tensor([6.2032e-05, 6.3398e-05, 1.3509e-04, 6.4067e-05, 1.0107e-04, 6.5830e-05, 1.4076e-04, 9.8163e-05], device='cuda:5') 2023-04-27 14:10:12,905 INFO [train.py:904] (5/8) Epoch 1, batch 5200, loss[loss=0.3029, simple_loss=0.3635, pruned_loss=0.1211, over 16876.00 frames. ], tot_loss[loss=0.308, simple_loss=0.3717, pruned_loss=0.1221, over 3198587.91 frames. ], batch size: 96, lr: 4.16e-02, grad_scale: 8.0 2023-04-27 14:10:30,246 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.582e+02 4.001e+02 4.794e+02 6.093e+02 1.086e+03, threshold=9.588e+02, percent-clipped=0.0 2023-04-27 14:11:26,081 INFO [train.py:904] (5/8) Epoch 1, batch 5250, loss[loss=0.2653, simple_loss=0.3441, pruned_loss=0.09328, over 16866.00 frames. ], tot_loss[loss=0.305, simple_loss=0.3677, pruned_loss=0.1211, over 3209368.55 frames. ], batch size: 96, lr: 4.15e-02, grad_scale: 8.0 2023-04-27 14:12:37,178 INFO [train.py:904] (5/8) Epoch 1, batch 5300, loss[loss=0.2508, simple_loss=0.3201, pruned_loss=0.0908, over 16706.00 frames. ], tot_loss[loss=0.3008, simple_loss=0.3636, pruned_loss=0.119, over 3217783.71 frames. ], batch size: 62, lr: 4.14e-02, grad_scale: 8.0 2023-04-27 14:12:54,711 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.864e+02 4.537e+02 5.350e+02 6.217e+02 1.130e+03, threshold=1.070e+03, percent-clipped=3.0 2023-04-27 14:13:49,540 INFO [train.py:904] (5/8) Epoch 1, batch 5350, loss[loss=0.2995, simple_loss=0.36, pruned_loss=0.1196, over 16474.00 frames. ], tot_loss[loss=0.2969, simple_loss=0.3599, pruned_loss=0.1169, over 3221870.92 frames. ], batch size: 62, lr: 4.13e-02, grad_scale: 8.0 2023-04-27 14:14:37,323 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6566, 3.8103, 3.3783, 1.7045, 2.9831, 2.4300, 3.3246, 4.0237], device='cuda:5'), covar=tensor([0.0264, 0.0228, 0.0313, 0.1831, 0.0752, 0.0958, 0.0636, 0.0115], device='cuda:5'), in_proj_covar=tensor([0.0079, 0.0055, 0.0092, 0.0138, 0.0127, 0.0124, 0.0112, 0.0052], device='cuda:5'), out_proj_covar=tensor([1.2935e-04, 9.4928e-05, 1.2831e-04, 1.7504e-04, 1.7328e-04, 1.6058e-04, 1.6892e-04, 9.0166e-05], device='cuda:5') 2023-04-27 14:14:42,415 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-04-27 14:15:01,000 INFO [train.py:904] (5/8) Epoch 1, batch 5400, loss[loss=0.3322, simple_loss=0.3941, pruned_loss=0.1351, over 16646.00 frames. ], tot_loss[loss=0.3021, simple_loss=0.3648, pruned_loss=0.1197, over 3216583.34 frames. ], batch size: 134, lr: 4.12e-02, grad_scale: 8.0 2023-04-27 14:15:18,319 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.226e+02 4.427e+02 5.350e+02 6.399e+02 9.942e+02, threshold=1.070e+03, percent-clipped=0.0 2023-04-27 14:16:19,551 INFO [train.py:904] (5/8) Epoch 1, batch 5450, loss[loss=0.4667, simple_loss=0.4874, pruned_loss=0.223, over 15293.00 frames. ], tot_loss[loss=0.3087, simple_loss=0.3698, pruned_loss=0.1238, over 3206261.10 frames. ], batch size: 190, lr: 4.11e-02, grad_scale: 8.0 2023-04-27 14:17:37,161 INFO [train.py:904] (5/8) Epoch 1, batch 5500, loss[loss=0.3764, simple_loss=0.4251, pruned_loss=0.1638, over 16413.00 frames. ], tot_loss[loss=0.3241, simple_loss=0.3811, pruned_loss=0.1335, over 3193084.05 frames. ], batch size: 146, lr: 4.10e-02, grad_scale: 8.0 2023-04-27 14:17:53,978 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 14:17:56,203 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.545e+02 5.862e+02 7.575e+02 9.403e+02 2.285e+03, threshold=1.515e+03, percent-clipped=16.0 2023-04-27 14:18:33,390 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2654, 2.7804, 2.6980, 3.4386, 2.6878, 3.3326, 2.8682, 2.7037], device='cuda:5'), covar=tensor([0.0264, 0.0377, 0.0331, 0.0276, 0.0980, 0.0224, 0.0501, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0074, 0.0080, 0.0070, 0.0081, 0.0147, 0.0080, 0.0099, 0.0091], device='cuda:5'), out_proj_covar=tensor([9.2785e-05, 9.6305e-05, 8.2228e-05, 1.0398e-04, 1.7761e-04, 9.6101e-05, 1.0868e-04, 1.1890e-04], device='cuda:5') 2023-04-27 14:18:39,590 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 14:18:57,390 INFO [train.py:904] (5/8) Epoch 1, batch 5550, loss[loss=0.3555, simple_loss=0.401, pruned_loss=0.1551, over 16631.00 frames. ], tot_loss[loss=0.3417, simple_loss=0.3929, pruned_loss=0.1452, over 3138640.41 frames. ], batch size: 57, lr: 4.09e-02, grad_scale: 8.0 2023-04-27 14:19:11,575 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9179, 4.0541, 3.7912, 4.0165, 3.5734, 3.9303, 3.8162, 3.9139], device='cuda:5'), covar=tensor([0.0397, 0.0649, 0.0760, 0.0344, 0.0621, 0.0452, 0.0529, 0.0544], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0182, 0.0170, 0.0118, 0.0147, 0.0121, 0.0161, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 14:20:17,845 INFO [train.py:904] (5/8) Epoch 1, batch 5600, loss[loss=0.3232, simple_loss=0.3766, pruned_loss=0.135, over 16575.00 frames. ], tot_loss[loss=0.3527, simple_loss=0.4003, pruned_loss=0.1526, over 3099437.52 frames. ], batch size: 57, lr: 4.08e-02, grad_scale: 8.0 2023-04-27 14:20:37,625 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.538e+02 5.637e+02 6.821e+02 8.546e+02 1.933e+03, threshold=1.364e+03, percent-clipped=3.0 2023-04-27 14:21:39,931 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:21:40,607 INFO [train.py:904] (5/8) Epoch 1, batch 5650, loss[loss=0.5406, simple_loss=0.5184, pruned_loss=0.2814, over 11451.00 frames. ], tot_loss[loss=0.3654, simple_loss=0.409, pruned_loss=0.1609, over 3081919.35 frames. ], batch size: 249, lr: 4.07e-02, grad_scale: 8.0 2023-04-27 14:22:59,489 INFO [train.py:904] (5/8) Epoch 1, batch 5700, loss[loss=0.4008, simple_loss=0.4462, pruned_loss=0.1777, over 15322.00 frames. ], tot_loss[loss=0.3686, simple_loss=0.411, pruned_loss=0.1632, over 3066267.32 frames. ], batch size: 190, lr: 4.06e-02, grad_scale: 8.0 2023-04-27 14:23:15,471 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:23:17,966 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.964e+02 6.202e+02 7.607e+02 9.700e+02 2.079e+03, threshold=1.521e+03, percent-clipped=5.0 2023-04-27 14:23:22,745 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-27 14:24:21,193 INFO [train.py:904] (5/8) Epoch 1, batch 5750, loss[loss=0.3668, simple_loss=0.4209, pruned_loss=0.1564, over 16667.00 frames. ], tot_loss[loss=0.3722, simple_loss=0.4144, pruned_loss=0.165, over 3048211.47 frames. ], batch size: 134, lr: 4.05e-02, grad_scale: 8.0 2023-04-27 14:24:28,891 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:25:42,934 INFO [train.py:904] (5/8) Epoch 1, batch 5800, loss[loss=0.3363, simple_loss=0.3966, pruned_loss=0.138, over 16852.00 frames. ], tot_loss[loss=0.3691, simple_loss=0.4134, pruned_loss=0.1625, over 3056968.62 frames. ], batch size: 116, lr: 4.04e-02, grad_scale: 8.0 2023-04-27 14:26:01,840 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.312e+02 5.567e+02 6.621e+02 8.933e+02 1.804e+03, threshold=1.324e+03, percent-clipped=2.0 2023-04-27 14:26:06,077 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 14:26:06,984 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:26:16,712 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-27 14:27:02,351 INFO [train.py:904] (5/8) Epoch 1, batch 5850, loss[loss=0.3041, simple_loss=0.37, pruned_loss=0.1191, over 16927.00 frames. ], tot_loss[loss=0.3643, simple_loss=0.41, pruned_loss=0.1593, over 3062916.02 frames. ], batch size: 96, lr: 4.03e-02, grad_scale: 8.0 2023-04-27 14:27:15,058 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:25,820 INFO [train.py:904] (5/8) Epoch 1, batch 5900, loss[loss=0.3356, simple_loss=0.3949, pruned_loss=0.1382, over 16748.00 frames. ], tot_loss[loss=0.3609, simple_loss=0.4081, pruned_loss=0.1569, over 3076773.58 frames. ], batch size: 134, lr: 4.02e-02, grad_scale: 8.0 2023-04-27 14:28:39,839 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8697, 5.0323, 5.4423, 5.4366, 5.6739, 5.0443, 5.0889, 5.2011], device='cuda:5'), covar=tensor([0.0263, 0.0199, 0.0351, 0.0345, 0.0224, 0.0220, 0.0640, 0.0231], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0103, 0.0131, 0.0128, 0.0138, 0.0111, 0.0145, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 14:28:48,065 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.490e+02 5.425e+02 6.564e+02 8.517e+02 1.690e+03, threshold=1.313e+03, percent-clipped=2.0 2023-04-27 14:28:58,980 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:29:18,267 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 14:29:20,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8919, 1.4521, 1.4466, 1.6444, 1.6210, 1.7599, 1.6322, 1.7850], device='cuda:5'), covar=tensor([0.0097, 0.0508, 0.0210, 0.0196, 0.0215, 0.0230, 0.0227, 0.0177], device='cuda:5'), in_proj_covar=tensor([0.0031, 0.0055, 0.0036, 0.0034, 0.0034, 0.0039, 0.0030, 0.0032], device='cuda:5'), out_proj_covar=tensor([3.5137e-05, 7.7235e-05, 4.4091e-05, 4.3276e-05, 3.9694e-05, 4.5201e-05, 3.8211e-05, 4.0220e-05], device='cuda:5') 2023-04-27 14:29:49,237 INFO [train.py:904] (5/8) Epoch 1, batch 5950, loss[loss=0.3265, simple_loss=0.3938, pruned_loss=0.1296, over 16884.00 frames. ], tot_loss[loss=0.3583, simple_loss=0.4083, pruned_loss=0.1542, over 3098605.73 frames. ], batch size: 96, lr: 4.01e-02, grad_scale: 8.0 2023-04-27 14:30:31,767 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9366, 1.8136, 1.8466, 2.5113, 2.4808, 2.5945, 2.2249, 2.3130], device='cuda:5'), covar=tensor([0.0151, 0.0759, 0.0317, 0.0144, 0.0097, 0.0117, 0.0166, 0.0150], device='cuda:5'), in_proj_covar=tensor([0.0037, 0.0080, 0.0058, 0.0042, 0.0036, 0.0036, 0.0045, 0.0037], device='cuda:5'), out_proj_covar=tensor([6.0721e-05, 1.4362e-04, 1.0453e-04, 7.0225e-05, 5.8971e-05, 5.9063e-05, 6.8067e-05, 6.1417e-05], device='cuda:5') 2023-04-27 14:30:47,584 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4079, 4.9217, 5.1615, 5.3587, 4.5121, 5.2400, 4.8396, 4.8146], device='cuda:5'), covar=tensor([0.0141, 0.0108, 0.0126, 0.0079, 0.0620, 0.0122, 0.0105, 0.0120], device='cuda:5'), in_proj_covar=tensor([0.0081, 0.0061, 0.0110, 0.0087, 0.0137, 0.0085, 0.0075, 0.0089], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-27 14:30:52,707 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:31:14,077 INFO [train.py:904] (5/8) Epoch 1, batch 6000, loss[loss=0.3453, simple_loss=0.39, pruned_loss=0.1503, over 17025.00 frames. ], tot_loss[loss=0.3583, simple_loss=0.4076, pruned_loss=0.1545, over 3094371.69 frames. ], batch size: 50, lr: 4.00e-02, grad_scale: 8.0 2023-04-27 14:31:14,077 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 14:31:23,950 INFO [train.py:938] (5/8) Epoch 1, validation: loss=0.2762, simple_loss=0.3752, pruned_loss=0.08863, over 944034.00 frames. 2023-04-27 14:31:23,951 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17316MB 2023-04-27 14:31:31,563 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:31:41,458 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.294e+02 5.611e+02 7.114e+02 9.006e+02 1.900e+03, threshold=1.423e+03, percent-clipped=2.0 2023-04-27 14:32:30,089 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:42,919 INFO [train.py:904] (5/8) Epoch 1, batch 6050, loss[loss=0.3148, simple_loss=0.3824, pruned_loss=0.1236, over 16862.00 frames. ], tot_loss[loss=0.3563, simple_loss=0.4055, pruned_loss=0.1536, over 3090097.85 frames. ], batch size: 102, lr: 3.99e-02, grad_scale: 8.0 2023-04-27 14:32:44,879 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:03,535 INFO [train.py:904] (5/8) Epoch 1, batch 6100, loss[loss=0.31, simple_loss=0.3821, pruned_loss=0.119, over 16504.00 frames. ], tot_loss[loss=0.3507, simple_loss=0.4029, pruned_loss=0.1493, over 3112445.08 frames. ], batch size: 68, lr: 3.98e-02, grad_scale: 8.0 2023-04-27 14:34:09,267 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:21,440 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:24,017 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.820e+02 4.604e+02 6.189e+02 8.357e+02 1.892e+03, threshold=1.238e+03, percent-clipped=2.0 2023-04-27 14:34:59,819 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:03,693 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8250, 1.6950, 1.6153, 1.5209, 1.9231, 1.7993, 1.9382, 2.1230], device='cuda:5'), covar=tensor([0.0099, 0.0335, 0.0156, 0.0198, 0.0132, 0.0235, 0.0164, 0.0116], device='cuda:5'), in_proj_covar=tensor([0.0030, 0.0055, 0.0035, 0.0033, 0.0034, 0.0037, 0.0029, 0.0030], device='cuda:5'), out_proj_covar=tensor([3.4665e-05, 7.7133e-05, 4.4113e-05, 4.2397e-05, 4.1349e-05, 4.4114e-05, 3.8866e-05, 3.8166e-05], device='cuda:5') 2023-04-27 14:35:25,405 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.2909, 3.3273, 3.0742, 2.3438, 3.0956, 3.2439, 3.2324, 1.9643], device='cuda:5'), covar=tensor([0.1497, 0.0087, 0.0131, 0.0375, 0.0120, 0.0114, 0.0098, 0.0529], device='cuda:5'), in_proj_covar=tensor([0.0097, 0.0037, 0.0038, 0.0062, 0.0037, 0.0037, 0.0043, 0.0058], device='cuda:5'), out_proj_covar=tensor([1.7863e-04, 7.4357e-05, 7.7150e-05, 1.1782e-04, 7.4301e-05, 8.0601e-05, 8.2913e-05, 1.1301e-04], device='cuda:5') 2023-04-27 14:35:26,096 INFO [train.py:904] (5/8) Epoch 1, batch 6150, loss[loss=0.3295, simple_loss=0.3884, pruned_loss=0.1353, over 16364.00 frames. ], tot_loss[loss=0.3473, simple_loss=0.3999, pruned_loss=0.1473, over 3123255.62 frames. ], batch size: 146, lr: 3.97e-02, grad_scale: 8.0 2023-04-27 14:35:57,282 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6170.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:39,078 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:46,264 INFO [train.py:904] (5/8) Epoch 1, batch 6200, loss[loss=0.336, simple_loss=0.3947, pruned_loss=0.1386, over 16733.00 frames. ], tot_loss[loss=0.3476, simple_loss=0.399, pruned_loss=0.1482, over 3110884.99 frames. ], batch size: 89, lr: 3.96e-02, grad_scale: 8.0 2023-04-27 14:36:49,270 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:03,951 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:05,933 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.990e+02 5.160e+02 6.786e+02 8.603e+02 1.824e+03, threshold=1.357e+03, percent-clipped=7.0 2023-04-27 14:37:08,088 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:33,374 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:02,379 INFO [train.py:904] (5/8) Epoch 1, batch 6250, loss[loss=0.3502, simple_loss=0.4022, pruned_loss=0.1491, over 16258.00 frames. ], tot_loss[loss=0.3478, simple_loss=0.3987, pruned_loss=0.1485, over 3090410.23 frames. ], batch size: 165, lr: 3.95e-02, grad_scale: 8.0 2023-04-27 14:38:20,991 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:22,882 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3834, 4.3931, 3.9430, 1.8219, 3.3028, 2.5341, 3.6315, 4.4379], device='cuda:5'), covar=tensor([0.0203, 0.0231, 0.0281, 0.2181, 0.0847, 0.1275, 0.0780, 0.0203], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0063, 0.0100, 0.0140, 0.0137, 0.0130, 0.0118, 0.0058], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 14:38:24,269 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:32,014 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:39:17,026 INFO [train.py:904] (5/8) Epoch 1, batch 6300, loss[loss=0.3885, simple_loss=0.4106, pruned_loss=0.1831, over 11403.00 frames. ], tot_loss[loss=0.3466, simple_loss=0.3984, pruned_loss=0.1474, over 3098815.44 frames. ], batch size: 247, lr: 3.94e-02, grad_scale: 8.0 2023-04-27 14:39:25,832 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:39:36,725 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.400e+02 5.487e+02 6.666e+02 8.343e+02 1.856e+03, threshold=1.333e+03, percent-clipped=2.0 2023-04-27 14:39:59,139 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:28,624 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:36,342 INFO [train.py:904] (5/8) Epoch 1, batch 6350, loss[loss=0.3676, simple_loss=0.4186, pruned_loss=0.1583, over 16691.00 frames. ], tot_loss[loss=0.3505, simple_loss=0.4, pruned_loss=0.1505, over 3090505.91 frames. ], batch size: 89, lr: 3.93e-02, grad_scale: 8.0 2023-04-27 14:40:41,032 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:40:50,582 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 14:41:38,278 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0233, 1.6489, 1.7015, 1.6378, 2.5149, 2.4137, 2.5209, 2.0008], device='cuda:5'), covar=tensor([0.0098, 0.0427, 0.0173, 0.0221, 0.0127, 0.0132, 0.0100, 0.0243], device='cuda:5'), in_proj_covar=tensor([0.0030, 0.0058, 0.0038, 0.0036, 0.0035, 0.0039, 0.0031, 0.0031], device='cuda:5'), out_proj_covar=tensor([3.5829e-05, 8.2010e-05, 4.8995e-05, 4.8877e-05, 4.3649e-05, 4.7753e-05, 4.2435e-05, 4.1529e-05], device='cuda:5') 2023-04-27 14:41:50,541 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:53,487 INFO [train.py:904] (5/8) Epoch 1, batch 6400, loss[loss=0.3132, simple_loss=0.3771, pruned_loss=0.1247, over 16466.00 frames. ], tot_loss[loss=0.3521, simple_loss=0.4005, pruned_loss=0.1519, over 3076834.93 frames. ], batch size: 75, lr: 3.92e-02, grad_scale: 8.0 2023-04-27 14:42:08,637 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6411.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:42:10,411 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 4.039e+02 5.999e+02 7.400e+02 9.043e+02 1.587e+03, threshold=1.480e+03, percent-clipped=2.0 2023-04-27 14:43:09,622 INFO [train.py:904] (5/8) Epoch 1, batch 6450, loss[loss=0.2481, simple_loss=0.3357, pruned_loss=0.08023, over 16802.00 frames. ], tot_loss[loss=0.3463, simple_loss=0.3969, pruned_loss=0.1479, over 3085821.58 frames. ], batch size: 83, lr: 3.91e-02, grad_scale: 8.0 2023-04-27 14:43:22,745 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:26,567 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:33,845 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:12,253 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:26,986 INFO [train.py:904] (5/8) Epoch 1, batch 6500, loss[loss=0.3703, simple_loss=0.3912, pruned_loss=0.1747, over 11528.00 frames. ], tot_loss[loss=0.3436, simple_loss=0.3941, pruned_loss=0.1465, over 3096635.08 frames. ], batch size: 247, lr: 3.90e-02, grad_scale: 16.0 2023-04-27 14:44:45,183 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.861e+02 4.801e+02 5.896e+02 8.038e+02 1.836e+03, threshold=1.179e+03, percent-clipped=2.0 2023-04-27 14:44:46,922 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:00,312 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:45:05,087 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:07,734 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:42,983 INFO [train.py:904] (5/8) Epoch 1, batch 6550, loss[loss=0.3377, simple_loss=0.4102, pruned_loss=0.1327, over 17045.00 frames. ], tot_loss[loss=0.3468, simple_loss=0.3977, pruned_loss=0.148, over 3090198.94 frames. ], batch size: 53, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:45:55,663 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:56,181 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-04-27 14:46:00,537 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:08,823 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:59,705 INFO [train.py:904] (5/8) Epoch 1, batch 6600, loss[loss=0.3852, simple_loss=0.4195, pruned_loss=0.1754, over 16740.00 frames. ], tot_loss[loss=0.35, simple_loss=0.4011, pruned_loss=0.1494, over 3086677.74 frames. ], batch size: 62, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:47:18,178 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.822e+02 5.266e+02 6.465e+02 8.044e+02 1.550e+03, threshold=1.293e+03, percent-clipped=3.0 2023-04-27 14:47:32,881 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:47:43,249 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:10,816 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3685, 4.5037, 4.4430, 4.5344, 4.4868, 4.9877, 4.8427, 4.4241], device='cuda:5'), covar=tensor([0.0840, 0.1173, 0.0916, 0.1297, 0.1908, 0.0750, 0.0821, 0.2059], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0202, 0.0166, 0.0172, 0.0216, 0.0164, 0.0159, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 14:48:10,846 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:18,092 INFO [train.py:904] (5/8) Epoch 1, batch 6650, loss[loss=0.3273, simple_loss=0.3841, pruned_loss=0.1352, over 16509.00 frames. ], tot_loss[loss=0.3509, simple_loss=0.4014, pruned_loss=0.1502, over 3092996.57 frames. ], batch size: 68, lr: 3.88e-02, grad_scale: 16.0 2023-04-27 14:48:31,305 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 14:49:18,677 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:21,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1337, 3.7684, 4.0316, 4.2126, 3.5148, 3.9531, 3.9763, 3.7482], device='cuda:5'), covar=tensor([0.0234, 0.0184, 0.0189, 0.0104, 0.0809, 0.0231, 0.0248, 0.0198], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0063, 0.0115, 0.0089, 0.0139, 0.0090, 0.0079, 0.0094], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 14:49:24,166 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:32,160 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:34,325 INFO [train.py:904] (5/8) Epoch 1, batch 6700, loss[loss=0.4162, simple_loss=0.4284, pruned_loss=0.202, over 11080.00 frames. ], tot_loss[loss=0.3503, simple_loss=0.4002, pruned_loss=0.1502, over 3090049.11 frames. ], batch size: 247, lr: 3.87e-02, grad_scale: 16.0 2023-04-27 14:49:52,616 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.128e+02 5.771e+02 7.025e+02 8.722e+02 1.711e+03, threshold=1.405e+03, percent-clipped=7.0 2023-04-27 14:50:45,227 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:50:50,998 INFO [train.py:904] (5/8) Epoch 1, batch 6750, loss[loss=0.2998, simple_loss=0.3647, pruned_loss=0.1174, over 16710.00 frames. ], tot_loss[loss=0.3499, simple_loss=0.3995, pruned_loss=0.1502, over 3082340.89 frames. ], batch size: 76, lr: 3.86e-02, grad_scale: 16.0 2023-04-27 14:51:51,418 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:05,978 INFO [train.py:904] (5/8) Epoch 1, batch 6800, loss[loss=0.3435, simple_loss=0.397, pruned_loss=0.145, over 16657.00 frames. ], tot_loss[loss=0.3488, simple_loss=0.399, pruned_loss=0.1494, over 3085353.76 frames. ], batch size: 134, lr: 3.85e-02, grad_scale: 16.0 2023-04-27 14:52:24,954 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.031e+02 5.369e+02 6.197e+02 7.365e+02 1.417e+03, threshold=1.239e+03, percent-clipped=1.0 2023-04-27 14:52:33,899 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:52:41,652 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:46,891 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:06,061 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:23,171 INFO [train.py:904] (5/8) Epoch 1, batch 6850, loss[loss=0.3572, simple_loss=0.4241, pruned_loss=0.1452, over 16915.00 frames. ], tot_loss[loss=0.3501, simple_loss=0.4002, pruned_loss=0.1499, over 3088980.05 frames. ], batch size: 116, lr: 3.84e-02, grad_scale: 16.0 2023-04-27 14:53:28,398 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 14:53:35,286 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:46,437 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:53,853 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 14:53:57,789 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:03,466 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:54:14,200 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0108, 5.1608, 4.9643, 5.1455, 4.5984, 4.8052, 4.7419, 5.3444], device='cuda:5'), covar=tensor([0.0418, 0.0653, 0.0670, 0.0274, 0.0595, 0.0305, 0.0528, 0.0350], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0199, 0.0190, 0.0126, 0.0161, 0.0129, 0.0179, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 14:54:37,400 INFO [train.py:904] (5/8) Epoch 1, batch 6900, loss[loss=0.3709, simple_loss=0.419, pruned_loss=0.1614, over 16450.00 frames. ], tot_loss[loss=0.3495, simple_loss=0.4018, pruned_loss=0.1486, over 3102920.81 frames. ], batch size: 146, lr: 3.83e-02, grad_scale: 16.0 2023-04-27 14:54:47,071 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:49,771 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:49,875 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8168, 3.0803, 1.8170, 3.5838, 3.7653, 3.6548, 2.0460, 3.1823], device='cuda:5'), covar=tensor([0.2408, 0.0289, 0.2110, 0.0104, 0.0126, 0.0279, 0.0998, 0.0381], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0084, 0.0154, 0.0052, 0.0057, 0.0070, 0.0121, 0.0098], device='cuda:5'), out_proj_covar=tensor([2.1343e-04, 1.2483e-04, 2.0500e-04, 8.7950e-05, 9.7909e-05, 1.2620e-04, 1.7319e-04, 1.4565e-04], device='cuda:5') 2023-04-27 14:54:55,480 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.157e+02 5.195e+02 6.018e+02 7.768e+02 1.319e+03, threshold=1.204e+03, percent-clipped=1.0 2023-04-27 14:54:59,973 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:55:09,445 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:55:09,551 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6546, 2.4949, 2.3917, 2.2111, 2.4806, 2.4837, 2.6666, 1.9257], device='cuda:5'), covar=tensor([0.0939, 0.0120, 0.0127, 0.0304, 0.0120, 0.0122, 0.0096, 0.0484], device='cuda:5'), in_proj_covar=tensor([0.0102, 0.0041, 0.0040, 0.0066, 0.0038, 0.0038, 0.0043, 0.0066], device='cuda:5'), out_proj_covar=tensor([1.9474e-04, 8.4671e-05, 8.8335e-05, 1.3269e-04, 8.0496e-05, 8.5575e-05, 8.6658e-05, 1.3436e-04], device='cuda:5') 2023-04-27 14:55:36,611 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:55:54,255 INFO [train.py:904] (5/8) Epoch 1, batch 6950, loss[loss=0.314, simple_loss=0.3752, pruned_loss=0.1264, over 16601.00 frames. ], tot_loss[loss=0.3545, simple_loss=0.4052, pruned_loss=0.1519, over 3094425.25 frames. ], batch size: 57, lr: 3.82e-02, grad_scale: 16.0 2023-04-27 14:56:25,070 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:25,219 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:47,704 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:50,599 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 2023-04-27 14:56:55,707 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9156, 5.0902, 4.7943, 5.0445, 4.4352, 4.6222, 4.6032, 5.2237], device='cuda:5'), covar=tensor([0.0345, 0.0608, 0.0764, 0.0249, 0.0635, 0.0335, 0.0485, 0.0343], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0207, 0.0195, 0.0128, 0.0164, 0.0133, 0.0182, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 14:57:12,234 INFO [train.py:904] (5/8) Epoch 1, batch 7000, loss[loss=0.3954, simple_loss=0.447, pruned_loss=0.1719, over 16434.00 frames. ], tot_loss[loss=0.3511, simple_loss=0.4036, pruned_loss=0.1493, over 3103044.85 frames. ], batch size: 146, lr: 3.81e-02, grad_scale: 16.0 2023-04-27 14:57:30,864 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.210e+02 5.287e+02 6.266e+02 8.014e+02 1.368e+03, threshold=1.253e+03, percent-clipped=3.0 2023-04-27 14:58:27,040 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:58:31,417 INFO [train.py:904] (5/8) Epoch 1, batch 7050, loss[loss=0.3106, simple_loss=0.3776, pruned_loss=0.1218, over 16349.00 frames. ], tot_loss[loss=0.3514, simple_loss=0.4044, pruned_loss=0.1491, over 3116851.93 frames. ], batch size: 35, lr: 3.80e-02, grad_scale: 16.0 2023-04-27 14:59:51,873 INFO [train.py:904] (5/8) Epoch 1, batch 7100, loss[loss=0.36, simple_loss=0.4095, pruned_loss=0.1552, over 16696.00 frames. ], tot_loss[loss=0.3492, simple_loss=0.4018, pruned_loss=0.1483, over 3108503.63 frames. ], batch size: 134, lr: 3.79e-02, grad_scale: 16.0 2023-04-27 15:00:05,871 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:00:11,240 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.812e+02 5.623e+02 6.703e+02 8.193e+02 2.007e+03, threshold=1.341e+03, percent-clipped=3.0 2023-04-27 15:00:20,535 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:00:27,746 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:00:39,653 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 15:01:11,095 INFO [train.py:904] (5/8) Epoch 1, batch 7150, loss[loss=0.3119, simple_loss=0.3807, pruned_loss=0.1216, over 16913.00 frames. ], tot_loss[loss=0.3476, simple_loss=0.3996, pruned_loss=0.1478, over 3089086.64 frames. ], batch size: 116, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:01:28,282 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8032, 3.9018, 1.7855, 3.8781, 2.3638, 3.9193, 1.8908, 2.6251], device='cuda:5'), covar=tensor([0.0070, 0.0112, 0.1728, 0.0064, 0.0815, 0.0164, 0.1514, 0.0616], device='cuda:5'), in_proj_covar=tensor([0.0058, 0.0062, 0.0130, 0.0058, 0.0112, 0.0066, 0.0140, 0.0101], device='cuda:5'), out_proj_covar=tensor([9.9868e-05, 1.1242e-04, 2.0194e-04, 9.3643e-05, 1.7720e-04, 1.2842e-04, 2.1656e-04, 1.7065e-04], device='cuda:5') 2023-04-27 15:01:30,127 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:34,846 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:39,070 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5680, 3.2811, 3.1275, 1.4139, 3.3124, 3.3389, 2.8942, 3.0254], device='cuda:5'), covar=tensor([0.0411, 0.0110, 0.0170, 0.2110, 0.0111, 0.0075, 0.0225, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0083, 0.0065, 0.0062, 0.0149, 0.0065, 0.0058, 0.0068, 0.0082], device='cuda:5'), out_proj_covar=tensor([1.3122e-04, 1.0259e-04, 1.0517e-04, 2.2472e-04, 1.0830e-04, 9.5110e-05, 1.2391e-04, 1.3020e-04], device='cuda:5') 2023-04-27 15:01:42,495 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:02:27,368 INFO [train.py:904] (5/8) Epoch 1, batch 7200, loss[loss=0.295, simple_loss=0.3677, pruned_loss=0.1111, over 16197.00 frames. ], tot_loss[loss=0.3437, simple_loss=0.3967, pruned_loss=0.1454, over 3085246.10 frames. ], batch size: 165, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:02:46,735 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.108e+02 4.516e+02 5.501e+02 7.177e+02 1.508e+03, threshold=1.100e+03, percent-clipped=3.0 2023-04-27 15:02:57,915 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5214, 3.2534, 2.7728, 3.8854, 2.7139, 3.9063, 3.0744, 2.7872], device='cuda:5'), covar=tensor([0.0324, 0.0323, 0.0319, 0.0232, 0.1161, 0.0164, 0.0513, 0.1143], device='cuda:5'), in_proj_covar=tensor([0.0105, 0.0104, 0.0087, 0.0113, 0.0185, 0.0099, 0.0126, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:5') 2023-04-27 15:03:02,603 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:03:08,074 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3577, 3.8455, 4.1516, 4.3451, 3.6926, 4.1899, 4.1508, 4.0281], device='cuda:5'), covar=tensor([0.0316, 0.0228, 0.0192, 0.0125, 0.0782, 0.0202, 0.0266, 0.0183], device='cuda:5'), in_proj_covar=tensor([0.0092, 0.0065, 0.0117, 0.0091, 0.0143, 0.0093, 0.0084, 0.0097], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:03:19,583 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:03:45,811 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6062, 2.6611, 2.5303, 2.2271, 2.6205, 2.5522, 2.6383, 1.9310], device='cuda:5'), covar=tensor([0.1108, 0.0101, 0.0111, 0.0342, 0.0090, 0.0108, 0.0119, 0.0534], device='cuda:5'), in_proj_covar=tensor([0.0103, 0.0039, 0.0040, 0.0067, 0.0037, 0.0038, 0.0043, 0.0070], device='cuda:5'), out_proj_covar=tensor([1.9980e-04, 8.3387e-05, 8.9590e-05, 1.3780e-04, 7.9597e-05, 8.7925e-05, 8.9263e-05, 1.4363e-04], device='cuda:5') 2023-04-27 15:03:46,518 INFO [train.py:904] (5/8) Epoch 1, batch 7250, loss[loss=0.3336, simple_loss=0.3766, pruned_loss=0.1453, over 11640.00 frames. ], tot_loss[loss=0.3395, simple_loss=0.3931, pruned_loss=0.143, over 3076066.87 frames. ], batch size: 246, lr: 3.77e-02, grad_scale: 8.0 2023-04-27 15:04:06,715 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:21,317 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:34,480 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 15:04:35,896 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:58,552 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.22 vs. limit=5.0 2023-04-27 15:04:58,723 INFO [train.py:904] (5/8) Epoch 1, batch 7300, loss[loss=0.3212, simple_loss=0.3863, pruned_loss=0.1281, over 16876.00 frames. ], tot_loss[loss=0.3367, simple_loss=0.3911, pruned_loss=0.1411, over 3090904.75 frames. ], batch size: 90, lr: 3.76e-02, grad_scale: 8.0 2023-04-27 15:05:19,385 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.541e+02 5.301e+02 6.672e+02 8.069e+02 1.507e+03, threshold=1.334e+03, percent-clipped=6.0 2023-04-27 15:05:47,846 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2348, 1.3259, 1.5469, 1.1691, 2.0108, 1.9515, 2.2452, 2.1782], device='cuda:5'), covar=tensor([0.0045, 0.0392, 0.0200, 0.0246, 0.0086, 0.0160, 0.0081, 0.0085], device='cuda:5'), in_proj_covar=tensor([0.0030, 0.0065, 0.0044, 0.0047, 0.0040, 0.0046, 0.0031, 0.0035], device='cuda:5'), out_proj_covar=tensor([3.5894e-05, 9.6783e-05, 6.1850e-05, 6.5881e-05, 5.4812e-05, 6.4136e-05, 4.6737e-05, 5.0173e-05], device='cuda:5') 2023-04-27 15:05:49,057 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:53,682 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:15,336 INFO [train.py:904] (5/8) Epoch 1, batch 7350, loss[loss=0.3503, simple_loss=0.398, pruned_loss=0.1513, over 17261.00 frames. ], tot_loss[loss=0.3342, simple_loss=0.3892, pruned_loss=0.1396, over 3080767.96 frames. ], batch size: 52, lr: 3.75e-02, grad_scale: 8.0 2023-04-27 15:07:30,968 INFO [train.py:904] (5/8) Epoch 1, batch 7400, loss[loss=0.3415, simple_loss=0.4018, pruned_loss=0.1406, over 16896.00 frames. ], tot_loss[loss=0.3366, simple_loss=0.3909, pruned_loss=0.1411, over 3083633.80 frames. ], batch size: 96, lr: 3.74e-02, grad_scale: 8.0 2023-04-27 15:07:36,109 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:07:40,579 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:07:50,857 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.431e+02 4.999e+02 6.259e+02 7.546e+02 1.554e+03, threshold=1.252e+03, percent-clipped=1.0 2023-04-27 15:08:11,744 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 15:08:52,569 INFO [train.py:904] (5/8) Epoch 1, batch 7450, loss[loss=0.416, simple_loss=0.4349, pruned_loss=0.1986, over 11170.00 frames. ], tot_loss[loss=0.3383, simple_loss=0.392, pruned_loss=0.1422, over 3087345.33 frames. ], batch size: 247, lr: 3.73e-02, grad_scale: 8.0 2023-04-27 15:09:22,471 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:10:15,211 INFO [train.py:904] (5/8) Epoch 1, batch 7500, loss[loss=0.3238, simple_loss=0.3817, pruned_loss=0.133, over 16587.00 frames. ], tot_loss[loss=0.3394, simple_loss=0.3933, pruned_loss=0.1427, over 3081494.24 frames. ], batch size: 62, lr: 3.72e-02, grad_scale: 8.0 2023-04-27 15:10:35,039 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.564e+02 5.230e+02 6.451e+02 8.258e+02 1.814e+03, threshold=1.290e+03, percent-clipped=2.0 2023-04-27 15:10:43,275 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:11:06,859 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:11:31,907 INFO [train.py:904] (5/8) Epoch 1, batch 7550, loss[loss=0.3542, simple_loss=0.401, pruned_loss=0.1537, over 16576.00 frames. ], tot_loss[loss=0.3408, simple_loss=0.3937, pruned_loss=0.144, over 3058669.48 frames. ], batch size: 75, lr: 3.72e-02, grad_scale: 4.0 2023-04-27 15:11:51,688 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7443, 4.3512, 4.5977, 4.7777, 4.0164, 4.5494, 4.5241, 4.2772], device='cuda:5'), covar=tensor([0.0224, 0.0144, 0.0149, 0.0097, 0.0823, 0.0163, 0.0170, 0.0207], device='cuda:5'), in_proj_covar=tensor([0.0090, 0.0062, 0.0117, 0.0092, 0.0142, 0.0093, 0.0083, 0.0097], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:11:54,814 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:12:21,505 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:12:50,412 INFO [train.py:904] (5/8) Epoch 1, batch 7600, loss[loss=0.3214, simple_loss=0.3733, pruned_loss=0.1347, over 16707.00 frames. ], tot_loss[loss=0.3405, simple_loss=0.3926, pruned_loss=0.1442, over 3059192.81 frames. ], batch size: 134, lr: 3.71e-02, grad_scale: 8.0 2023-04-27 15:12:57,293 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3190, 1.5479, 1.8950, 2.6992, 2.7746, 2.7471, 1.7214, 2.6921], device='cuda:5'), covar=tensor([0.0078, 0.0718, 0.0296, 0.0134, 0.0100, 0.0107, 0.0316, 0.0104], device='cuda:5'), in_proj_covar=tensor([0.0041, 0.0088, 0.0066, 0.0049, 0.0042, 0.0042, 0.0060, 0.0039], device='cuda:5'), out_proj_covar=tensor([7.4231e-05, 1.5903e-04, 1.2633e-04, 9.1377e-05, 7.3827e-05, 7.5176e-05, 1.0012e-04, 7.1115e-05], device='cuda:5') 2023-04-27 15:13:10,076 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:12,791 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.798e+02 5.452e+02 6.510e+02 8.243e+02 1.443e+03, threshold=1.302e+03, percent-clipped=3.0 2023-04-27 15:13:29,798 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:40,409 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:14:13,735 INFO [train.py:904] (5/8) Epoch 1, batch 7650, loss[loss=0.3387, simple_loss=0.3923, pruned_loss=0.1426, over 16694.00 frames. ], tot_loss[loss=0.3426, simple_loss=0.3938, pruned_loss=0.1457, over 3055910.15 frames. ], batch size: 134, lr: 3.70e-02, grad_scale: 8.0 2023-04-27 15:14:36,277 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 15:15:05,720 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-04-27 15:15:14,458 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:15:39,936 INFO [train.py:904] (5/8) Epoch 1, batch 7700, loss[loss=0.3019, simple_loss=0.3654, pruned_loss=0.1192, over 16800.00 frames. ], tot_loss[loss=0.3435, simple_loss=0.394, pruned_loss=0.1465, over 3063389.88 frames. ], batch size: 83, lr: 3.69e-02, grad_scale: 8.0 2023-04-27 15:15:45,436 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:16:00,936 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.630e+02 5.474e+02 6.817e+02 8.632e+02 3.010e+03, threshold=1.363e+03, percent-clipped=1.0 2023-04-27 15:16:45,758 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:57,065 INFO [train.py:904] (5/8) Epoch 1, batch 7750, loss[loss=0.3773, simple_loss=0.4272, pruned_loss=0.1637, over 15427.00 frames. ], tot_loss[loss=0.3447, simple_loss=0.3952, pruned_loss=0.1471, over 3049735.99 frames. ], batch size: 190, lr: 3.68e-02, grad_scale: 8.0 2023-04-27 15:16:59,264 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:17:16,091 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:14,288 INFO [train.py:904] (5/8) Epoch 1, batch 7800, loss[loss=0.3233, simple_loss=0.3914, pruned_loss=0.1276, over 16493.00 frames. ], tot_loss[loss=0.3462, simple_loss=0.3963, pruned_loss=0.148, over 3056115.73 frames. ], batch size: 75, lr: 3.67e-02, grad_scale: 8.0 2023-04-27 15:18:16,729 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:19,983 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:36,282 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.181e+02 4.983e+02 6.289e+02 7.590e+02 1.248e+03, threshold=1.258e+03, percent-clipped=0.0 2023-04-27 15:18:43,837 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:19:07,171 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-04-27 15:19:31,869 INFO [train.py:904] (5/8) Epoch 1, batch 7850, loss[loss=0.3284, simple_loss=0.397, pruned_loss=0.1299, over 16508.00 frames. ], tot_loss[loss=0.344, simple_loss=0.3961, pruned_loss=0.1459, over 3080354.62 frames. ], batch size: 68, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:19:51,364 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7863.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:19:56,872 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:20:29,993 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 15:20:49,094 INFO [train.py:904] (5/8) Epoch 1, batch 7900, loss[loss=0.3014, simple_loss=0.3735, pruned_loss=0.1146, over 16739.00 frames. ], tot_loss[loss=0.3411, simple_loss=0.3941, pruned_loss=0.144, over 3087709.44 frames. ], batch size: 83, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:21:12,000 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.293e+02 4.952e+02 6.401e+02 7.784e+02 1.850e+03, threshold=1.280e+03, percent-clipped=3.0 2023-04-27 15:21:39,404 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:02,592 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 15:22:08,697 INFO [train.py:904] (5/8) Epoch 1, batch 7950, loss[loss=0.4179, simple_loss=0.4298, pruned_loss=0.203, over 11204.00 frames. ], tot_loss[loss=0.3411, simple_loss=0.3939, pruned_loss=0.1441, over 3086295.98 frames. ], batch size: 246, lr: 3.65e-02, grad_scale: 8.0 2023-04-27 15:22:31,050 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1314, 3.0985, 3.0530, 3.4248, 3.3901, 3.2674, 3.3494, 3.3634], device='cuda:5'), covar=tensor([0.0419, 0.0404, 0.1141, 0.0400, 0.0442, 0.0540, 0.0421, 0.0374], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0172, 0.0261, 0.0181, 0.0155, 0.0161, 0.0145, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:22:54,879 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:56,164 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:23:30,893 INFO [train.py:904] (5/8) Epoch 1, batch 8000, loss[loss=0.3101, simple_loss=0.3755, pruned_loss=0.1224, over 16892.00 frames. ], tot_loss[loss=0.3415, simple_loss=0.3936, pruned_loss=0.1447, over 3064157.00 frames. ], batch size: 116, lr: 3.64e-02, grad_scale: 8.0 2023-04-27 15:23:51,330 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.520e+02 5.080e+02 6.216e+02 7.580e+02 1.182e+03, threshold=1.243e+03, percent-clipped=0.0 2023-04-27 15:24:00,609 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-27 15:24:45,928 INFO [train.py:904] (5/8) Epoch 1, batch 8050, loss[loss=0.318, simple_loss=0.3874, pruned_loss=0.1243, over 16500.00 frames. ], tot_loss[loss=0.3417, simple_loss=0.3936, pruned_loss=0.1449, over 3054874.21 frames. ], batch size: 68, lr: 3.63e-02, grad_scale: 8.0 2023-04-27 15:25:04,699 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:57,840 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:00,733 INFO [train.py:904] (5/8) Epoch 1, batch 8100, loss[loss=0.3347, simple_loss=0.3799, pruned_loss=0.1447, over 16326.00 frames. ], tot_loss[loss=0.3395, simple_loss=0.3923, pruned_loss=0.1433, over 3073410.78 frames. ], batch size: 35, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:26:15,158 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:22,781 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.844e+02 5.411e+02 6.350e+02 7.571e+02 1.310e+03, threshold=1.270e+03, percent-clipped=1.0 2023-04-27 15:27:16,455 INFO [train.py:904] (5/8) Epoch 1, batch 8150, loss[loss=0.3014, simple_loss=0.3628, pruned_loss=0.12, over 16631.00 frames. ], tot_loss[loss=0.3372, simple_loss=0.3903, pruned_loss=0.142, over 3080601.63 frames. ], batch size: 57, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:27:28,064 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:28:33,141 INFO [train.py:904] (5/8) Epoch 1, batch 8200, loss[loss=0.3861, simple_loss=0.4167, pruned_loss=0.1777, over 11521.00 frames. ], tot_loss[loss=0.3356, simple_loss=0.3884, pruned_loss=0.1414, over 3076153.41 frames. ], batch size: 247, lr: 3.61e-02, grad_scale: 4.0 2023-04-27 15:28:56,622 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.505e+02 5.388e+02 6.507e+02 7.835e+02 3.023e+03, threshold=1.301e+03, percent-clipped=3.0 2023-04-27 15:29:16,771 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0946, 3.7662, 3.6893, 3.4362, 3.9936, 2.3011, 3.7083, 3.8974], device='cuda:5'), covar=tensor([0.0058, 0.0080, 0.0073, 0.0235, 0.0042, 0.0820, 0.0066, 0.0084], device='cuda:5'), in_proj_covar=tensor([0.0048, 0.0039, 0.0056, 0.0077, 0.0040, 0.0083, 0.0055, 0.0053], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:29:53,200 INFO [train.py:904] (5/8) Epoch 1, batch 8250, loss[loss=0.3288, simple_loss=0.392, pruned_loss=0.1328, over 15330.00 frames. ], tot_loss[loss=0.3323, simple_loss=0.387, pruned_loss=0.1388, over 3084005.54 frames. ], batch size: 191, lr: 3.60e-02, grad_scale: 4.0 2023-04-27 15:30:15,989 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:41,226 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:14,617 INFO [train.py:904] (5/8) Epoch 1, batch 8300, loss[loss=0.2685, simple_loss=0.3548, pruned_loss=0.0911, over 16296.00 frames. ], tot_loss[loss=0.3235, simple_loss=0.3811, pruned_loss=0.1329, over 3059855.93 frames. ], batch size: 165, lr: 3.59e-02, grad_scale: 4.0 2023-04-27 15:31:17,766 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5710, 4.2885, 1.9732, 4.4641, 2.6665, 4.3760, 2.5026, 3.3992], device='cuda:5'), covar=tensor([0.0036, 0.0155, 0.1432, 0.0041, 0.0774, 0.0150, 0.1089, 0.0448], device='cuda:5'), in_proj_covar=tensor([0.0061, 0.0069, 0.0147, 0.0065, 0.0124, 0.0074, 0.0151, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:31:40,148 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.807e+02 4.177e+02 5.020e+02 6.033e+02 1.438e+03, threshold=1.004e+03, percent-clipped=1.0 2023-04-27 15:31:47,612 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1825, 3.1366, 2.8899, 1.9049, 2.7336, 2.0795, 2.9105, 3.1355], device='cuda:5'), covar=tensor([0.0228, 0.0341, 0.0331, 0.1804, 0.0775, 0.1246, 0.0780, 0.0248], device='cuda:5'), in_proj_covar=tensor([0.0100, 0.0078, 0.0116, 0.0153, 0.0147, 0.0139, 0.0135, 0.0070], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:31:56,760 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:59,950 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:33,505 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 15:32:36,035 INFO [train.py:904] (5/8) Epoch 1, batch 8350, loss[loss=0.268, simple_loss=0.3465, pruned_loss=0.09474, over 16681.00 frames. ], tot_loss[loss=0.3156, simple_loss=0.3767, pruned_loss=0.1272, over 3068970.99 frames. ], batch size: 83, lr: 3.58e-02, grad_scale: 4.0 2023-04-27 15:33:55,127 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:57,944 INFO [train.py:904] (5/8) Epoch 1, batch 8400, loss[loss=0.2852, simple_loss=0.3438, pruned_loss=0.1133, over 12023.00 frames. ], tot_loss[loss=0.309, simple_loss=0.3719, pruned_loss=0.1231, over 3049695.60 frames. ], batch size: 246, lr: 3.58e-02, grad_scale: 8.0 2023-04-27 15:34:21,545 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.723e+02 4.252e+02 5.243e+02 6.006e+02 1.213e+03, threshold=1.049e+03, percent-clipped=1.0 2023-04-27 15:35:12,414 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:18,239 INFO [train.py:904] (5/8) Epoch 1, batch 8450, loss[loss=0.2486, simple_loss=0.3302, pruned_loss=0.08353, over 16672.00 frames. ], tot_loss[loss=0.305, simple_loss=0.3691, pruned_loss=0.1205, over 3039935.96 frames. ], batch size: 134, lr: 3.57e-02, grad_scale: 8.0 2023-04-27 15:35:30,891 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:36:39,483 INFO [train.py:904] (5/8) Epoch 1, batch 8500, loss[loss=0.2542, simple_loss=0.3335, pruned_loss=0.08745, over 16671.00 frames. ], tot_loss[loss=0.2958, simple_loss=0.3621, pruned_loss=0.1148, over 3052573.26 frames. ], batch size: 76, lr: 3.56e-02, grad_scale: 8.0 2023-04-27 15:36:49,163 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:37:04,606 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.878e+02 4.840e+02 6.056e+02 7.383e+02 1.593e+03, threshold=1.211e+03, percent-clipped=7.0 2023-04-27 15:38:03,992 INFO [train.py:904] (5/8) Epoch 1, batch 8550, loss[loss=0.3296, simple_loss=0.3961, pruned_loss=0.1316, over 15328.00 frames. ], tot_loss[loss=0.292, simple_loss=0.3587, pruned_loss=0.1126, over 3042813.90 frames. ], batch size: 190, lr: 3.55e-02, grad_scale: 8.0 2023-04-27 15:38:56,849 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4875, 3.5200, 3.7943, 3.8412, 3.9461, 3.5938, 3.6575, 3.7970], device='cuda:5'), covar=tensor([0.0296, 0.0289, 0.0436, 0.0396, 0.0414, 0.0301, 0.0650, 0.0293], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0111, 0.0128, 0.0129, 0.0138, 0.0118, 0.0165, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:39:44,574 INFO [train.py:904] (5/8) Epoch 1, batch 8600, loss[loss=0.3003, simple_loss=0.3589, pruned_loss=0.1209, over 12430.00 frames. ], tot_loss[loss=0.2918, simple_loss=0.3592, pruned_loss=0.1122, over 3030050.57 frames. ], batch size: 248, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:40:17,886 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.023e+02 4.189e+02 5.220e+02 6.302e+02 1.296e+03, threshold=1.044e+03, percent-clipped=1.0 2023-04-27 15:40:28,419 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:41:26,191 INFO [train.py:904] (5/8) Epoch 1, batch 8650, loss[loss=0.2459, simple_loss=0.3333, pruned_loss=0.07925, over 16808.00 frames. ], tot_loss[loss=0.2871, simple_loss=0.3563, pruned_loss=0.109, over 3040869.48 frames. ], batch size: 124, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:42:24,181 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4923, 3.5044, 3.5438, 3.8942, 3.8042, 3.7279, 3.8736, 3.8025], device='cuda:5'), covar=tensor([0.0405, 0.0360, 0.0918, 0.0363, 0.0471, 0.0459, 0.0309, 0.0264], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0161, 0.0246, 0.0176, 0.0148, 0.0155, 0.0137, 0.0148], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:43:13,575 INFO [train.py:904] (5/8) Epoch 1, batch 8700, loss[loss=0.2467, simple_loss=0.3285, pruned_loss=0.08246, over 16362.00 frames. ], tot_loss[loss=0.2824, simple_loss=0.3522, pruned_loss=0.1063, over 3056149.32 frames. ], batch size: 146, lr: 3.53e-02, grad_scale: 8.0 2023-04-27 15:43:41,222 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.978e+02 4.653e+02 5.603e+02 6.924e+02 1.986e+03, threshold=1.121e+03, percent-clipped=4.0 2023-04-27 15:44:21,308 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0707, 4.2936, 4.2721, 4.1917, 4.3234, 4.6496, 4.6441, 4.2228], device='cuda:5'), covar=tensor([0.1076, 0.0911, 0.0705, 0.1449, 0.1687, 0.0715, 0.0561, 0.1706], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0206, 0.0169, 0.0175, 0.0213, 0.0174, 0.0149, 0.0226], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-27 15:44:49,927 INFO [train.py:904] (5/8) Epoch 1, batch 8750, loss[loss=0.2819, simple_loss=0.3656, pruned_loss=0.09905, over 16381.00 frames. ], tot_loss[loss=0.2809, simple_loss=0.3519, pruned_loss=0.105, over 3073872.74 frames. ], batch size: 146, lr: 3.52e-02, grad_scale: 8.0 2023-04-27 15:45:14,454 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:45:47,878 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7526, 3.4861, 2.8096, 3.9848, 2.4483, 3.9179, 2.8248, 2.3968], device='cuda:5'), covar=tensor([0.0270, 0.0245, 0.0286, 0.0239, 0.1463, 0.0165, 0.0579, 0.1359], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0123, 0.0098, 0.0133, 0.0199, 0.0117, 0.0139, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:46:42,415 INFO [train.py:904] (5/8) Epoch 1, batch 8800, loss[loss=0.2548, simple_loss=0.3334, pruned_loss=0.08804, over 16113.00 frames. ], tot_loss[loss=0.278, simple_loss=0.349, pruned_loss=0.1035, over 3058799.32 frames. ], batch size: 165, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:47:13,496 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.267e+02 4.433e+02 5.797e+02 7.279e+02 2.542e+03, threshold=1.159e+03, percent-clipped=4.0 2023-04-27 15:47:25,173 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:48:27,337 INFO [train.py:904] (5/8) Epoch 1, batch 8850, loss[loss=0.275, simple_loss=0.3605, pruned_loss=0.09472, over 16354.00 frames. ], tot_loss[loss=0.2772, simple_loss=0.3505, pruned_loss=0.1019, over 3061626.77 frames. ], batch size: 146, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:50:04,277 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0160, 3.6438, 2.9198, 4.3005, 2.7076, 4.1335, 2.9730, 2.7086], device='cuda:5'), covar=tensor([0.0240, 0.0231, 0.0259, 0.0200, 0.1148, 0.0166, 0.0517, 0.1151], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0124, 0.0099, 0.0135, 0.0199, 0.0119, 0.0138, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:50:13,551 INFO [train.py:904] (5/8) Epoch 1, batch 8900, loss[loss=0.3196, simple_loss=0.3872, pruned_loss=0.126, over 16946.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.3503, pruned_loss=0.1007, over 3068360.28 frames. ], batch size: 116, lr: 3.50e-02, grad_scale: 8.0 2023-04-27 15:50:42,918 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.973e+02 4.392e+02 5.224e+02 6.190e+02 1.359e+03, threshold=1.045e+03, percent-clipped=1.0 2023-04-27 15:50:56,822 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:13,687 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:17,782 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2111, 1.4574, 1.8616, 2.2454, 2.1096, 2.1120, 1.5823, 2.2971], device='cuda:5'), covar=tensor([0.0059, 0.0539, 0.0233, 0.0145, 0.0072, 0.0151, 0.0347, 0.0076], device='cuda:5'), in_proj_covar=tensor([0.0045, 0.0089, 0.0071, 0.0053, 0.0041, 0.0044, 0.0069, 0.0040], device='cuda:5'), out_proj_covar=tensor([8.3076e-05, 1.6186e-04, 1.3679e-04, 9.8268e-05, 7.4563e-05, 8.2637e-05, 1.1994e-04, 7.3995e-05], device='cuda:5') 2023-04-27 15:52:19,953 INFO [train.py:904] (5/8) Epoch 1, batch 8950, loss[loss=0.2414, simple_loss=0.336, pruned_loss=0.07343, over 16782.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3502, pruned_loss=0.101, over 3084612.04 frames. ], batch size: 83, lr: 3.49e-02, grad_scale: 8.0 2023-04-27 15:53:00,569 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:24,962 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-04-27 15:53:38,252 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:42,533 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:08,286 INFO [train.py:904] (5/8) Epoch 1, batch 9000, loss[loss=0.2494, simple_loss=0.328, pruned_loss=0.08542, over 15275.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3456, pruned_loss=0.09807, over 3086597.10 frames. ], batch size: 191, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:54:08,286 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 15:54:19,193 INFO [train.py:938] (5/8) Epoch 1, validation: loss=0.2299, simple_loss=0.3267, pruned_loss=0.06658, over 944034.00 frames. 2023-04-27 15:54:19,194 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-27 15:54:46,245 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9013.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:51,598 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.333e+02 3.931e+02 4.892e+02 5.908e+02 1.148e+03, threshold=9.783e+02, percent-clipped=2.0 2023-04-27 15:54:56,771 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5586, 4.5935, 4.8853, 4.9768, 5.0364, 4.6276, 4.7290, 4.8424], device='cuda:5'), covar=tensor([0.0235, 0.0208, 0.0362, 0.0374, 0.0299, 0.0213, 0.0531, 0.0203], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0110, 0.0126, 0.0126, 0.0140, 0.0116, 0.0164, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:55:56,645 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:56:02,944 INFO [train.py:904] (5/8) Epoch 1, batch 9050, loss[loss=0.261, simple_loss=0.3318, pruned_loss=0.09508, over 16645.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.349, pruned_loss=0.1008, over 3092002.45 frames. ], batch size: 134, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:56:31,976 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 2023-04-27 15:56:48,966 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:57:30,881 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.42 vs. limit=5.0 2023-04-27 15:57:45,082 INFO [train.py:904] (5/8) Epoch 1, batch 9100, loss[loss=0.2834, simple_loss=0.3617, pruned_loss=0.1025, over 16378.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3488, pruned_loss=0.1011, over 3099804.72 frames. ], batch size: 146, lr: 3.47e-02, grad_scale: 8.0 2023-04-27 15:58:03,598 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 15:58:13,310 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1327, 3.7535, 4.0288, 4.1942, 3.5808, 4.0274, 3.9020, 3.8342], device='cuda:5'), covar=tensor([0.0316, 0.0190, 0.0179, 0.0101, 0.0538, 0.0180, 0.0289, 0.0193], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0064, 0.0114, 0.0089, 0.0135, 0.0089, 0.0080, 0.0096], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 15:58:14,296 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.937e+02 4.687e+02 5.823e+02 7.250e+02 1.575e+03, threshold=1.165e+03, percent-clipped=5.0 2023-04-27 15:58:14,906 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:42,641 INFO [train.py:904] (5/8) Epoch 1, batch 9150, loss[loss=0.2621, simple_loss=0.338, pruned_loss=0.09308, over 16992.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3488, pruned_loss=0.1003, over 3095231.79 frames. ], batch size: 55, lr: 3.46e-02, grad_scale: 8.0 2023-04-27 16:01:27,887 INFO [train.py:904] (5/8) Epoch 1, batch 9200, loss[loss=0.2804, simple_loss=0.3545, pruned_loss=0.1031, over 16350.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3432, pruned_loss=0.09825, over 3097304.55 frames. ], batch size: 146, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:01:54,929 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.438e+02 5.266e+02 6.436e+02 8.220e+02 2.129e+03, threshold=1.287e+03, percent-clipped=5.0 2023-04-27 16:03:04,220 INFO [train.py:904] (5/8) Epoch 1, batch 9250, loss[loss=0.2735, simple_loss=0.3317, pruned_loss=0.1076, over 12694.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3421, pruned_loss=0.09767, over 3087850.58 frames. ], batch size: 248, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:03:06,254 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 16:04:17,165 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:04:58,127 INFO [train.py:904] (5/8) Epoch 1, batch 9300, loss[loss=0.2271, simple_loss=0.3047, pruned_loss=0.07473, over 16917.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3387, pruned_loss=0.09572, over 3078998.16 frames. ], batch size: 116, lr: 3.44e-02, grad_scale: 8.0 2023-04-27 16:05:33,425 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.238e+02 3.939e+02 4.601e+02 5.465e+02 1.094e+03, threshold=9.201e+02, percent-clipped=0.0 2023-04-27 16:06:27,898 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:06:44,332 INFO [train.py:904] (5/8) Epoch 1, batch 9350, loss[loss=0.2612, simple_loss=0.3253, pruned_loss=0.09858, over 12304.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3391, pruned_loss=0.09578, over 3075010.15 frames. ], batch size: 248, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:07:24,451 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:07:31,000 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8681, 4.1350, 4.3399, 4.4070, 4.5228, 4.0531, 3.9639, 4.2979], device='cuda:5'), covar=tensor([0.0446, 0.0377, 0.0530, 0.0546, 0.0430, 0.0399, 0.0844, 0.0294], device='cuda:5'), in_proj_covar=tensor([0.0113, 0.0108, 0.0120, 0.0122, 0.0134, 0.0110, 0.0152, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:07:31,093 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7735, 3.8072, 3.7140, 1.6700, 3.8548, 3.8385, 3.1871, 3.4009], device='cuda:5'), covar=tensor([0.0433, 0.0069, 0.0149, 0.1730, 0.0077, 0.0076, 0.0285, 0.0200], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0068, 0.0067, 0.0143, 0.0061, 0.0065, 0.0078, 0.0088], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:08:26,600 INFO [train.py:904] (5/8) Epoch 1, batch 9400, loss[loss=0.2361, simple_loss=0.3127, pruned_loss=0.07972, over 12300.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3388, pruned_loss=0.09487, over 3078639.25 frames. ], batch size: 248, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:08:54,866 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:08:56,917 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.254e+02 4.024e+02 4.732e+02 5.961e+02 1.474e+03, threshold=9.465e+02, percent-clipped=2.0 2023-04-27 16:08:58,008 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:08,181 INFO [train.py:904] (5/8) Epoch 1, batch 9450, loss[loss=0.2627, simple_loss=0.3316, pruned_loss=0.09683, over 12319.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3409, pruned_loss=0.09548, over 3074796.54 frames. ], batch size: 246, lr: 3.42e-02, grad_scale: 8.0 2023-04-27 16:10:33,962 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:58,517 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:50,040 INFO [train.py:904] (5/8) Epoch 1, batch 9500, loss[loss=0.24, simple_loss=0.3076, pruned_loss=0.08614, over 12918.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3389, pruned_loss=0.09437, over 3061549.34 frames. ], batch size: 248, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:12:21,414 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.089e+02 4.155e+02 5.314e+02 6.779e+02 1.064e+03, threshold=1.063e+03, percent-clipped=4.0 2023-04-27 16:13:37,019 INFO [train.py:904] (5/8) Epoch 1, batch 9550, loss[loss=0.2642, simple_loss=0.3326, pruned_loss=0.09785, over 12903.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3382, pruned_loss=0.09474, over 3044193.92 frames. ], batch size: 248, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:14:46,675 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:14:48,703 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:15:18,634 INFO [train.py:904] (5/8) Epoch 1, batch 9600, loss[loss=0.3051, simple_loss=0.3768, pruned_loss=0.1167, over 15390.00 frames. ], tot_loss[loss=0.266, simple_loss=0.3402, pruned_loss=0.09594, over 3044631.06 frames. ], batch size: 191, lr: 3.40e-02, grad_scale: 8.0 2023-04-27 16:15:48,624 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 4.688e+02 5.707e+02 6.655e+02 1.542e+03, threshold=1.141e+03, percent-clipped=4.0 2023-04-27 16:16:20,307 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:45,981 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:53,492 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:08,565 INFO [train.py:904] (5/8) Epoch 1, batch 9650, loss[loss=0.2642, simple_loss=0.343, pruned_loss=0.0927, over 16171.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3428, pruned_loss=0.09671, over 3043382.21 frames. ], batch size: 165, lr: 3.39e-02, grad_scale: 8.0 2023-04-27 16:17:26,834 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:52,735 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:35,171 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:57,055 INFO [train.py:904] (5/8) Epoch 1, batch 9700, loss[loss=0.302, simple_loss=0.3727, pruned_loss=0.1156, over 16696.00 frames. ], tot_loss[loss=0.2675, simple_loss=0.3416, pruned_loss=0.09663, over 3037754.31 frames. ], batch size: 134, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:19:20,001 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 16:19:23,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3355, 4.6240, 4.5819, 4.5846, 4.6532, 5.0767, 4.9014, 4.5707], device='cuda:5'), covar=tensor([0.0701, 0.1051, 0.0832, 0.1278, 0.1574, 0.0614, 0.0625, 0.1398], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0216, 0.0180, 0.0186, 0.0222, 0.0178, 0.0154, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-27 16:19:24,741 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.013e+02 4.098e+02 5.059e+02 6.165e+02 1.399e+03, threshold=1.012e+03, percent-clipped=2.0 2023-04-27 16:19:28,144 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:33,573 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:49,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-04-27 16:20:20,898 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1437, 4.5683, 4.3240, 4.5116, 3.9915, 4.0387, 4.1932, 4.5678], device='cuda:5'), covar=tensor([0.0478, 0.0668, 0.0618, 0.0250, 0.0510, 0.0487, 0.0429, 0.0489], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0218, 0.0192, 0.0129, 0.0160, 0.0131, 0.0178, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:20:27,200 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:20:40,567 INFO [train.py:904] (5/8) Epoch 1, batch 9750, loss[loss=0.2552, simple_loss=0.3403, pruned_loss=0.08508, over 16336.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3407, pruned_loss=0.09668, over 3037764.18 frames. ], batch size: 146, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:21:17,390 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:21:21,071 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 16:22:19,301 INFO [train.py:904] (5/8) Epoch 1, batch 9800, loss[loss=0.2669, simple_loss=0.3523, pruned_loss=0.09078, over 15288.00 frames. ], tot_loss[loss=0.2654, simple_loss=0.3407, pruned_loss=0.09503, over 3061948.23 frames. ], batch size: 191, lr: 3.37e-02, grad_scale: 8.0 2023-04-27 16:22:26,937 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:47,906 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.893e+02 4.491e+02 5.387e+02 6.683e+02 1.234e+03, threshold=1.077e+03, percent-clipped=5.0 2023-04-27 16:23:14,777 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2036, 4.3934, 4.3596, 4.3742, 4.3669, 4.7978, 4.6905, 4.3262], device='cuda:5'), covar=tensor([0.0817, 0.1096, 0.0825, 0.1382, 0.1865, 0.0803, 0.0829, 0.1877], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0219, 0.0184, 0.0191, 0.0226, 0.0192, 0.0155, 0.0236], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:24:05,894 INFO [train.py:904] (5/8) Epoch 1, batch 9850, loss[loss=0.2714, simple_loss=0.3485, pruned_loss=0.09718, over 16322.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3413, pruned_loss=0.0939, over 3095224.30 frames. ], batch size: 146, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:25:58,737 INFO [train.py:904] (5/8) Epoch 1, batch 9900, loss[loss=0.2613, simple_loss=0.3502, pruned_loss=0.08625, over 16915.00 frames. ], tot_loss[loss=0.2634, simple_loss=0.3407, pruned_loss=0.09308, over 3084781.58 frames. ], batch size: 116, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:26:31,900 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.967e+02 4.199e+02 5.185e+02 6.475e+02 1.120e+03, threshold=1.037e+03, percent-clipped=1.0 2023-04-27 16:27:32,795 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:27:57,277 INFO [train.py:904] (5/8) Epoch 1, batch 9950, loss[loss=0.2519, simple_loss=0.3323, pruned_loss=0.08582, over 16521.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3418, pruned_loss=0.09293, over 3083051.58 frames. ], batch size: 68, lr: 3.35e-02, grad_scale: 8.0 2023-04-27 16:29:39,901 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7044, 3.7809, 3.7858, 3.8352, 3.9701, 4.2023, 4.2733, 3.9058], device='cuda:5'), covar=tensor([0.1823, 0.1461, 0.0922, 0.1179, 0.1797, 0.0893, 0.0565, 0.1743], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0208, 0.0175, 0.0177, 0.0214, 0.0180, 0.0150, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-27 16:30:02,400 INFO [train.py:904] (5/8) Epoch 1, batch 10000, loss[loss=0.2639, simple_loss=0.3291, pruned_loss=0.09937, over 12644.00 frames. ], tot_loss[loss=0.2615, simple_loss=0.3395, pruned_loss=0.09176, over 3099534.00 frames. ], batch size: 250, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:30:30,575 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:30:31,032 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-27 16:30:32,719 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.207e+02 4.345e+02 5.531e+02 6.703e+02 1.367e+03, threshold=1.106e+03, percent-clipped=3.0 2023-04-27 16:31:22,510 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:31:41,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8481, 3.8981, 2.6033, 4.9032, 4.7021, 4.5003, 2.6235, 3.7732], device='cuda:5'), covar=tensor([0.1780, 0.0299, 0.1465, 0.0055, 0.0071, 0.0215, 0.0987, 0.0419], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0099, 0.0160, 0.0060, 0.0070, 0.0084, 0.0142, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:31:43,691 INFO [train.py:904] (5/8) Epoch 1, batch 10050, loss[loss=0.3049, simple_loss=0.3671, pruned_loss=0.1214, over 12121.00 frames. ], tot_loss[loss=0.2606, simple_loss=0.339, pruned_loss=0.0911, over 3099946.71 frames. ], batch size: 248, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:32:06,766 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:32:23,186 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:16,692 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:18,972 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:33:19,592 INFO [train.py:904] (5/8) Epoch 1, batch 10100, loss[loss=0.2647, simple_loss=0.3397, pruned_loss=0.09489, over 16234.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3404, pruned_loss=0.09224, over 3098829.32 frames. ], batch size: 165, lr: 3.33e-02, grad_scale: 16.0 2023-04-27 16:33:49,435 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.152e+02 4.595e+02 5.786e+02 6.822e+02 2.551e+03, threshold=1.157e+03, percent-clipped=1.0 2023-04-27 16:33:53,663 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:34:06,031 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:34:32,992 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2195, 3.0478, 2.6562, 2.2776, 1.8972, 1.9827, 2.9352, 3.2590], device='cuda:5'), covar=tensor([0.1483, 0.0660, 0.1032, 0.0428, 0.1750, 0.1565, 0.0367, 0.0156], device='cuda:5'), in_proj_covar=tensor([0.0219, 0.0197, 0.0213, 0.0133, 0.0173, 0.0166, 0.0150, 0.0083], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 16:35:05,143 INFO [train.py:904] (5/8) Epoch 2, batch 0, loss[loss=0.4421, simple_loss=0.4501, pruned_loss=0.217, over 16263.00 frames. ], tot_loss[loss=0.4421, simple_loss=0.4501, pruned_loss=0.217, over 16263.00 frames. ], batch size: 165, lr: 3.26e-02, grad_scale: 8.0 2023-04-27 16:35:05,143 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 16:35:12,744 INFO [train.py:938] (5/8) Epoch 2, validation: loss=0.2235, simple_loss=0.3221, pruned_loss=0.06242, over 944034.00 frames. 2023-04-27 16:35:12,745 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-27 16:35:14,594 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-27 16:35:46,416 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9915, 1.5392, 1.5534, 1.2245, 1.7427, 1.6909, 1.5346, 1.9509], device='cuda:5'), covar=tensor([0.0029, 0.0182, 0.0125, 0.0151, 0.0084, 0.0112, 0.0056, 0.0059], device='cuda:5'), in_proj_covar=tensor([0.0036, 0.0077, 0.0065, 0.0071, 0.0062, 0.0069, 0.0039, 0.0046], device='cuda:5'), out_proj_covar=tensor([5.0514e-05, 1.1865e-04, 9.9956e-05, 1.1038e-04, 9.5770e-05, 1.0476e-04, 5.9811e-05, 7.5628e-05], device='cuda:5') 2023-04-27 16:36:22,866 INFO [train.py:904] (5/8) Epoch 2, batch 50, loss[loss=0.2448, simple_loss=0.3178, pruned_loss=0.08587, over 17014.00 frames. ], tot_loss[loss=0.3311, simple_loss=0.3758, pruned_loss=0.1432, over 746293.87 frames. ], batch size: 41, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:36:45,681 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.307e+02 5.068e+02 6.475e+02 7.712e+02 1.660e+03, threshold=1.295e+03, percent-clipped=4.0 2023-04-27 16:37:16,883 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:37:30,897 INFO [train.py:904] (5/8) Epoch 2, batch 100, loss[loss=0.2634, simple_loss=0.3311, pruned_loss=0.09782, over 17221.00 frames. ], tot_loss[loss=0.3113, simple_loss=0.3636, pruned_loss=0.1295, over 1321574.20 frames. ], batch size: 44, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:38:12,006 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2800, 3.4200, 2.7855, 4.6710, 2.6843, 4.4222, 3.1243, 2.6389], device='cuda:5'), covar=tensor([0.0234, 0.0292, 0.0284, 0.0193, 0.1252, 0.0129, 0.0476, 0.1259], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0136, 0.0109, 0.0157, 0.0214, 0.0128, 0.0150, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:38:22,579 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:38:38,651 INFO [train.py:904] (5/8) Epoch 2, batch 150, loss[loss=0.3555, simple_loss=0.3872, pruned_loss=0.1619, over 16752.00 frames. ], tot_loss[loss=0.3072, simple_loss=0.3614, pruned_loss=0.1265, over 1769295.26 frames. ], batch size: 124, lr: 3.24e-02, grad_scale: 4.0 2023-04-27 16:38:40,094 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5242, 4.3111, 4.5085, 4.7420, 3.5299, 4.4881, 4.6178, 4.1102], device='cuda:5'), covar=tensor([0.0477, 0.0268, 0.0325, 0.0182, 0.1389, 0.0284, 0.0280, 0.0303], device='cuda:5'), in_proj_covar=tensor([0.0105, 0.0075, 0.0137, 0.0106, 0.0169, 0.0109, 0.0097, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:38:56,220 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:01,496 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.010e+02 4.543e+02 5.695e+02 6.778e+02 1.541e+03, threshold=1.139e+03, percent-clipped=2.0 2023-04-27 16:39:15,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0520, 4.3178, 3.4745, 3.8177, 2.7198, 2.2686, 4.6389, 5.2046], device='cuda:5'), covar=tensor([0.1779, 0.0588, 0.0964, 0.0368, 0.2287, 0.1381, 0.0224, 0.0041], device='cuda:5'), in_proj_covar=tensor([0.0228, 0.0201, 0.0218, 0.0137, 0.0197, 0.0167, 0.0154, 0.0085], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 16:39:17,116 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 16:39:29,259 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9322, 3.9125, 3.8201, 2.9768, 4.0091, 3.6473, 4.0595, 2.0966], device='cuda:5'), covar=tensor([0.0994, 0.0078, 0.0094, 0.0371, 0.0052, 0.0107, 0.0073, 0.0748], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0050, 0.0058, 0.0094, 0.0052, 0.0055, 0.0057, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:5') 2023-04-27 16:39:47,465 INFO [train.py:904] (5/8) Epoch 2, batch 200, loss[loss=0.2605, simple_loss=0.332, pruned_loss=0.09455, over 17045.00 frames. ], tot_loss[loss=0.3022, simple_loss=0.3585, pruned_loss=0.123, over 2116221.83 frames. ], batch size: 55, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:40:03,411 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:29,365 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 16:40:50,472 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:40:55,179 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:58,067 INFO [train.py:904] (5/8) Epoch 2, batch 250, loss[loss=0.3223, simple_loss=0.3557, pruned_loss=0.1445, over 16810.00 frames. ], tot_loss[loss=0.2985, simple_loss=0.3534, pruned_loss=0.1217, over 2382813.78 frames. ], batch size: 116, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:41:10,964 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:41:17,467 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3449, 5.0579, 5.2596, 5.4511, 4.6083, 5.2160, 5.1878, 4.9711], device='cuda:5'), covar=tensor([0.0316, 0.0184, 0.0166, 0.0093, 0.0811, 0.0201, 0.0138, 0.0204], device='cuda:5'), in_proj_covar=tensor([0.0110, 0.0080, 0.0146, 0.0113, 0.0179, 0.0114, 0.0101, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 16:41:22,655 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.115e+02 4.151e+02 5.077e+02 6.310e+02 1.196e+03, threshold=1.015e+03, percent-clipped=1.0 2023-04-27 16:41:22,944 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:41:51,847 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-04-27 16:42:03,850 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:09,614 INFO [train.py:904] (5/8) Epoch 2, batch 300, loss[loss=0.283, simple_loss=0.3373, pruned_loss=0.1143, over 15506.00 frames. ], tot_loss[loss=0.2908, simple_loss=0.3485, pruned_loss=0.1166, over 2595304.01 frames. ], batch size: 190, lr: 3.22e-02, grad_scale: 4.0 2023-04-27 16:42:14,812 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:29,318 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4555, 4.6724, 5.0232, 5.0679, 5.1887, 4.6171, 4.6468, 4.8377], device='cuda:5'), covar=tensor([0.0431, 0.0254, 0.0633, 0.0534, 0.0472, 0.0317, 0.0803, 0.0298], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0126, 0.0146, 0.0151, 0.0167, 0.0138, 0.0209, 0.0137], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-04-27 16:42:36,281 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10470.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:43,786 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7052, 4.4119, 4.3560, 1.8633, 4.5022, 4.6025, 3.5128, 3.5046], device='cuda:5'), covar=tensor([0.0628, 0.0069, 0.0147, 0.1692, 0.0084, 0.0064, 0.0261, 0.0243], device='cuda:5'), in_proj_covar=tensor([0.0108, 0.0072, 0.0071, 0.0148, 0.0065, 0.0064, 0.0085, 0.0096], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:43:11,411 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6800, 4.7166, 4.5637, 2.1750, 4.7584, 4.8419, 3.8166, 3.8048], device='cuda:5'), covar=tensor([0.0664, 0.0061, 0.0131, 0.1605, 0.0060, 0.0033, 0.0221, 0.0231], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0074, 0.0074, 0.0151, 0.0067, 0.0064, 0.0087, 0.0097], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:43:19,879 INFO [train.py:904] (5/8) Epoch 2, batch 350, loss[loss=0.2624, simple_loss=0.3406, pruned_loss=0.09212, over 17073.00 frames. ], tot_loss[loss=0.2841, simple_loss=0.3432, pruned_loss=0.1125, over 2757837.54 frames. ], batch size: 53, lr: 3.21e-02, grad_scale: 4.0 2023-04-27 16:43:39,923 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:43:42,893 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.580e+02 4.016e+02 4.975e+02 5.908e+02 1.216e+03, threshold=9.949e+02, percent-clipped=2.0 2023-04-27 16:44:27,080 INFO [train.py:904] (5/8) Epoch 2, batch 400, loss[loss=0.3199, simple_loss=0.3527, pruned_loss=0.1435, over 16476.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.3418, pruned_loss=0.1125, over 2881386.09 frames. ], batch size: 75, lr: 3.21e-02, grad_scale: 8.0 2023-04-27 16:44:33,228 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6753, 4.4362, 4.5792, 4.6896, 4.0492, 4.5784, 4.5491, 4.2533], device='cuda:5'), covar=tensor([0.0289, 0.0173, 0.0171, 0.0123, 0.0715, 0.0169, 0.0217, 0.0222], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0084, 0.0154, 0.0121, 0.0188, 0.0120, 0.0107, 0.0136], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 16:44:37,586 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1711, 4.3138, 3.6692, 3.6028, 3.0117, 2.3477, 4.6265, 5.1915], device='cuda:5'), covar=tensor([0.1616, 0.0530, 0.0903, 0.0442, 0.1863, 0.1286, 0.0201, 0.0057], device='cuda:5'), in_proj_covar=tensor([0.0231, 0.0204, 0.0221, 0.0141, 0.0212, 0.0169, 0.0159, 0.0099], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 16:45:36,485 INFO [train.py:904] (5/8) Epoch 2, batch 450, loss[loss=0.2735, simple_loss=0.3307, pruned_loss=0.1081, over 16570.00 frames. ], tot_loss[loss=0.2809, simple_loss=0.3393, pruned_loss=0.1112, over 2982470.38 frames. ], batch size: 75, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:45:59,588 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.527e+02 4.134e+02 5.028e+02 5.858e+02 1.204e+03, threshold=1.006e+03, percent-clipped=2.0 2023-04-27 16:46:44,108 INFO [train.py:904] (5/8) Epoch 2, batch 500, loss[loss=0.2914, simple_loss=0.3415, pruned_loss=0.1207, over 16118.00 frames. ], tot_loss[loss=0.2792, simple_loss=0.3381, pruned_loss=0.1102, over 3053490.40 frames. ], batch size: 164, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:47:45,080 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:47:52,732 INFO [train.py:904] (5/8) Epoch 2, batch 550, loss[loss=0.3377, simple_loss=0.3733, pruned_loss=0.1511, over 16369.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.3372, pruned_loss=0.1091, over 3105484.68 frames. ], batch size: 146, lr: 3.19e-02, grad_scale: 8.0 2023-04-27 16:48:10,576 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1050, 3.0682, 3.4074, 2.8483, 3.5891, 3.3446, 3.6950, 1.9156], device='cuda:5'), covar=tensor([0.0870, 0.0200, 0.0098, 0.0364, 0.0060, 0.0125, 0.0060, 0.0780], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0052, 0.0056, 0.0094, 0.0051, 0.0054, 0.0058, 0.0098], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:5') 2023-04-27 16:48:17,046 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.652e+02 3.972e+02 4.860e+02 6.091e+02 1.021e+03, threshold=9.720e+02, percent-clipped=1.0 2023-04-27 16:48:17,446 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:48:18,727 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0972, 3.4422, 2.7691, 4.3815, 2.4743, 4.1957, 2.9117, 2.5392], device='cuda:5'), covar=tensor([0.0261, 0.0312, 0.0329, 0.0215, 0.1409, 0.0163, 0.0542, 0.1420], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0145, 0.0118, 0.0167, 0.0227, 0.0136, 0.0158, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:48:47,710 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6012, 4.5560, 4.2972, 1.5593, 2.9398, 2.6341, 3.5566, 4.6976], device='cuda:5'), covar=tensor([0.0333, 0.0338, 0.0337, 0.2354, 0.0974, 0.1153, 0.1010, 0.0411], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0091, 0.0134, 0.0157, 0.0148, 0.0141, 0.0145, 0.0089], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 16:48:51,499 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:48:52,785 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:02,282 INFO [train.py:904] (5/8) Epoch 2, batch 600, loss[loss=0.276, simple_loss=0.3252, pruned_loss=0.1134, over 16850.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3354, pruned_loss=0.1079, over 3163654.94 frames. ], batch size: 116, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:49:22,663 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:24,354 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:49:33,777 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:11,094 INFO [train.py:904] (5/8) Epoch 2, batch 650, loss[loss=0.2858, simple_loss=0.3362, pruned_loss=0.1177, over 16297.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3342, pruned_loss=0.1068, over 3197146.61 frames. ], batch size: 165, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:50:17,092 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:24,045 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:50:33,031 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.483e+02 4.152e+02 4.979e+02 6.118e+02 1.196e+03, threshold=9.959e+02, percent-clipped=5.0 2023-04-27 16:50:48,638 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7983, 3.8793, 2.9081, 3.4264, 2.7922, 2.0655, 3.8363, 4.5828], device='cuda:5'), covar=tensor([0.1716, 0.0503, 0.1034, 0.0429, 0.1704, 0.1586, 0.0385, 0.0107], device='cuda:5'), in_proj_covar=tensor([0.0231, 0.0208, 0.0218, 0.0144, 0.0221, 0.0168, 0.0162, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 16:50:57,026 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:51:03,742 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 16:51:19,294 INFO [train.py:904] (5/8) Epoch 2, batch 700, loss[loss=0.3285, simple_loss=0.3717, pruned_loss=0.1427, over 15323.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.334, pruned_loss=0.1063, over 3227973.31 frames. ], batch size: 190, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:52:06,310 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:52:26,662 INFO [train.py:904] (5/8) Epoch 2, batch 750, loss[loss=0.2835, simple_loss=0.3348, pruned_loss=0.1161, over 16910.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.3343, pruned_loss=0.1066, over 3245633.41 frames. ], batch size: 96, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:52:50,209 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 3.722e+02 4.836e+02 5.954e+02 8.013e+02, threshold=9.671e+02, percent-clipped=0.0 2023-04-27 16:53:28,049 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:53:33,387 INFO [train.py:904] (5/8) Epoch 2, batch 800, loss[loss=0.2722, simple_loss=0.3408, pruned_loss=0.1019, over 17238.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.334, pruned_loss=0.1064, over 3256691.87 frames. ], batch size: 52, lr: 3.16e-02, grad_scale: 8.0 2023-04-27 16:54:00,198 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7339, 4.4980, 4.5258, 4.6800, 3.8891, 4.5360, 4.5037, 4.2233], device='cuda:5'), covar=tensor([0.0348, 0.0196, 0.0229, 0.0159, 0.1072, 0.0238, 0.0242, 0.0288], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0087, 0.0162, 0.0130, 0.0199, 0.0126, 0.0110, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 16:54:21,149 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 16:54:43,056 INFO [train.py:904] (5/8) Epoch 2, batch 850, loss[loss=0.2868, simple_loss=0.3365, pruned_loss=0.1185, over 15596.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3328, pruned_loss=0.1053, over 3277450.72 frames. ], batch size: 190, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:54:45,122 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-27 16:55:06,015 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.379e+02 3.969e+02 4.868e+02 6.050e+02 1.097e+03, threshold=9.736e+02, percent-clipped=1.0 2023-04-27 16:55:26,217 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4883, 4.9695, 5.2079, 5.3888, 4.6887, 5.2157, 5.1818, 4.9270], device='cuda:5'), covar=tensor([0.0203, 0.0172, 0.0146, 0.0097, 0.0715, 0.0151, 0.0114, 0.0177], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0087, 0.0159, 0.0128, 0.0194, 0.0122, 0.0110, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 16:55:49,289 INFO [train.py:904] (5/8) Epoch 2, batch 900, loss[loss=0.2769, simple_loss=0.33, pruned_loss=0.1119, over 16893.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3308, pruned_loss=0.1031, over 3288092.15 frames. ], batch size: 116, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:56:10,041 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:37,564 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6545, 4.7710, 4.7959, 4.9057, 4.7535, 5.3093, 5.1630, 4.7594], device='cuda:5'), covar=tensor([0.0789, 0.1258, 0.1010, 0.1219, 0.2177, 0.0743, 0.0791, 0.1777], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0268, 0.0233, 0.0228, 0.0288, 0.0230, 0.0198, 0.0293], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 16:56:57,476 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:58,328 INFO [train.py:904] (5/8) Epoch 2, batch 950, loss[loss=0.2575, simple_loss=0.3345, pruned_loss=0.09024, over 16776.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3306, pruned_loss=0.1034, over 3289990.49 frames. ], batch size: 57, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:57:03,593 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2763, 3.7135, 3.6873, 1.4614, 3.8058, 3.7800, 3.3110, 3.2128], device='cuda:5'), covar=tensor([0.0677, 0.0105, 0.0209, 0.1637, 0.0092, 0.0094, 0.0232, 0.0218], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0076, 0.0076, 0.0150, 0.0066, 0.0068, 0.0088, 0.0101], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:57:04,045 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-04-27 16:57:10,357 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:57:14,732 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:57:19,042 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0499, 4.3717, 2.4430, 5.2984, 5.1792, 4.5582, 2.8879, 4.1865], device='cuda:5'), covar=tensor([0.1748, 0.0284, 0.1593, 0.0066, 0.0123, 0.0340, 0.0945, 0.0384], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0103, 0.0159, 0.0063, 0.0083, 0.0092, 0.0141, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:57:21,475 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.464e+02 3.927e+02 4.944e+02 5.832e+02 1.325e+03, threshold=9.889e+02, percent-clipped=4.0 2023-04-27 16:57:38,037 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:57:48,442 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-27 16:58:06,962 INFO [train.py:904] (5/8) Epoch 2, batch 1000, loss[loss=0.2669, simple_loss=0.3321, pruned_loss=0.1009, over 17131.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.3284, pruned_loss=0.1024, over 3299662.20 frames. ], batch size: 47, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:58:16,902 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:37,575 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:59:13,979 INFO [train.py:904] (5/8) Epoch 2, batch 1050, loss[loss=0.2447, simple_loss=0.3153, pruned_loss=0.08704, over 16702.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3283, pruned_loss=0.1019, over 3309977.30 frames. ], batch size: 57, lr: 3.13e-02, grad_scale: 8.0 2023-04-27 16:59:19,821 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1828, 5.5641, 5.1696, 5.5005, 4.8468, 4.9309, 4.9658, 5.6395], device='cuda:5'), covar=tensor([0.0420, 0.0654, 0.0713, 0.0253, 0.0580, 0.0358, 0.0463, 0.0523], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0288, 0.0253, 0.0168, 0.0197, 0.0163, 0.0231, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 16:59:26,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6915, 2.9303, 2.6554, 3.8726, 2.3325, 3.7048, 2.6473, 2.5742], device='cuda:5'), covar=tensor([0.0242, 0.0322, 0.0293, 0.0201, 0.1201, 0.0152, 0.0546, 0.1021], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0147, 0.0121, 0.0170, 0.0227, 0.0139, 0.0160, 0.0203], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 16:59:36,192 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.907e+02 3.907e+02 4.719e+02 5.702e+02 1.083e+03, threshold=9.439e+02, percent-clipped=2.0 2023-04-27 16:59:55,649 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8036, 3.5473, 3.3560, 1.7317, 2.7758, 2.0833, 3.3573, 3.5925], device='cuda:5'), covar=tensor([0.0327, 0.0461, 0.0360, 0.1959, 0.0843, 0.1130, 0.0674, 0.0358], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0094, 0.0132, 0.0153, 0.0144, 0.0135, 0.0141, 0.0091], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 17:00:00,088 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:00:08,468 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:00:21,899 INFO [train.py:904] (5/8) Epoch 2, batch 1100, loss[loss=0.2944, simple_loss=0.3631, pruned_loss=0.1129, over 17053.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3276, pruned_loss=0.102, over 3313359.88 frames. ], batch size: 53, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:00:52,294 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2288, 3.4737, 2.9006, 4.3669, 2.3119, 4.1754, 2.5068, 2.3843], device='cuda:5'), covar=tensor([0.0249, 0.0384, 0.0337, 0.0246, 0.1807, 0.0190, 0.0800, 0.1800], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0146, 0.0122, 0.0170, 0.0226, 0.0141, 0.0159, 0.0203], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 17:01:28,365 INFO [train.py:904] (5/8) Epoch 2, batch 1150, loss[loss=0.2512, simple_loss=0.3158, pruned_loss=0.09329, over 16489.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3271, pruned_loss=0.1012, over 3308547.74 frames. ], batch size: 68, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:01:52,664 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.647e+02 4.017e+02 4.974e+02 5.619e+02 1.017e+03, threshold=9.949e+02, percent-clipped=1.0 2023-04-27 17:02:39,329 INFO [train.py:904] (5/8) Epoch 2, batch 1200, loss[loss=0.2841, simple_loss=0.3594, pruned_loss=0.1044, over 17048.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3258, pruned_loss=0.1008, over 3315987.21 frames. ], batch size: 53, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:02:51,888 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 2023-04-27 17:03:46,869 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11400.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:47,498 INFO [train.py:904] (5/8) Epoch 2, batch 1250, loss[loss=0.2875, simple_loss=0.355, pruned_loss=0.11, over 17122.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3267, pruned_loss=0.1012, over 3323871.55 frames. ], batch size: 47, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:04:10,335 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.776e+02 4.393e+02 5.372e+02 6.605e+02 1.355e+03, threshold=1.074e+03, percent-clipped=5.0 2023-04-27 17:04:26,807 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:49,880 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:53,740 INFO [train.py:904] (5/8) Epoch 2, batch 1300, loss[loss=0.3062, simple_loss=0.3517, pruned_loss=0.1303, over 16913.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3265, pruned_loss=0.1017, over 3326621.01 frames. ], batch size: 116, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:04:57,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 17:05:30,776 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11478.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:01,552 INFO [train.py:904] (5/8) Epoch 2, batch 1350, loss[loss=0.2718, simple_loss=0.3467, pruned_loss=0.09843, over 17128.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.3261, pruned_loss=0.1011, over 3324222.88 frames. ], batch size: 48, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:06:24,756 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.664e+02 4.249e+02 5.319e+02 6.551e+02 1.548e+03, threshold=1.064e+03, percent-clipped=4.0 2023-04-27 17:06:41,826 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:45,134 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:57,542 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11541.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 17:07:09,721 INFO [train.py:904] (5/8) Epoch 2, batch 1400, loss[loss=0.2468, simple_loss=0.3264, pruned_loss=0.08362, over 17018.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3267, pruned_loss=0.1011, over 3317774.83 frames. ], batch size: 50, lr: 3.09e-02, grad_scale: 8.0 2023-04-27 17:07:16,978 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2258, 2.1908, 2.5431, 2.9265, 3.5759, 3.5290, 2.6993, 3.5825], device='cuda:5'), covar=tensor([0.0074, 0.0369, 0.0218, 0.0147, 0.0056, 0.0103, 0.0193, 0.0070], device='cuda:5'), in_proj_covar=tensor([0.0066, 0.0103, 0.0087, 0.0074, 0.0050, 0.0053, 0.0084, 0.0051], device='cuda:5'), out_proj_covar=tensor([1.1894e-04, 1.8453e-04, 1.6904e-04, 1.4131e-04, 9.0793e-05, 1.0332e-04, 1.4671e-04, 9.6905e-05], device='cuda:5') 2023-04-27 17:07:32,137 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-04-27 17:07:33,952 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8017, 2.7095, 1.5976, 2.7429, 2.1725, 2.7668, 1.9132, 2.4114], device='cuda:5'), covar=tensor([0.0105, 0.0194, 0.1342, 0.0078, 0.0618, 0.0299, 0.1113, 0.0508], device='cuda:5'), in_proj_covar=tensor([0.0080, 0.0094, 0.0167, 0.0078, 0.0152, 0.0116, 0.0173, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 17:08:03,122 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:08:09,086 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:08:19,425 INFO [train.py:904] (5/8) Epoch 2, batch 1450, loss[loss=0.2494, simple_loss=0.3313, pruned_loss=0.08378, over 17234.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.3253, pruned_loss=0.1004, over 3324166.28 frames. ], batch size: 52, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:08:43,793 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.475e+02 3.985e+02 4.876e+02 5.970e+02 8.943e+02, threshold=9.752e+02, percent-clipped=0.0 2023-04-27 17:08:52,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8018, 4.4542, 4.5804, 4.7016, 4.0429, 4.5419, 4.5495, 4.2148], device='cuda:5'), covar=tensor([0.0242, 0.0177, 0.0198, 0.0137, 0.0821, 0.0197, 0.0224, 0.0273], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0089, 0.0164, 0.0132, 0.0202, 0.0128, 0.0113, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 17:09:06,434 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:09:26,120 INFO [train.py:904] (5/8) Epoch 2, batch 1500, loss[loss=0.2185, simple_loss=0.2935, pruned_loss=0.07182, over 16741.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.325, pruned_loss=0.1002, over 3322631.74 frames. ], batch size: 39, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:09:52,648 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1859, 4.6627, 2.6230, 5.2743, 5.3430, 4.6558, 2.9194, 4.2510], device='cuda:5'), covar=tensor([0.1578, 0.0242, 0.1492, 0.0102, 0.0103, 0.0314, 0.0933, 0.0372], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0107, 0.0162, 0.0067, 0.0089, 0.0099, 0.0148, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 17:10:30,093 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:33,591 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:36,158 INFO [train.py:904] (5/8) Epoch 2, batch 1550, loss[loss=0.2313, simple_loss=0.2973, pruned_loss=0.08268, over 16976.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3267, pruned_loss=0.1013, over 3321665.04 frames. ], batch size: 41, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:10:58,965 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.956e+02 4.303e+02 5.262e+02 6.472e+02 2.309e+03, threshold=1.052e+03, percent-clipped=9.0 2023-04-27 17:11:44,741 INFO [train.py:904] (5/8) Epoch 2, batch 1600, loss[loss=0.2778, simple_loss=0.3287, pruned_loss=0.1134, over 16774.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3298, pruned_loss=0.1036, over 3319774.49 frames. ], batch size: 83, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:11:51,984 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 17:11:57,735 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:12:12,255 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1807, 4.2026, 2.0313, 4.2312, 2.6748, 4.1539, 2.1794, 3.2525], device='cuda:5'), covar=tensor([0.0043, 0.0099, 0.1434, 0.0048, 0.0812, 0.0237, 0.1342, 0.0513], device='cuda:5'), in_proj_covar=tensor([0.0080, 0.0095, 0.0167, 0.0077, 0.0155, 0.0116, 0.0172, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 17:12:53,457 INFO [train.py:904] (5/8) Epoch 2, batch 1650, loss[loss=0.252, simple_loss=0.322, pruned_loss=0.09097, over 17202.00 frames. ], tot_loss[loss=0.27, simple_loss=0.3317, pruned_loss=0.1041, over 3305184.85 frames. ], batch size: 44, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:13:15,766 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:17,809 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.782e+02 3.949e+02 4.848e+02 6.012e+02 1.253e+03, threshold=9.697e+02, percent-clipped=3.0 2023-04-27 17:13:29,675 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 17:13:35,579 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:45,094 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8737, 3.9583, 3.0063, 3.3111, 2.7450, 2.0203, 3.9773, 4.5733], device='cuda:5'), covar=tensor([0.1609, 0.0506, 0.1038, 0.0452, 0.1866, 0.1367, 0.0340, 0.0157], device='cuda:5'), in_proj_covar=tensor([0.0234, 0.0214, 0.0224, 0.0152, 0.0237, 0.0170, 0.0168, 0.0111], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 17:13:54,547 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7672, 5.1514, 5.0468, 5.1194, 5.0348, 5.6327, 5.4108, 5.0317], device='cuda:5'), covar=tensor([0.0815, 0.1024, 0.1096, 0.1153, 0.2169, 0.0713, 0.0715, 0.1503], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0265, 0.0238, 0.0231, 0.0295, 0.0239, 0.0211, 0.0299], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 17:13:56,976 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:02,897 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3801, 4.2765, 4.7399, 4.8363, 4.8956, 4.3601, 4.4966, 4.6553], device='cuda:5'), covar=tensor([0.0253, 0.0260, 0.0347, 0.0339, 0.0316, 0.0279, 0.0728, 0.0232], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0143, 0.0166, 0.0165, 0.0193, 0.0158, 0.0244, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 17:14:03,720 INFO [train.py:904] (5/8) Epoch 2, batch 1700, loss[loss=0.2712, simple_loss=0.3355, pruned_loss=0.1035, over 16513.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3354, pruned_loss=0.1062, over 3300855.87 frames. ], batch size: 75, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:14:23,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9821, 2.2006, 2.6513, 2.9874, 3.8214, 3.5213, 2.3047, 3.7295], device='cuda:5'), covar=tensor([0.0072, 0.0311, 0.0152, 0.0120, 0.0031, 0.0094, 0.0204, 0.0045], device='cuda:5'), in_proj_covar=tensor([0.0065, 0.0101, 0.0089, 0.0073, 0.0050, 0.0052, 0.0086, 0.0051], device='cuda:5'), out_proj_covar=tensor([1.1926e-04, 1.8195e-04, 1.7127e-04, 1.3846e-04, 8.8333e-05, 9.9272e-05, 1.5000e-04, 9.6228e-05], device='cuda:5') 2023-04-27 17:14:40,674 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:41,662 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:55,443 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:15:01,573 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 17:15:13,210 INFO [train.py:904] (5/8) Epoch 2, batch 1750, loss[loss=0.3163, simple_loss=0.3627, pruned_loss=0.1349, over 16325.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3357, pruned_loss=0.106, over 3309565.58 frames. ], batch size: 165, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:15:21,929 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:15:36,934 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.686e+02 4.328e+02 5.055e+02 5.950e+02 1.641e+03, threshold=1.011e+03, percent-clipped=5.0 2023-04-27 17:15:39,703 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:15:43,038 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 17:16:17,797 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:21,486 INFO [train.py:904] (5/8) Epoch 2, batch 1800, loss[loss=0.2367, simple_loss=0.3067, pruned_loss=0.08332, over 17201.00 frames. ], tot_loss[loss=0.2731, simple_loss=0.3357, pruned_loss=0.1052, over 3313287.16 frames. ], batch size: 46, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:17:03,376 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:04,327 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8362, 3.8349, 3.6929, 3.3253, 3.8049, 2.0208, 3.5804, 3.7448], device='cuda:5'), covar=tensor([0.0107, 0.0070, 0.0097, 0.0324, 0.0073, 0.1173, 0.0094, 0.0140], device='cuda:5'), in_proj_covar=tensor([0.0064, 0.0054, 0.0076, 0.0100, 0.0058, 0.0103, 0.0072, 0.0076], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 17:17:07,192 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:09,780 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-04-27 17:17:16,357 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:33,777 INFO [train.py:904] (5/8) Epoch 2, batch 1850, loss[loss=0.2874, simple_loss=0.3654, pruned_loss=0.1047, over 16688.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.3371, pruned_loss=0.1051, over 3322287.31 frames. ], batch size: 57, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:17:45,629 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:57,888 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.566e+02 4.018e+02 4.936e+02 6.102e+02 1.131e+03, threshold=9.871e+02, percent-clipped=1.0 2023-04-27 17:18:04,025 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 17:18:08,275 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-27 17:18:13,603 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8432, 4.4036, 4.3154, 1.6531, 4.5429, 4.5369, 3.5318, 3.7742], device='cuda:5'), covar=tensor([0.0646, 0.0066, 0.0211, 0.1667, 0.0100, 0.0049, 0.0249, 0.0203], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0076, 0.0077, 0.0151, 0.0070, 0.0067, 0.0094, 0.0108], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 17:18:35,069 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:18:41,210 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7525, 6.0686, 5.6801, 5.9425, 5.2909, 4.8424, 5.5278, 6.1323], device='cuda:5'), covar=tensor([0.0378, 0.0458, 0.0779, 0.0282, 0.0490, 0.0421, 0.0414, 0.0408], device='cuda:5'), in_proj_covar=tensor([0.0217, 0.0298, 0.0267, 0.0181, 0.0207, 0.0174, 0.0237, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 17:18:43,223 INFO [train.py:904] (5/8) Epoch 2, batch 1900, loss[loss=0.2427, simple_loss=0.3191, pruned_loss=0.08319, over 17215.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.336, pruned_loss=0.1043, over 3319921.08 frames. ], batch size: 46, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:18:49,571 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:17,209 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:51,471 INFO [train.py:904] (5/8) Epoch 2, batch 1950, loss[loss=0.3003, simple_loss=0.3513, pruned_loss=0.1247, over 16840.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.335, pruned_loss=0.1027, over 3328098.28 frames. ], batch size: 102, lr: 3.03e-02, grad_scale: 8.0 2023-04-27 17:20:14,621 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.438e+02 4.259e+02 5.225e+02 6.323e+02 1.109e+03, threshold=1.045e+03, percent-clipped=2.0 2023-04-27 17:20:40,787 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:20:42,653 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9466, 4.0457, 3.3594, 3.5783, 3.0779, 2.2940, 4.4523, 5.0053], device='cuda:5'), covar=tensor([0.1611, 0.0562, 0.0908, 0.0417, 0.1844, 0.1204, 0.0232, 0.0084], device='cuda:5'), in_proj_covar=tensor([0.0242, 0.0218, 0.0227, 0.0156, 0.0243, 0.0174, 0.0169, 0.0112], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 17:20:46,700 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9688, 4.1656, 3.5788, 3.5432, 3.1097, 2.1559, 4.6838, 5.2807], device='cuda:5'), covar=tensor([0.1669, 0.0553, 0.0859, 0.0465, 0.1969, 0.1293, 0.0196, 0.0058], device='cuda:5'), in_proj_covar=tensor([0.0241, 0.0218, 0.0227, 0.0155, 0.0243, 0.0174, 0.0169, 0.0111], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-27 17:20:59,925 INFO [train.py:904] (5/8) Epoch 2, batch 2000, loss[loss=0.302, simple_loss=0.338, pruned_loss=0.133, over 16852.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3347, pruned_loss=0.1026, over 3326370.81 frames. ], batch size: 102, lr: 3.02e-02, grad_scale: 8.0 2023-04-27 17:21:28,889 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:21:51,888 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:09,011 INFO [train.py:904] (5/8) Epoch 2, batch 2050, loss[loss=0.2486, simple_loss=0.3303, pruned_loss=0.08351, over 16691.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3353, pruned_loss=0.1038, over 3324944.85 frames. ], batch size: 57, lr: 3.02e-02, grad_scale: 16.0 2023-04-27 17:22:11,067 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:30,399 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8300, 1.9527, 1.9541, 1.8806, 2.9292, 2.5470, 3.5627, 3.0588], device='cuda:5'), covar=tensor([0.0029, 0.0158, 0.0130, 0.0172, 0.0068, 0.0127, 0.0045, 0.0097], device='cuda:5'), in_proj_covar=tensor([0.0044, 0.0087, 0.0078, 0.0085, 0.0075, 0.0086, 0.0049, 0.0060], device='cuda:5'), out_proj_covar=tensor([7.1836e-05, 1.3469e-04, 1.2122e-04, 1.3427e-04, 1.2119e-04, 1.4025e-04, 7.7636e-05, 1.0352e-04], device='cuda:5') 2023-04-27 17:22:32,914 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.569e+02 4.228e+02 5.159e+02 6.169e+02 1.128e+03, threshold=1.032e+03, percent-clipped=1.0 2023-04-27 17:22:59,000 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:20,413 INFO [train.py:904] (5/8) Epoch 2, batch 2100, loss[loss=0.2196, simple_loss=0.2958, pruned_loss=0.07173, over 17208.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3358, pruned_loss=0.1039, over 3319039.81 frames. ], batch size: 44, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:23:53,753 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:54,167 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 17:24:15,567 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:28,735 INFO [train.py:904] (5/8) Epoch 2, batch 2150, loss[loss=0.2622, simple_loss=0.3258, pruned_loss=0.09929, over 17026.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3374, pruned_loss=0.1056, over 3318791.71 frames. ], batch size: 41, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:24:33,520 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:51,837 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.440e+02 4.527e+02 5.287e+02 6.148e+02 1.100e+03, threshold=1.057e+03, percent-clipped=4.0 2023-04-27 17:25:22,226 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:23,388 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:25,473 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1700, 4.1418, 4.4289, 4.4882, 4.6103, 4.3021, 3.9732, 4.4295], device='cuda:5'), covar=tensor([0.0400, 0.0350, 0.0494, 0.0523, 0.0488, 0.0322, 0.1231, 0.0378], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0155, 0.0182, 0.0169, 0.0202, 0.0160, 0.0256, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 17:25:33,138 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0892, 3.7676, 2.4993, 4.4688, 4.3494, 4.2872, 1.8941, 3.2944], device='cuda:5'), covar=tensor([0.1457, 0.0275, 0.1227, 0.0062, 0.0158, 0.0221, 0.1089, 0.0475], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0109, 0.0165, 0.0068, 0.0097, 0.0101, 0.0148, 0.0134], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-04-27 17:25:38,419 INFO [train.py:904] (5/8) Epoch 2, batch 2200, loss[loss=0.3162, simple_loss=0.3747, pruned_loss=0.1289, over 16675.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.339, pruned_loss=0.1071, over 3323139.90 frames. ], batch size: 57, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:25:44,430 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:26:48,780 INFO [train.py:904] (5/8) Epoch 2, batch 2250, loss[loss=0.2759, simple_loss=0.3382, pruned_loss=0.1068, over 16505.00 frames. ], tot_loss[loss=0.2756, simple_loss=0.3388, pruned_loss=0.1061, over 3321996.08 frames. ], batch size: 146, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:26:52,142 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:12,088 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.514e+02 4.351e+02 5.087e+02 6.619e+02 9.905e+02, threshold=1.017e+03, percent-clipped=0.0 2023-04-27 17:27:32,277 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:51,314 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:57,828 INFO [train.py:904] (5/8) Epoch 2, batch 2300, loss[loss=0.2905, simple_loss=0.338, pruned_loss=0.1215, over 16904.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3389, pruned_loss=0.1066, over 3318819.89 frames. ], batch size: 109, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:28:20,674 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-27 17:28:27,149 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:06,515 INFO [train.py:904] (5/8) Epoch 2, batch 2350, loss[loss=0.2794, simple_loss=0.3472, pruned_loss=0.1058, over 15926.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3394, pruned_loss=0.1063, over 3321166.86 frames. ], batch size: 35, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:29:07,844 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:15,030 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:31,567 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.397e+02 4.650e+02 5.724e+02 7.343e+02 1.924e+03, threshold=1.145e+03, percent-clipped=10.0 2023-04-27 17:29:33,899 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:43,370 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0141, 3.1037, 3.7040, 3.0225, 3.6895, 3.7668, 3.9305, 1.9383], device='cuda:5'), covar=tensor([0.0684, 0.0233, 0.0082, 0.0251, 0.0055, 0.0057, 0.0042, 0.0592], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0051, 0.0060, 0.0103, 0.0052, 0.0060, 0.0061, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 17:30:14,734 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:30:15,618 INFO [train.py:904] (5/8) Epoch 2, batch 2400, loss[loss=0.2717, simple_loss=0.3494, pruned_loss=0.09694, over 16651.00 frames. ], tot_loss[loss=0.2765, simple_loss=0.34, pruned_loss=0.1065, over 3326081.01 frames. ], batch size: 57, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:30:52,393 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:15,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7789, 4.8278, 5.2854, 5.2802, 5.4073, 4.9846, 4.9665, 5.0635], device='cuda:5'), covar=tensor([0.0253, 0.0243, 0.0309, 0.0409, 0.0349, 0.0220, 0.0664, 0.0228], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0152, 0.0173, 0.0167, 0.0201, 0.0157, 0.0251, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 17:31:26,500 INFO [train.py:904] (5/8) Epoch 2, batch 2450, loss[loss=0.2764, simple_loss=0.3535, pruned_loss=0.0997, over 17090.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3393, pruned_loss=0.1052, over 3335194.47 frames. ], batch size: 55, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:31:31,725 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:51,027 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.702e+02 4.143e+02 4.943e+02 6.147e+02 1.117e+03, threshold=9.887e+02, percent-clipped=0.0 2023-04-27 17:31:59,248 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:20,313 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:35,558 INFO [train.py:904] (5/8) Epoch 2, batch 2500, loss[loss=0.2878, simple_loss=0.3405, pruned_loss=0.1175, over 16890.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3399, pruned_loss=0.1059, over 3334237.01 frames. ], batch size: 96, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:32:36,842 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:27,072 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:43,147 INFO [train.py:904] (5/8) Epoch 2, batch 2550, loss[loss=0.2863, simple_loss=0.3358, pruned_loss=0.1184, over 16679.00 frames. ], tot_loss[loss=0.2752, simple_loss=0.3398, pruned_loss=0.1053, over 3336580.55 frames. ], batch size: 134, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:34:05,123 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-27 17:34:08,156 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.860e+02 4.439e+02 5.440e+02 6.842e+02 1.130e+03, threshold=1.088e+03, percent-clipped=2.0 2023-04-27 17:34:27,590 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:53,771 INFO [train.py:904] (5/8) Epoch 2, batch 2600, loss[loss=0.2752, simple_loss=0.3326, pruned_loss=0.1089, over 16809.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.3388, pruned_loss=0.1043, over 3336033.64 frames. ], batch size: 124, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:35:08,906 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:32,904 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:36:01,534 INFO [train.py:904] (5/8) Epoch 2, batch 2650, loss[loss=0.2839, simple_loss=0.3441, pruned_loss=0.1119, over 16887.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3389, pruned_loss=0.1038, over 3342747.98 frames. ], batch size: 116, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:36:03,346 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:36:27,047 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.569e+02 4.221e+02 4.762e+02 5.639e+02 1.002e+03, threshold=9.524e+02, percent-clipped=0.0 2023-04-27 17:36:34,012 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:37:09,367 INFO [train.py:904] (5/8) Epoch 2, batch 2700, loss[loss=0.3059, simple_loss=0.356, pruned_loss=0.1279, over 16189.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3392, pruned_loss=0.1037, over 3343157.79 frames. ], batch size: 165, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:38:04,055 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8499, 3.6950, 3.2959, 1.6396, 2.6559, 2.0874, 3.2825, 3.6762], device='cuda:5'), covar=tensor([0.0267, 0.0304, 0.0411, 0.1719, 0.0786, 0.1056, 0.0598, 0.0301], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0107, 0.0141, 0.0155, 0.0146, 0.0138, 0.0149, 0.0097], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 17:38:19,508 INFO [train.py:904] (5/8) Epoch 2, batch 2750, loss[loss=0.2504, simple_loss=0.3309, pruned_loss=0.08491, over 16087.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3385, pruned_loss=0.102, over 3340821.34 frames. ], batch size: 35, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:38:41,653 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.401e+02 4.125e+02 4.810e+02 5.955e+02 1.093e+03, threshold=9.620e+02, percent-clipped=2.0 2023-04-27 17:39:26,263 INFO [train.py:904] (5/8) Epoch 2, batch 2800, loss[loss=0.2807, simple_loss=0.3344, pruned_loss=0.1135, over 16832.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3386, pruned_loss=0.1029, over 3333726.65 frames. ], batch size: 96, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:39:51,783 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-04-27 17:40:05,128 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0546, 4.5648, 4.5071, 2.2407, 4.7483, 4.7526, 3.7047, 3.7106], device='cuda:5'), covar=tensor([0.0487, 0.0048, 0.0131, 0.1110, 0.0053, 0.0025, 0.0191, 0.0191], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0079, 0.0078, 0.0148, 0.0071, 0.0068, 0.0096, 0.0108], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 17:40:33,621 INFO [train.py:904] (5/8) Epoch 2, batch 2850, loss[loss=0.2693, simple_loss=0.3497, pruned_loss=0.09441, over 17257.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.338, pruned_loss=0.1026, over 3336073.94 frames. ], batch size: 52, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:40:55,835 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-27 17:40:57,335 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.754e+02 4.789e+02 5.762e+02 6.825e+02 1.913e+03, threshold=1.152e+03, percent-clipped=8.0 2023-04-27 17:41:16,885 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6263, 5.9496, 5.5348, 5.8411, 5.1169, 4.9234, 5.4793, 6.0117], device='cuda:5'), covar=tensor([0.0398, 0.0544, 0.0759, 0.0358, 0.0523, 0.0376, 0.0390, 0.0416], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0310, 0.0273, 0.0190, 0.0212, 0.0178, 0.0246, 0.0212], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 17:41:39,154 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-27 17:41:41,140 INFO [train.py:904] (5/8) Epoch 2, batch 2900, loss[loss=0.2492, simple_loss=0.331, pruned_loss=0.0837, over 17100.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3369, pruned_loss=0.1032, over 3332946.14 frames. ], batch size: 47, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:42:01,327 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7607, 1.7244, 2.1964, 2.6554, 2.8589, 2.5727, 1.7684, 2.6472], device='cuda:5'), covar=tensor([0.0057, 0.0257, 0.0168, 0.0115, 0.0046, 0.0126, 0.0236, 0.0053], device='cuda:5'), in_proj_covar=tensor([0.0070, 0.0107, 0.0091, 0.0081, 0.0055, 0.0055, 0.0092, 0.0054], device='cuda:5'), out_proj_covar=tensor([1.2931e-04, 1.9350e-04, 1.7396e-04, 1.5425e-04, 9.9680e-05, 1.0490e-04, 1.6167e-04, 1.0204e-04], device='cuda:5') 2023-04-27 17:42:18,768 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 17:42:42,659 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1531, 4.9650, 4.8627, 3.8729, 4.7650, 2.2985, 4.5520, 4.8532], device='cuda:5'), covar=tensor([0.0074, 0.0068, 0.0083, 0.0494, 0.0085, 0.1192, 0.0089, 0.0145], device='cuda:5'), in_proj_covar=tensor([0.0066, 0.0056, 0.0080, 0.0106, 0.0063, 0.0104, 0.0076, 0.0083], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 17:42:49,010 INFO [train.py:904] (5/8) Epoch 2, batch 2950, loss[loss=0.263, simple_loss=0.3396, pruned_loss=0.09319, over 17236.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3367, pruned_loss=0.1043, over 3324795.07 frames. ], batch size: 52, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:42:50,418 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:42:50,660 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1445, 3.2668, 2.6861, 4.4506, 2.5297, 4.3552, 2.7635, 2.7467], device='cuda:5'), covar=tensor([0.0204, 0.0303, 0.0278, 0.0138, 0.1044, 0.0124, 0.0508, 0.0991], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0158, 0.0133, 0.0186, 0.0234, 0.0146, 0.0164, 0.0215], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 17:43:05,790 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3037, 2.2943, 1.7186, 1.8981, 2.6584, 2.6119, 3.2307, 2.6616], device='cuda:5'), covar=tensor([0.0068, 0.0133, 0.0144, 0.0147, 0.0065, 0.0109, 0.0050, 0.0087], device='cuda:5'), in_proj_covar=tensor([0.0045, 0.0087, 0.0080, 0.0083, 0.0075, 0.0086, 0.0053, 0.0063], device='cuda:5'), out_proj_covar=tensor([7.5631e-05, 1.3562e-04, 1.2381e-04, 1.3118e-04, 1.2356e-04, 1.4060e-04, 8.4822e-05, 1.0725e-04], device='cuda:5') 2023-04-27 17:43:13,805 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.070e+02 4.352e+02 4.979e+02 5.894e+02 1.092e+03, threshold=9.958e+02, percent-clipped=0.0 2023-04-27 17:43:14,149 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:43:53,714 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:54,563 INFO [train.py:904] (5/8) Epoch 2, batch 3000, loss[loss=0.2577, simple_loss=0.3365, pruned_loss=0.08945, over 17112.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.3362, pruned_loss=0.1035, over 3330642.47 frames. ], batch size: 49, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:43:54,564 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 17:44:03,915 INFO [train.py:938] (5/8) Epoch 2, validation: loss=0.1858, simple_loss=0.2917, pruned_loss=0.04, over 944034.00 frames. 2023-04-27 17:44:03,916 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-27 17:44:16,373 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 17:45:09,407 INFO [train.py:904] (5/8) Epoch 2, batch 3050, loss[loss=0.2258, simple_loss=0.3064, pruned_loss=0.07258, over 17216.00 frames. ], tot_loss[loss=0.2722, simple_loss=0.3365, pruned_loss=0.1039, over 3327500.33 frames. ], batch size: 45, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:45:33,140 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.792e+02 4.585e+02 5.657e+02 6.871e+02 1.163e+03, threshold=1.131e+03, percent-clipped=2.0 2023-04-27 17:45:45,658 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8206, 3.7397, 2.1591, 4.3250, 4.1534, 4.1434, 1.9089, 2.8095], device='cuda:5'), covar=tensor([0.1524, 0.0272, 0.1486, 0.0096, 0.0241, 0.0283, 0.1020, 0.0633], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0110, 0.0162, 0.0068, 0.0104, 0.0108, 0.0146, 0.0134], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 17:46:15,667 INFO [train.py:904] (5/8) Epoch 2, batch 3100, loss[loss=0.2388, simple_loss=0.3077, pruned_loss=0.08492, over 15974.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3356, pruned_loss=0.1033, over 3327849.52 frames. ], batch size: 35, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:47:22,160 INFO [train.py:904] (5/8) Epoch 2, batch 3150, loss[loss=0.2716, simple_loss=0.3474, pruned_loss=0.09786, over 16757.00 frames. ], tot_loss[loss=0.27, simple_loss=0.334, pruned_loss=0.103, over 3326675.65 frames. ], batch size: 57, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:47:44,590 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.682e+02 3.986e+02 4.862e+02 6.258e+02 1.253e+03, threshold=9.723e+02, percent-clipped=2.0 2023-04-27 17:48:28,282 INFO [train.py:904] (5/8) Epoch 2, batch 3200, loss[loss=0.2519, simple_loss=0.3143, pruned_loss=0.09478, over 16758.00 frames. ], tot_loss[loss=0.2693, simple_loss=0.3328, pruned_loss=0.1029, over 3316715.50 frames. ], batch size: 89, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:16,604 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 17:49:34,234 INFO [train.py:904] (5/8) Epoch 2, batch 3250, loss[loss=0.2869, simple_loss=0.3449, pruned_loss=0.1145, over 16648.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.3323, pruned_loss=0.1021, over 3327165.04 frames. ], batch size: 134, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:39,460 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 17:49:58,158 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.960e+02 4.105e+02 4.918e+02 6.860e+02 1.329e+03, threshold=9.835e+02, percent-clipped=3.0 2023-04-27 17:49:58,595 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:50:02,823 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5082, 4.3829, 4.3380, 3.7902, 4.3259, 2.0823, 4.1722, 4.4049], device='cuda:5'), covar=tensor([0.0067, 0.0063, 0.0063, 0.0285, 0.0057, 0.0997, 0.0068, 0.0094], device='cuda:5'), in_proj_covar=tensor([0.0067, 0.0057, 0.0081, 0.0107, 0.0064, 0.0105, 0.0078, 0.0084], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 17:50:42,668 INFO [train.py:904] (5/8) Epoch 2, batch 3300, loss[loss=0.2425, simple_loss=0.3134, pruned_loss=0.08584, over 16820.00 frames. ], tot_loss[loss=0.27, simple_loss=0.3339, pruned_loss=0.103, over 3330604.00 frames. ], batch size: 42, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:51:03,261 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:51:16,198 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1045, 4.4121, 2.5946, 5.3314, 5.1758, 4.7004, 2.1427, 3.7745], device='cuda:5'), covar=tensor([0.1584, 0.0282, 0.1367, 0.0061, 0.0171, 0.0354, 0.1236, 0.0494], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0113, 0.0166, 0.0070, 0.0109, 0.0110, 0.0150, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 17:51:26,913 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7080, 1.5790, 2.2216, 2.4145, 2.7727, 2.6025, 1.7804, 2.5936], device='cuda:5'), covar=tensor([0.0050, 0.0274, 0.0141, 0.0130, 0.0038, 0.0086, 0.0236, 0.0045], device='cuda:5'), in_proj_covar=tensor([0.0072, 0.0108, 0.0091, 0.0081, 0.0057, 0.0057, 0.0092, 0.0054], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 17:51:48,098 INFO [train.py:904] (5/8) Epoch 2, batch 3350, loss[loss=0.3352, simple_loss=0.3759, pruned_loss=0.1473, over 16160.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3332, pruned_loss=0.1019, over 3334004.01 frames. ], batch size: 165, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:51:51,960 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 17:51:54,816 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2395, 5.4968, 5.1926, 5.4350, 4.8240, 4.4595, 5.0518, 5.6371], device='cuda:5'), covar=tensor([0.0426, 0.0548, 0.0782, 0.0260, 0.0482, 0.0484, 0.0416, 0.0444], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0311, 0.0273, 0.0193, 0.0208, 0.0183, 0.0244, 0.0214], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 17:52:07,178 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 17:52:13,283 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.954e+02 4.311e+02 5.188e+02 6.478e+02 1.182e+03, threshold=1.038e+03, percent-clipped=6.0 2023-04-27 17:52:24,452 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8991, 3.3316, 3.3797, 1.3600, 3.4142, 3.4815, 3.0393, 2.8970], device='cuda:5'), covar=tensor([0.0858, 0.0108, 0.0161, 0.1565, 0.0117, 0.0093, 0.0243, 0.0338], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0084, 0.0081, 0.0150, 0.0076, 0.0072, 0.0098, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 17:52:48,142 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 17:52:56,208 INFO [train.py:904] (5/8) Epoch 2, batch 3400, loss[loss=0.2534, simple_loss=0.3314, pruned_loss=0.08769, over 17263.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3331, pruned_loss=0.1006, over 3336048.33 frames. ], batch size: 52, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:54:05,267 INFO [train.py:904] (5/8) Epoch 2, batch 3450, loss[loss=0.3333, simple_loss=0.3727, pruned_loss=0.1469, over 16463.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3317, pruned_loss=0.09988, over 3331374.11 frames. ], batch size: 146, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:54:16,429 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4515, 4.2156, 4.3443, 4.7672, 4.8226, 4.3812, 4.6430, 4.6306], device='cuda:5'), covar=tensor([0.0450, 0.0553, 0.1212, 0.0399, 0.0399, 0.0582, 0.0535, 0.0451], device='cuda:5'), in_proj_covar=tensor([0.0248, 0.0290, 0.0423, 0.0309, 0.0240, 0.0232, 0.0227, 0.0242], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 17:54:29,798 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.037e+02 4.547e+02 5.342e+02 6.587e+02 1.202e+03, threshold=1.068e+03, percent-clipped=2.0 2023-04-27 17:55:11,514 INFO [train.py:904] (5/8) Epoch 2, batch 3500, loss[loss=0.2762, simple_loss=0.3483, pruned_loss=0.102, over 17066.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3302, pruned_loss=0.09902, over 3325341.87 frames. ], batch size: 55, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:56:21,461 INFO [train.py:904] (5/8) Epoch 2, batch 3550, loss[loss=0.2882, simple_loss=0.3427, pruned_loss=0.1169, over 16419.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3283, pruned_loss=0.09791, over 3327643.26 frames. ], batch size: 146, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:56:32,012 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 17:56:45,025 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.316e+02 4.227e+02 5.059e+02 5.919e+02 1.365e+03, threshold=1.012e+03, percent-clipped=3.0 2023-04-27 17:57:11,466 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:57:29,808 INFO [train.py:904] (5/8) Epoch 2, batch 3600, loss[loss=0.3, simple_loss=0.3457, pruned_loss=0.1271, over 16773.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3263, pruned_loss=0.09649, over 3332495.01 frames. ], batch size: 116, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:58:37,370 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:58:40,991 INFO [train.py:904] (5/8) Epoch 2, batch 3650, loss[loss=0.2523, simple_loss=0.305, pruned_loss=0.09983, over 16819.00 frames. ], tot_loss[loss=0.2606, simple_loss=0.3259, pruned_loss=0.09763, over 3320066.18 frames. ], batch size: 116, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 17:59:08,122 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.959e+02 4.047e+02 4.833e+02 5.657e+02 1.025e+03, threshold=9.667e+02, percent-clipped=2.0 2023-04-27 17:59:54,540 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-04-27 17:59:54,879 INFO [train.py:904] (5/8) Epoch 2, batch 3700, loss[loss=0.2957, simple_loss=0.3347, pruned_loss=0.1283, over 11290.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3245, pruned_loss=0.099, over 3292890.41 frames. ], batch size: 248, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 18:01:07,718 INFO [train.py:904] (5/8) Epoch 2, batch 3750, loss[loss=0.2746, simple_loss=0.3241, pruned_loss=0.1126, over 16798.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.325, pruned_loss=0.1012, over 3295037.21 frames. ], batch size: 102, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:01:22,580 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4777, 3.8747, 3.8705, 1.8619, 3.9603, 3.9551, 3.3677, 3.1307], device='cuda:5'), covar=tensor([0.0697, 0.0099, 0.0142, 0.1411, 0.0090, 0.0047, 0.0257, 0.0305], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0079, 0.0077, 0.0145, 0.0075, 0.0068, 0.0099, 0.0110], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:01:33,976 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.812e+02 4.240e+02 4.945e+02 6.119e+02 1.020e+03, threshold=9.889e+02, percent-clipped=4.0 2023-04-27 18:02:23,400 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3007, 4.1172, 4.1670, 4.2951, 3.5082, 4.2284, 4.1256, 3.9104], device='cuda:5'), covar=tensor([0.0299, 0.0182, 0.0247, 0.0160, 0.0886, 0.0206, 0.0365, 0.0269], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0090, 0.0166, 0.0127, 0.0195, 0.0135, 0.0113, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:02:24,617 INFO [train.py:904] (5/8) Epoch 2, batch 3800, loss[loss=0.302, simple_loss=0.354, pruned_loss=0.125, over 16689.00 frames. ], tot_loss[loss=0.2663, simple_loss=0.3267, pruned_loss=0.103, over 3280277.51 frames. ], batch size: 124, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:03:28,840 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-27 18:03:40,206 INFO [train.py:904] (5/8) Epoch 2, batch 3850, loss[loss=0.2635, simple_loss=0.3262, pruned_loss=0.1004, over 15524.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3257, pruned_loss=0.1028, over 3278681.65 frames. ], batch size: 190, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:04:07,045 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.317e+02 4.007e+02 4.656e+02 5.451e+02 1.276e+03, threshold=9.312e+02, percent-clipped=3.0 2023-04-27 18:04:08,799 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3503, 4.2584, 4.1975, 3.4693, 4.2594, 2.0139, 4.0005, 4.2636], device='cuda:5'), covar=tensor([0.0095, 0.0065, 0.0084, 0.0406, 0.0062, 0.1336, 0.0097, 0.0117], device='cuda:5'), in_proj_covar=tensor([0.0064, 0.0056, 0.0081, 0.0103, 0.0063, 0.0106, 0.0076, 0.0080], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:04:53,523 INFO [train.py:904] (5/8) Epoch 2, batch 3900, loss[loss=0.2693, simple_loss=0.3371, pruned_loss=0.1008, over 17101.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3245, pruned_loss=0.103, over 3283317.77 frames. ], batch size: 48, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:05:02,682 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1787, 3.8427, 2.7115, 4.7193, 4.6397, 4.4744, 2.2681, 3.2484], device='cuda:5'), covar=tensor([0.1390, 0.0299, 0.1153, 0.0049, 0.0147, 0.0172, 0.0981, 0.0528], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0116, 0.0169, 0.0069, 0.0107, 0.0109, 0.0154, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 18:05:25,157 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:05:55,293 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14093.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:06:05,401 INFO [train.py:904] (5/8) Epoch 2, batch 3950, loss[loss=0.2514, simple_loss=0.3093, pruned_loss=0.09675, over 16874.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3223, pruned_loss=0.1029, over 3294602.37 frames. ], batch size: 96, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:06:09,435 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14103.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:06:31,642 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.483e+02 3.936e+02 4.888e+02 5.915e+02 1.002e+03, threshold=9.776e+02, percent-clipped=1.0 2023-04-27 18:06:34,580 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:06:52,522 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14133.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:06:53,768 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9417, 3.2353, 2.6752, 4.3501, 2.3806, 4.1511, 2.7114, 2.5964], device='cuda:5'), covar=tensor([0.0242, 0.0315, 0.0305, 0.0182, 0.1204, 0.0139, 0.0531, 0.1077], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0168, 0.0141, 0.0198, 0.0248, 0.0157, 0.0175, 0.0229], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:07:17,643 INFO [train.py:904] (5/8) Epoch 2, batch 4000, loss[loss=0.2279, simple_loss=0.298, pruned_loss=0.07889, over 16833.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3226, pruned_loss=0.1038, over 3278578.24 frames. ], batch size: 96, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:07:37,809 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:08:02,226 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 18:08:15,155 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9243, 3.6905, 3.9688, 4.2786, 4.3021, 3.9671, 4.2396, 4.2346], device='cuda:5'), covar=tensor([0.0522, 0.0524, 0.1021, 0.0341, 0.0419, 0.0594, 0.0428, 0.0325], device='cuda:5'), in_proj_covar=tensor([0.0232, 0.0267, 0.0377, 0.0277, 0.0226, 0.0214, 0.0213, 0.0216], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:08:31,022 INFO [train.py:904] (5/8) Epoch 2, batch 4050, loss[loss=0.2017, simple_loss=0.2875, pruned_loss=0.05792, over 16718.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3205, pruned_loss=0.1002, over 3282990.96 frames. ], batch size: 89, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:08:56,317 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.111e+02 3.521e+02 4.461e+02 5.736e+02 1.012e+03, threshold=8.922e+02, percent-clipped=2.0 2023-04-27 18:09:06,925 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 18:09:43,239 INFO [train.py:904] (5/8) Epoch 2, batch 4100, loss[loss=0.3397, simple_loss=0.383, pruned_loss=0.1482, over 11809.00 frames. ], tot_loss[loss=0.2578, simple_loss=0.3202, pruned_loss=0.0977, over 3275806.36 frames. ], batch size: 247, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:10:11,934 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2351, 3.6809, 3.7269, 1.6092, 4.0608, 3.9735, 3.3468, 3.1243], device='cuda:5'), covar=tensor([0.0770, 0.0108, 0.0172, 0.1474, 0.0049, 0.0049, 0.0231, 0.0333], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0080, 0.0076, 0.0151, 0.0073, 0.0071, 0.0102, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:10:54,691 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4276, 4.8032, 4.7730, 4.8404, 4.7513, 5.2985, 5.0356, 4.8059], device='cuda:5'), covar=tensor([0.0687, 0.0948, 0.0798, 0.1151, 0.1619, 0.0635, 0.0620, 0.1467], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0255, 0.0234, 0.0226, 0.0284, 0.0244, 0.0200, 0.0302], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:10:59,433 INFO [train.py:904] (5/8) Epoch 2, batch 4150, loss[loss=0.3777, simple_loss=0.4082, pruned_loss=0.1736, over 11355.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3298, pruned_loss=0.1022, over 3258320.52 frames. ], batch size: 248, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:11:25,037 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 18:11:25,251 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.427e+02 3.836e+02 4.498e+02 5.328e+02 1.203e+03, threshold=8.995e+02, percent-clipped=3.0 2023-04-27 18:11:31,677 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1238, 4.6496, 4.8410, 4.9942, 4.3663, 4.8641, 4.7511, 4.4838], device='cuda:5'), covar=tensor([0.0180, 0.0161, 0.0150, 0.0086, 0.0606, 0.0157, 0.0134, 0.0172], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0082, 0.0151, 0.0116, 0.0175, 0.0122, 0.0104, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:12:14,217 INFO [train.py:904] (5/8) Epoch 2, batch 4200, loss[loss=0.2821, simple_loss=0.36, pruned_loss=0.1021, over 16493.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.339, pruned_loss=0.106, over 3216359.96 frames. ], batch size: 75, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:15,544 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:26,263 INFO [train.py:904] (5/8) Epoch 2, batch 4250, loss[loss=0.2397, simple_loss=0.3253, pruned_loss=0.07705, over 16876.00 frames. ], tot_loss[loss=0.2773, simple_loss=0.3423, pruned_loss=0.1061, over 3204002.09 frames. ], batch size: 102, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:44,214 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:52,837 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.326e+02 4.073e+02 4.856e+02 5.593e+02 1.302e+03, threshold=9.713e+02, percent-clipped=3.0 2023-04-27 18:14:05,670 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14428.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:24,092 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:38,556 INFO [train.py:904] (5/8) Epoch 2, batch 4300, loss[loss=0.275, simple_loss=0.3496, pruned_loss=0.1002, over 16576.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3415, pruned_loss=0.1034, over 3200936.61 frames. ], batch size: 57, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:14:49,278 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:10,481 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:12,751 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:15:29,944 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:49,389 INFO [train.py:904] (5/8) Epoch 2, batch 4350, loss[loss=0.2741, simple_loss=0.3462, pruned_loss=0.101, over 16761.00 frames. ], tot_loss[loss=0.2787, simple_loss=0.3462, pruned_loss=0.1056, over 3197205.60 frames. ], batch size: 83, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:16:13,772 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.262e+02 4.011e+02 4.650e+02 6.201e+02 1.293e+03, threshold=9.301e+02, percent-clipped=5.0 2023-04-27 18:16:40,337 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5144, 4.5639, 5.1231, 4.9948, 5.2049, 4.6713, 4.5375, 4.7323], device='cuda:5'), covar=tensor([0.0342, 0.0276, 0.0289, 0.0446, 0.0424, 0.0309, 0.0847, 0.0300], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0144, 0.0166, 0.0163, 0.0192, 0.0158, 0.0243, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 18:16:56,238 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14548.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:16:59,399 INFO [train.py:904] (5/8) Epoch 2, batch 4400, loss[loss=0.2896, simple_loss=0.3603, pruned_loss=0.1095, over 16852.00 frames. ], tot_loss[loss=0.2822, simple_loss=0.3493, pruned_loss=0.1076, over 3179298.19 frames. ], batch size: 42, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:17:47,073 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 18:18:09,499 INFO [train.py:904] (5/8) Epoch 2, batch 4450, loss[loss=0.2956, simple_loss=0.3752, pruned_loss=0.108, over 16894.00 frames. ], tot_loss[loss=0.2831, simple_loss=0.3516, pruned_loss=0.1073, over 3176803.70 frames. ], batch size: 116, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:18:34,423 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.338e+02 3.415e+02 4.231e+02 5.024e+02 9.561e+02, threshold=8.461e+02, percent-clipped=1.0 2023-04-27 18:19:17,781 INFO [train.py:904] (5/8) Epoch 2, batch 4500, loss[loss=0.2849, simple_loss=0.3524, pruned_loss=0.1087, over 16469.00 frames. ], tot_loss[loss=0.2814, simple_loss=0.3504, pruned_loss=0.1063, over 3180473.26 frames. ], batch size: 146, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:19:27,205 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7295, 1.6046, 2.4277, 3.3055, 3.5767, 3.6342, 1.9621, 3.8103], device='cuda:5'), covar=tensor([0.0025, 0.0259, 0.0123, 0.0066, 0.0018, 0.0035, 0.0166, 0.0022], device='cuda:5'), in_proj_covar=tensor([0.0070, 0.0108, 0.0090, 0.0080, 0.0055, 0.0057, 0.0091, 0.0052], device='cuda:5'), out_proj_covar=tensor([1.2625e-04, 1.9339e-04, 1.6863e-04, 1.5046e-04, 9.6497e-05, 1.0500e-04, 1.5787e-04, 9.5128e-05], device='cuda:5') 2023-04-27 18:20:09,451 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-04-27 18:20:28,470 INFO [train.py:904] (5/8) Epoch 2, batch 4550, loss[loss=0.2889, simple_loss=0.3642, pruned_loss=0.1068, over 16534.00 frames. ], tot_loss[loss=0.282, simple_loss=0.351, pruned_loss=0.1065, over 3182648.20 frames. ], batch size: 68, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:20:54,776 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.252e+02 3.646e+02 4.280e+02 5.239e+02 1.001e+03, threshold=8.560e+02, percent-clipped=1.0 2023-04-27 18:21:08,348 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:25,089 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:40,573 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2363, 4.8175, 4.9835, 5.2849, 4.0782, 5.2367, 5.0412, 4.8725], device='cuda:5'), covar=tensor([0.0245, 0.0180, 0.0221, 0.0111, 0.0977, 0.0139, 0.0102, 0.0157], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0084, 0.0148, 0.0114, 0.0178, 0.0121, 0.0108, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:21:43,186 INFO [train.py:904] (5/8) Epoch 2, batch 4600, loss[loss=0.3312, simple_loss=0.3747, pruned_loss=0.1438, over 11543.00 frames. ], tot_loss[loss=0.2809, simple_loss=0.3506, pruned_loss=0.1056, over 3190605.08 frames. ], batch size: 246, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:21:53,356 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:02,258 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8827, 1.3291, 2.0665, 2.8295, 2.8807, 2.8816, 1.7746, 2.8044], device='cuda:5'), covar=tensor([0.0031, 0.0277, 0.0117, 0.0057, 0.0023, 0.0046, 0.0183, 0.0044], device='cuda:5'), in_proj_covar=tensor([0.0068, 0.0106, 0.0087, 0.0077, 0.0053, 0.0055, 0.0089, 0.0052], device='cuda:5'), out_proj_covar=tensor([1.2248e-04, 1.9106e-04, 1.6362e-04, 1.4468e-04, 9.3511e-05, 1.0234e-04, 1.5485e-04, 9.5050e-05], device='cuda:5') 2023-04-27 18:22:10,207 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:15,143 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0749, 4.9226, 4.7088, 3.7096, 5.2880, 2.1095, 4.8856, 5.1613], device='cuda:5'), covar=tensor([0.0107, 0.0072, 0.0071, 0.0428, 0.0034, 0.1281, 0.0049, 0.0090], device='cuda:5'), in_proj_covar=tensor([0.0060, 0.0049, 0.0072, 0.0095, 0.0056, 0.0102, 0.0068, 0.0072], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:22:18,599 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:18,654 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:22:54,480 INFO [train.py:904] (5/8) Epoch 2, batch 4650, loss[loss=0.3739, simple_loss=0.4106, pruned_loss=0.1685, over 11528.00 frames. ], tot_loss[loss=0.2786, simple_loss=0.3484, pruned_loss=0.1044, over 3190695.84 frames. ], batch size: 246, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:22:55,641 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:03,205 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14807.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:07,195 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7696, 2.9260, 2.3103, 3.8178, 3.7005, 3.5690, 1.5322, 2.8385], device='cuda:5'), covar=tensor([0.1617, 0.0368, 0.1430, 0.0064, 0.0166, 0.0281, 0.1372, 0.0627], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0115, 0.0166, 0.0064, 0.0096, 0.0105, 0.0149, 0.0140], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 18:23:07,255 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1882, 2.5168, 2.1899, 3.4889, 2.0398, 3.3879, 2.3698, 2.1040], device='cuda:5'), covar=tensor([0.0287, 0.0419, 0.0347, 0.0209, 0.1367, 0.0162, 0.0583, 0.1181], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0168, 0.0142, 0.0198, 0.0256, 0.0155, 0.0175, 0.0237], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:23:20,767 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 3.221e+02 3.729e+02 4.286e+02 9.218e+02, threshold=7.459e+02, percent-clipped=2.0 2023-04-27 18:23:27,865 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:23:27,929 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3687, 4.4537, 4.7717, 4.9203, 4.9852, 4.3461, 4.3456, 4.5091], device='cuda:5'), covar=tensor([0.0328, 0.0243, 0.0361, 0.0420, 0.0375, 0.0298, 0.0742, 0.0277], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0134, 0.0156, 0.0152, 0.0180, 0.0148, 0.0225, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-04-27 18:23:54,906 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:24:06,756 INFO [train.py:904] (5/8) Epoch 2, batch 4700, loss[loss=0.3027, simple_loss=0.3491, pruned_loss=0.1282, over 11284.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3455, pruned_loss=0.1029, over 3200375.35 frames. ], batch size: 246, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:25:20,216 INFO [train.py:904] (5/8) Epoch 2, batch 4750, loss[loss=0.2552, simple_loss=0.3302, pruned_loss=0.09016, over 16382.00 frames. ], tot_loss[loss=0.271, simple_loss=0.341, pruned_loss=0.1005, over 3209111.66 frames. ], batch size: 146, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:25:27,085 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 18:25:45,829 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.318e+02 3.402e+02 4.045e+02 5.011e+02 1.217e+03, threshold=8.089e+02, percent-clipped=2.0 2023-04-27 18:26:30,655 INFO [train.py:904] (5/8) Epoch 2, batch 4800, loss[loss=0.2872, simple_loss=0.3586, pruned_loss=0.1079, over 16373.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3384, pruned_loss=0.09921, over 3190849.41 frames. ], batch size: 146, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:27:31,163 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9043, 2.7088, 2.6960, 1.7300, 2.8728, 2.8312, 2.5019, 2.4855], device='cuda:5'), covar=tensor([0.0748, 0.0130, 0.0189, 0.1133, 0.0098, 0.0089, 0.0307, 0.0321], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0080, 0.0082, 0.0147, 0.0071, 0.0070, 0.0099, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:27:43,268 INFO [train.py:904] (5/8) Epoch 2, batch 4850, loss[loss=0.2469, simple_loss=0.3312, pruned_loss=0.08133, over 16407.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3407, pruned_loss=0.1001, over 3158164.76 frames. ], batch size: 75, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:28:11,114 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.137e+02 3.566e+02 4.230e+02 5.018e+02 9.509e+02, threshold=8.461e+02, percent-clipped=3.0 2023-04-27 18:28:22,279 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8644, 3.7115, 3.5935, 4.1672, 4.1097, 3.8991, 4.0553, 4.1751], device='cuda:5'), covar=tensor([0.0382, 0.0449, 0.1168, 0.0396, 0.0521, 0.0593, 0.0444, 0.0275], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0259, 0.0357, 0.0266, 0.0211, 0.0198, 0.0200, 0.0205], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:28:28,189 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 18:28:54,419 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1862, 3.0685, 3.2086, 3.4629, 3.4260, 3.2110, 3.4204, 3.4030], device='cuda:5'), covar=tensor([0.0429, 0.0420, 0.0880, 0.0366, 0.0397, 0.1097, 0.0380, 0.0311], device='cuda:5'), in_proj_covar=tensor([0.0220, 0.0254, 0.0355, 0.0264, 0.0207, 0.0194, 0.0199, 0.0201], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:28:58,616 INFO [train.py:904] (5/8) Epoch 2, batch 4900, loss[loss=0.273, simple_loss=0.3497, pruned_loss=0.09813, over 16744.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.3398, pruned_loss=0.09836, over 3155764.77 frames. ], batch size: 124, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:29:24,624 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:04,915 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:07,522 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8211, 3.8906, 4.1971, 4.1734, 4.2329, 3.8747, 3.8816, 3.9654], device='cuda:5'), covar=tensor([0.0216, 0.0218, 0.0294, 0.0330, 0.0286, 0.0217, 0.0563, 0.0287], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0141, 0.0164, 0.0158, 0.0184, 0.0151, 0.0229, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-04-27 18:30:10,710 INFO [train.py:904] (5/8) Epoch 2, batch 4950, loss[loss=0.2563, simple_loss=0.335, pruned_loss=0.08882, over 16740.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3389, pruned_loss=0.09699, over 3174228.87 frames. ], batch size: 89, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:30:34,534 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:37,413 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.417e+02 3.788e+02 4.518e+02 5.433e+02 9.876e+02, threshold=9.036e+02, percent-clipped=3.0 2023-04-27 18:30:56,283 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6209, 2.7622, 2.0358, 4.0033, 3.9022, 3.8462, 1.5010, 2.6371], device='cuda:5'), covar=tensor([0.1527, 0.0474, 0.1484, 0.0052, 0.0134, 0.0264, 0.1282, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0120, 0.0169, 0.0067, 0.0100, 0.0109, 0.0150, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 18:31:12,943 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:31:22,568 INFO [train.py:904] (5/8) Epoch 2, batch 5000, loss[loss=0.27, simple_loss=0.3455, pruned_loss=0.09727, over 16849.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.3406, pruned_loss=0.09706, over 3199228.78 frames. ], batch size: 42, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:32:21,457 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:26,371 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:34,519 INFO [train.py:904] (5/8) Epoch 2, batch 5050, loss[loss=0.2333, simple_loss=0.3115, pruned_loss=0.07755, over 17135.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.341, pruned_loss=0.09721, over 3192770.89 frames. ], batch size: 48, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:32:38,577 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-27 18:33:00,586 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.434e+02 3.699e+02 4.750e+02 5.849e+02 1.293e+03, threshold=9.501e+02, percent-clipped=2.0 2023-04-27 18:33:45,529 INFO [train.py:904] (5/8) Epoch 2, batch 5100, loss[loss=0.235, simple_loss=0.3116, pruned_loss=0.07922, over 16831.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3394, pruned_loss=0.09642, over 3194688.94 frames. ], batch size: 42, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:53,363 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:34:30,687 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4194, 3.3202, 3.1997, 3.3171, 3.0498, 3.3066, 3.1568, 3.1390], device='cuda:5'), covar=tensor([0.0271, 0.0151, 0.0177, 0.0121, 0.0481, 0.0145, 0.0648, 0.0210], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0091, 0.0153, 0.0116, 0.0178, 0.0126, 0.0104, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:34:43,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5841, 1.8210, 1.8345, 1.7066, 2.4556, 2.3073, 2.8045, 2.6854], device='cuda:5'), covar=tensor([0.0020, 0.0160, 0.0137, 0.0183, 0.0067, 0.0131, 0.0026, 0.0053], device='cuda:5'), in_proj_covar=tensor([0.0043, 0.0094, 0.0093, 0.0097, 0.0083, 0.0097, 0.0052, 0.0066], device='cuda:5'), out_proj_covar=tensor([6.5903e-05, 1.4777e-04, 1.4296e-04, 1.5710e-04, 1.3640e-04, 1.5887e-04, 8.4512e-05, 1.1025e-04], device='cuda:5') 2023-04-27 18:34:56,482 INFO [train.py:904] (5/8) Epoch 2, batch 5150, loss[loss=0.2887, simple_loss=0.3668, pruned_loss=0.1052, over 15268.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.34, pruned_loss=0.09566, over 3187203.43 frames. ], batch size: 190, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:35:22,260 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.454e+02 3.560e+02 4.122e+02 5.032e+02 1.152e+03, threshold=8.244e+02, percent-clipped=1.0 2023-04-27 18:35:52,184 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 18:36:10,033 INFO [train.py:904] (5/8) Epoch 2, batch 5200, loss[loss=0.2941, simple_loss=0.3656, pruned_loss=0.1113, over 15453.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3388, pruned_loss=0.09548, over 3188356.55 frames. ], batch size: 190, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:36:53,442 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0590, 4.6173, 4.8095, 4.9570, 4.2735, 4.7691, 4.8342, 4.5869], device='cuda:5'), covar=tensor([0.0299, 0.0269, 0.0161, 0.0113, 0.0822, 0.0205, 0.0141, 0.0225], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0095, 0.0161, 0.0123, 0.0190, 0.0133, 0.0110, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:37:12,837 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0333, 2.0791, 1.8573, 1.8794, 2.6519, 2.5864, 3.1662, 2.9618], device='cuda:5'), covar=tensor([0.0015, 0.0155, 0.0172, 0.0193, 0.0066, 0.0128, 0.0026, 0.0049], device='cuda:5'), in_proj_covar=tensor([0.0042, 0.0096, 0.0096, 0.0099, 0.0085, 0.0099, 0.0054, 0.0068], device='cuda:5'), out_proj_covar=tensor([6.5120e-05, 1.5198e-04, 1.4775e-04, 1.6082e-04, 1.4133e-04, 1.6096e-04, 8.8231e-05, 1.1474e-04], device='cuda:5') 2023-04-27 18:37:16,300 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:37:23,464 INFO [train.py:904] (5/8) Epoch 2, batch 5250, loss[loss=0.206, simple_loss=0.2944, pruned_loss=0.05879, over 16842.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3358, pruned_loss=0.0949, over 3206672.51 frames. ], batch size: 102, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:37:48,868 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.244e+02 3.713e+02 4.444e+02 5.638e+02 1.103e+03, threshold=8.887e+02, percent-clipped=2.0 2023-04-27 18:37:56,614 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2480, 3.1503, 1.4379, 3.1967, 2.2592, 3.2132, 1.7819, 2.3721], device='cuda:5'), covar=tensor([0.0080, 0.0157, 0.1761, 0.0063, 0.0870, 0.0286, 0.1474, 0.0726], device='cuda:5'), in_proj_covar=tensor([0.0075, 0.0099, 0.0167, 0.0079, 0.0156, 0.0124, 0.0174, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 18:38:23,202 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:38:32,784 INFO [train.py:904] (5/8) Epoch 2, batch 5300, loss[loss=0.2831, simple_loss=0.3434, pruned_loss=0.1114, over 12107.00 frames. ], tot_loss[loss=0.2589, simple_loss=0.3316, pruned_loss=0.09308, over 3201814.50 frames. ], batch size: 247, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:39:43,336 INFO [train.py:904] (5/8) Epoch 2, batch 5350, loss[loss=0.2692, simple_loss=0.3444, pruned_loss=0.09699, over 16717.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.329, pruned_loss=0.09143, over 3214334.84 frames. ], batch size: 124, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:40:09,985 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.455e+02 3.880e+02 4.934e+02 5.714e+02 1.287e+03, threshold=9.868e+02, percent-clipped=1.0 2023-04-27 18:40:14,765 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 18:40:56,837 INFO [train.py:904] (5/8) Epoch 2, batch 5400, loss[loss=0.266, simple_loss=0.3385, pruned_loss=0.09676, over 16520.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3334, pruned_loss=0.09362, over 3205203.12 frames. ], batch size: 62, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:40:57,221 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 18:42:10,440 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 18:42:14,670 INFO [train.py:904] (5/8) Epoch 2, batch 5450, loss[loss=0.3689, simple_loss=0.4185, pruned_loss=0.1596, over 15315.00 frames. ], tot_loss[loss=0.2644, simple_loss=0.3367, pruned_loss=0.096, over 3179143.19 frames. ], batch size: 191, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:42:43,049 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 4.170e+02 4.882e+02 6.060e+02 1.199e+03, threshold=9.765e+02, percent-clipped=3.0 2023-04-27 18:43:34,037 INFO [train.py:904] (5/8) Epoch 2, batch 5500, loss[loss=0.3079, simple_loss=0.3773, pruned_loss=0.1193, over 16668.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3483, pruned_loss=0.1055, over 3153112.93 frames. ], batch size: 134, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:43:51,390 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6273, 3.4163, 3.0793, 1.6605, 2.6055, 2.0899, 3.1762, 3.5732], device='cuda:5'), covar=tensor([0.0290, 0.0367, 0.0389, 0.1672, 0.0705, 0.1015, 0.0692, 0.0302], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0108, 0.0147, 0.0152, 0.0144, 0.0136, 0.0148, 0.0096], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 18:44:48,104 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-27 18:44:50,703 INFO [train.py:904] (5/8) Epoch 2, batch 5550, loss[loss=0.3425, simple_loss=0.3943, pruned_loss=0.1453, over 16222.00 frames. ], tot_loss[loss=0.2934, simple_loss=0.3583, pruned_loss=0.1143, over 3127999.28 frames. ], batch size: 165, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:45:13,085 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3480, 1.4564, 1.9236, 2.3314, 2.3055, 2.6179, 1.6085, 2.4695], device='cuda:5'), covar=tensor([0.0054, 0.0196, 0.0104, 0.0075, 0.0036, 0.0037, 0.0147, 0.0037], device='cuda:5'), in_proj_covar=tensor([0.0070, 0.0109, 0.0093, 0.0081, 0.0058, 0.0057, 0.0091, 0.0054], device='cuda:5'), out_proj_covar=tensor([1.2455e-04, 1.9361e-04, 1.7134e-04, 1.5217e-04, 1.0023e-04, 1.0307e-04, 1.5742e-04, 9.4812e-05], device='cuda:5') 2023-04-27 18:45:16,849 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5612, 3.6357, 1.4291, 3.5590, 2.5138, 3.7176, 1.9259, 2.5283], device='cuda:5'), covar=tensor([0.0049, 0.0125, 0.1726, 0.0051, 0.0649, 0.0219, 0.1359, 0.0718], device='cuda:5'), in_proj_covar=tensor([0.0072, 0.0099, 0.0165, 0.0075, 0.0153, 0.0124, 0.0170, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 18:45:19,406 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.692e+02 5.827e+02 6.984e+02 8.601e+02 1.757e+03, threshold=1.397e+03, percent-clipped=15.0 2023-04-27 18:45:21,636 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8337, 4.0659, 3.7763, 3.9436, 3.5511, 3.6586, 3.7825, 3.9869], device='cuda:5'), covar=tensor([0.0416, 0.0566, 0.0818, 0.0321, 0.0471, 0.0702, 0.0438, 0.0648], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0285, 0.0266, 0.0183, 0.0192, 0.0178, 0.0232, 0.0200], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:46:11,052 INFO [train.py:904] (5/8) Epoch 2, batch 5600, loss[loss=0.3939, simple_loss=0.4164, pruned_loss=0.1857, over 11107.00 frames. ], tot_loss[loss=0.3041, simple_loss=0.3658, pruned_loss=0.1212, over 3111061.90 frames. ], batch size: 247, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:46:39,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6694, 3.5590, 3.6516, 3.5498, 3.6495, 4.0612, 3.9365, 3.5114], device='cuda:5'), covar=tensor([0.1391, 0.1636, 0.1216, 0.2259, 0.2276, 0.1067, 0.1006, 0.2362], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0265, 0.0242, 0.0231, 0.0299, 0.0253, 0.0203, 0.0308], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:47:34,300 INFO [train.py:904] (5/8) Epoch 2, batch 5650, loss[loss=0.3745, simple_loss=0.4115, pruned_loss=0.1687, over 15227.00 frames. ], tot_loss[loss=0.3145, simple_loss=0.373, pruned_loss=0.128, over 3084874.73 frames. ], batch size: 190, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:01,995 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.886e+02 5.674e+02 6.824e+02 8.519e+02 2.118e+03, threshold=1.365e+03, percent-clipped=2.0 2023-04-27 18:48:07,038 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7957, 4.3264, 4.3218, 2.0164, 4.6834, 4.6948, 3.3119, 3.5442], device='cuda:5'), covar=tensor([0.0867, 0.0103, 0.0155, 0.1573, 0.0048, 0.0033, 0.0345, 0.0332], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0082, 0.0077, 0.0154, 0.0074, 0.0072, 0.0106, 0.0115], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:48:53,336 INFO [train.py:904] (5/8) Epoch 2, batch 5700, loss[loss=0.3666, simple_loss=0.3971, pruned_loss=0.1681, over 11497.00 frames. ], tot_loss[loss=0.3167, simple_loss=0.3743, pruned_loss=0.1295, over 3081540.97 frames. ], batch size: 248, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:53,762 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:49:28,498 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5462, 4.3950, 1.4124, 4.4554, 2.9391, 4.5420, 2.1540, 3.0521], device='cuda:5'), covar=tensor([0.0029, 0.0088, 0.2087, 0.0030, 0.0834, 0.0187, 0.1427, 0.0604], device='cuda:5'), in_proj_covar=tensor([0.0072, 0.0101, 0.0169, 0.0075, 0.0160, 0.0127, 0.0174, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 18:49:44,691 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15883.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:09,228 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:12,116 INFO [train.py:904] (5/8) Epoch 2, batch 5750, loss[loss=0.3458, simple_loss=0.3766, pruned_loss=0.1575, over 11342.00 frames. ], tot_loss[loss=0.3204, simple_loss=0.3773, pruned_loss=0.1317, over 3054086.00 frames. ], batch size: 246, lr: 2.69e-02, grad_scale: 8.0 2023-04-27 18:50:42,016 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.285e+02 4.919e+02 6.542e+02 8.153e+02 1.932e+03, threshold=1.308e+03, percent-clipped=2.0 2023-04-27 18:51:22,903 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:51:33,508 INFO [train.py:904] (5/8) Epoch 2, batch 5800, loss[loss=0.2651, simple_loss=0.3451, pruned_loss=0.0926, over 16617.00 frames. ], tot_loss[loss=0.3176, simple_loss=0.3763, pruned_loss=0.1295, over 3068696.31 frames. ], batch size: 76, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:52:56,021 INFO [train.py:904] (5/8) Epoch 2, batch 5850, loss[loss=0.3136, simple_loss=0.3776, pruned_loss=0.1248, over 16277.00 frames. ], tot_loss[loss=0.3125, simple_loss=0.373, pruned_loss=0.126, over 3084624.80 frames. ], batch size: 165, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:52:57,397 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5119, 5.1180, 5.2180, 5.3100, 4.7286, 5.2368, 5.1378, 4.9679], device='cuda:5'), covar=tensor([0.0212, 0.0161, 0.0120, 0.0093, 0.0666, 0.0140, 0.0118, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0088, 0.0146, 0.0113, 0.0177, 0.0121, 0.0103, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 18:52:57,454 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16001.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:53:25,313 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.722e+02 4.866e+02 5.967e+02 7.152e+02 1.262e+03, threshold=1.193e+03, percent-clipped=0.0 2023-04-27 18:53:55,950 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-04-27 18:54:18,258 INFO [train.py:904] (5/8) Epoch 2, batch 5900, loss[loss=0.2838, simple_loss=0.3558, pruned_loss=0.1059, over 16815.00 frames. ], tot_loss[loss=0.3118, simple_loss=0.3723, pruned_loss=0.1257, over 3071853.80 frames. ], batch size: 124, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:54:40,591 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:17,952 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:29,759 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3625, 3.9169, 3.9236, 1.5969, 4.0516, 4.1018, 3.2791, 3.0552], device='cuda:5'), covar=tensor([0.0840, 0.0095, 0.0137, 0.1620, 0.0058, 0.0048, 0.0258, 0.0368], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0080, 0.0074, 0.0149, 0.0073, 0.0072, 0.0103, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:55:42,187 INFO [train.py:904] (5/8) Epoch 2, batch 5950, loss[loss=0.3121, simple_loss=0.3784, pruned_loss=0.1229, over 16855.00 frames. ], tot_loss[loss=0.311, simple_loss=0.3737, pruned_loss=0.1242, over 3092816.36 frames. ], batch size: 116, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:56:13,606 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.088e+02 4.747e+02 6.406e+02 8.221e+02 1.979e+03, threshold=1.281e+03, percent-clipped=5.0 2023-04-27 18:56:21,908 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3612, 3.2882, 2.6540, 2.6784, 2.4523, 1.9350, 3.4939, 3.8579], device='cuda:5'), covar=tensor([0.1674, 0.0519, 0.0945, 0.0580, 0.1594, 0.1154, 0.0319, 0.0148], device='cuda:5'), in_proj_covar=tensor([0.0249, 0.0219, 0.0236, 0.0174, 0.0264, 0.0178, 0.0189, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:56:56,108 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:02,049 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:04,569 INFO [train.py:904] (5/8) Epoch 2, batch 6000, loss[loss=0.3199, simple_loss=0.3753, pruned_loss=0.1322, over 15383.00 frames. ], tot_loss[loss=0.3106, simple_loss=0.3731, pruned_loss=0.124, over 3105415.52 frames. ], batch size: 190, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:57:04,570 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 18:57:15,925 INFO [train.py:938] (5/8) Epoch 2, validation: loss=0.2302, simple_loss=0.3372, pruned_loss=0.06166, over 944034.00 frames. 2023-04-27 18:57:15,927 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-27 18:58:20,001 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3941, 3.3055, 2.6331, 2.5507, 2.5539, 1.9487, 3.2434, 3.7322], device='cuda:5'), covar=tensor([0.1674, 0.0603, 0.1052, 0.0661, 0.1489, 0.1214, 0.0435, 0.0185], device='cuda:5'), in_proj_covar=tensor([0.0248, 0.0220, 0.0237, 0.0175, 0.0269, 0.0177, 0.0192, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 18:58:34,801 INFO [train.py:904] (5/8) Epoch 2, batch 6050, loss[loss=0.2753, simple_loss=0.353, pruned_loss=0.09882, over 17021.00 frames. ], tot_loss[loss=0.3082, simple_loss=0.3712, pruned_loss=0.1226, over 3106600.69 frames. ], batch size: 50, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 18:58:41,036 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 18:58:48,691 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16210.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:04,139 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.028e+02 4.577e+02 5.718e+02 7.421e+02 1.323e+03, threshold=1.144e+03, percent-clipped=1.0 2023-04-27 18:59:33,669 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:52,056 INFO [train.py:904] (5/8) Epoch 2, batch 6100, loss[loss=0.2791, simple_loss=0.348, pruned_loss=0.1051, over 16564.00 frames. ], tot_loss[loss=0.3055, simple_loss=0.3702, pruned_loss=0.1204, over 3113526.22 frames. ], batch size: 62, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:00:22,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7441, 4.5753, 4.1192, 5.0500, 5.1072, 4.4399, 5.0641, 4.8867], device='cuda:5'), covar=tensor([0.0449, 0.0470, 0.1719, 0.0529, 0.0479, 0.0383, 0.0469, 0.0632], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0269, 0.0371, 0.0275, 0.0210, 0.0193, 0.0206, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:01:17,152 INFO [train.py:904] (5/8) Epoch 2, batch 6150, loss[loss=0.2982, simple_loss=0.363, pruned_loss=0.1167, over 16300.00 frames. ], tot_loss[loss=0.3033, simple_loss=0.3677, pruned_loss=0.1194, over 3107848.60 frames. ], batch size: 165, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:01:24,177 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 19:01:45,850 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.137e+02 4.737e+02 5.827e+02 7.174e+02 1.739e+03, threshold=1.165e+03, percent-clipped=5.0 2023-04-27 19:02:19,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9863, 2.5239, 2.0723, 3.1431, 2.2845, 3.0511, 2.4266, 2.1598], device='cuda:5'), covar=tensor([0.0247, 0.0300, 0.0256, 0.0179, 0.0902, 0.0171, 0.0442, 0.0832], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0175, 0.0149, 0.0204, 0.0259, 0.0162, 0.0180, 0.0238], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:02:34,542 INFO [train.py:904] (5/8) Epoch 2, batch 6200, loss[loss=0.3232, simple_loss=0.3736, pruned_loss=0.1364, over 15400.00 frames. ], tot_loss[loss=0.3031, simple_loss=0.3664, pruned_loss=0.1199, over 3075236.12 frames. ], batch size: 191, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:02:45,552 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:02,414 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 19:03:06,859 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:21,327 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6301, 4.8773, 4.5866, 4.6879, 4.3089, 4.2098, 4.4298, 4.9327], device='cuda:5'), covar=tensor([0.0434, 0.0534, 0.0760, 0.0354, 0.0397, 0.0507, 0.0434, 0.0499], device='cuda:5'), in_proj_covar=tensor([0.0218, 0.0304, 0.0278, 0.0192, 0.0209, 0.0185, 0.0246, 0.0214], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:03:25,442 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9956, 3.7588, 3.4447, 1.6523, 2.7008, 2.0573, 3.4559, 3.8763], device='cuda:5'), covar=tensor([0.0261, 0.0333, 0.0409, 0.1777, 0.0770, 0.1196, 0.0556, 0.0334], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0106, 0.0148, 0.0151, 0.0144, 0.0134, 0.0147, 0.0099], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 19:03:51,377 INFO [train.py:904] (5/8) Epoch 2, batch 6250, loss[loss=0.3021, simple_loss=0.3756, pruned_loss=0.1143, over 16918.00 frames. ], tot_loss[loss=0.3025, simple_loss=0.366, pruned_loss=0.1195, over 3094382.44 frames. ], batch size: 109, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:04:18,709 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.383e+02 4.905e+02 5.910e+02 7.685e+02 1.899e+03, threshold=1.182e+03, percent-clipped=6.0 2023-04-27 19:04:37,377 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:04:51,288 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:04:53,509 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 19:05:05,100 INFO [train.py:904] (5/8) Epoch 2, batch 6300, loss[loss=0.3553, simple_loss=0.4141, pruned_loss=0.1483, over 16862.00 frames. ], tot_loss[loss=0.3017, simple_loss=0.3657, pruned_loss=0.1188, over 3103143.69 frames. ], batch size: 96, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:06:22,088 INFO [train.py:904] (5/8) Epoch 2, batch 6350, loss[loss=0.2794, simple_loss=0.3448, pruned_loss=0.107, over 16828.00 frames. ], tot_loss[loss=0.3049, simple_loss=0.3672, pruned_loss=0.1212, over 3091416.16 frames. ], batch size: 116, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:06:29,793 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:06:52,068 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.677e+02 5.217e+02 6.547e+02 8.022e+02 1.954e+03, threshold=1.309e+03, percent-clipped=7.0 2023-04-27 19:07:13,650 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-04-27 19:07:21,362 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16539.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:07:39,368 INFO [train.py:904] (5/8) Epoch 2, batch 6400, loss[loss=0.396, simple_loss=0.4247, pruned_loss=0.1836, over 10937.00 frames. ], tot_loss[loss=0.3081, simple_loss=0.3689, pruned_loss=0.1237, over 3073835.91 frames. ], batch size: 248, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:08:07,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4290, 2.5131, 2.2536, 3.8881, 1.8019, 3.4293, 2.1215, 2.2443], device='cuda:5'), covar=tensor([0.0376, 0.0525, 0.0380, 0.0211, 0.1694, 0.0249, 0.0813, 0.1138], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0178, 0.0147, 0.0209, 0.0258, 0.0166, 0.0180, 0.0238], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:08:34,558 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:08:34,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9833, 3.4474, 2.5286, 4.6754, 4.6650, 4.3800, 1.7255, 3.4367], device='cuda:5'), covar=tensor([0.1319, 0.0400, 0.1163, 0.0073, 0.0164, 0.0207, 0.1300, 0.0518], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0120, 0.0164, 0.0067, 0.0106, 0.0114, 0.0150, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 19:08:55,796 INFO [train.py:904] (5/8) Epoch 2, batch 6450, loss[loss=0.2725, simple_loss=0.3431, pruned_loss=0.1009, over 16701.00 frames. ], tot_loss[loss=0.3048, simple_loss=0.367, pruned_loss=0.1213, over 3075165.37 frames. ], batch size: 76, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:09:25,708 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.052e+02 4.859e+02 5.740e+02 6.998e+02 1.216e+03, threshold=1.148e+03, percent-clipped=0.0 2023-04-27 19:09:50,216 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9395, 2.4444, 2.1015, 3.1946, 2.2700, 3.0327, 2.3200, 1.9868], device='cuda:5'), covar=tensor([0.0270, 0.0373, 0.0287, 0.0209, 0.0950, 0.0174, 0.0542, 0.1013], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0181, 0.0151, 0.0211, 0.0262, 0.0167, 0.0185, 0.0243], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:10:13,793 INFO [train.py:904] (5/8) Epoch 2, batch 6500, loss[loss=0.2604, simple_loss=0.3412, pruned_loss=0.08984, over 16833.00 frames. ], tot_loss[loss=0.3004, simple_loss=0.363, pruned_loss=0.1189, over 3077085.63 frames. ], batch size: 102, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:10:24,352 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:15,075 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 19:11:32,129 INFO [train.py:904] (5/8) Epoch 2, batch 6550, loss[loss=0.3147, simple_loss=0.3893, pruned_loss=0.12, over 16873.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3655, pruned_loss=0.1196, over 3085300.47 frames. ], batch size: 116, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:11:37,741 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:57,428 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 19:11:59,598 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.469e+02 4.692e+02 5.928e+02 7.455e+02 1.589e+03, threshold=1.186e+03, percent-clipped=4.0 2023-04-27 19:12:11,260 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:33,006 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:43,759 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6350, 3.5590, 3.6530, 3.9372, 3.9605, 3.6261, 3.8926, 3.8845], device='cuda:5'), covar=tensor([0.0455, 0.0436, 0.0943, 0.0340, 0.0360, 0.0822, 0.0389, 0.0308], device='cuda:5'), in_proj_covar=tensor([0.0236, 0.0267, 0.0372, 0.0279, 0.0214, 0.0194, 0.0210, 0.0219], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:12:47,626 INFO [train.py:904] (5/8) Epoch 2, batch 6600, loss[loss=0.2715, simple_loss=0.3454, pruned_loss=0.09883, over 16844.00 frames. ], tot_loss[loss=0.3047, simple_loss=0.3681, pruned_loss=0.1207, over 3066986.12 frames. ], batch size: 96, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:13:47,418 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:06,691 INFO [train.py:904] (5/8) Epoch 2, batch 6650, loss[loss=0.2794, simple_loss=0.3525, pruned_loss=0.1031, over 16373.00 frames. ], tot_loss[loss=0.3062, simple_loss=0.3686, pruned_loss=0.1219, over 3076178.53 frames. ], batch size: 146, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:14:13,175 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:35,747 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.480e+02 5.041e+02 6.003e+02 7.722e+02 1.259e+03, threshold=1.201e+03, percent-clipped=2.0 2023-04-27 19:15:05,806 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 19:15:23,520 INFO [train.py:904] (5/8) Epoch 2, batch 6700, loss[loss=0.2871, simple_loss=0.3554, pruned_loss=0.1095, over 16969.00 frames. ], tot_loss[loss=0.3049, simple_loss=0.3669, pruned_loss=0.1215, over 3072941.51 frames. ], batch size: 109, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:15:26,815 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:15:50,682 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:16:38,387 INFO [train.py:904] (5/8) Epoch 2, batch 6750, loss[loss=0.3651, simple_loss=0.397, pruned_loss=0.1666, over 11386.00 frames. ], tot_loss[loss=0.3045, simple_loss=0.3661, pruned_loss=0.1215, over 3081492.78 frames. ], batch size: 248, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:16:47,153 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:17:07,559 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.622e+02 5.251e+02 6.489e+02 8.459e+02 1.507e+03, threshold=1.298e+03, percent-clipped=4.0 2023-04-27 19:17:22,319 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:17:53,320 INFO [train.py:904] (5/8) Epoch 2, batch 6800, loss[loss=0.3092, simple_loss=0.3786, pruned_loss=0.1199, over 16684.00 frames. ], tot_loss[loss=0.3033, simple_loss=0.3657, pruned_loss=0.1205, over 3093247.50 frames. ], batch size: 89, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:18:18,750 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:19:10,220 INFO [train.py:904] (5/8) Epoch 2, batch 6850, loss[loss=0.3753, simple_loss=0.4095, pruned_loss=0.1705, over 11623.00 frames. ], tot_loss[loss=0.3025, simple_loss=0.366, pruned_loss=0.1196, over 3112177.63 frames. ], batch size: 248, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:19:38,488 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.045e+02 4.695e+02 5.738e+02 7.185e+02 1.728e+03, threshold=1.148e+03, percent-clipped=4.0 2023-04-27 19:19:49,853 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:20:24,240 INFO [train.py:904] (5/8) Epoch 2, batch 6900, loss[loss=0.3774, simple_loss=0.4055, pruned_loss=0.1746, over 11482.00 frames. ], tot_loss[loss=0.3044, simple_loss=0.3687, pruned_loss=0.1201, over 3091554.53 frames. ], batch size: 246, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:21:01,221 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:40,736 INFO [train.py:904] (5/8) Epoch 2, batch 6950, loss[loss=0.3634, simple_loss=0.3942, pruned_loss=0.1663, over 11418.00 frames. ], tot_loss[loss=0.3064, simple_loss=0.3702, pruned_loss=0.1213, over 3120889.44 frames. ], batch size: 248, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:22:09,951 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.451e+02 5.662e+02 6.855e+02 8.747e+02 1.724e+03, threshold=1.371e+03, percent-clipped=6.0 2023-04-27 19:22:54,972 INFO [train.py:904] (5/8) Epoch 2, batch 7000, loss[loss=0.2921, simple_loss=0.3734, pruned_loss=0.1054, over 16536.00 frames. ], tot_loss[loss=0.3059, simple_loss=0.3704, pruned_loss=0.1208, over 3113958.04 frames. ], batch size: 62, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:23:06,371 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 19:23:29,287 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8226, 2.6801, 1.7030, 2.8556, 2.2017, 2.7702, 1.8820, 2.3445], device='cuda:5'), covar=tensor([0.0071, 0.0268, 0.1062, 0.0054, 0.0571, 0.0371, 0.1020, 0.0460], device='cuda:5'), in_proj_covar=tensor([0.0072, 0.0105, 0.0168, 0.0072, 0.0158, 0.0134, 0.0176, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 19:23:53,768 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:24:10,872 INFO [train.py:904] (5/8) Epoch 2, batch 7050, loss[loss=0.3581, simple_loss=0.4156, pruned_loss=0.1503, over 16263.00 frames. ], tot_loss[loss=0.308, simple_loss=0.3722, pruned_loss=0.1219, over 3110509.87 frames. ], batch size: 165, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:24:37,730 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.062e+02 5.191e+02 6.345e+02 7.855e+02 1.482e+03, threshold=1.269e+03, percent-clipped=3.0 2023-04-27 19:24:44,007 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:25:22,998 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:25:23,674 INFO [train.py:904] (5/8) Epoch 2, batch 7100, loss[loss=0.3409, simple_loss=0.3679, pruned_loss=0.1569, over 11267.00 frames. ], tot_loss[loss=0.3077, simple_loss=0.3707, pruned_loss=0.1223, over 3082860.72 frames. ], batch size: 248, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:25:40,696 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:25:59,145 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 19:26:38,673 INFO [train.py:904] (5/8) Epoch 2, batch 7150, loss[loss=0.3297, simple_loss=0.3955, pruned_loss=0.132, over 16367.00 frames. ], tot_loss[loss=0.3068, simple_loss=0.3689, pruned_loss=0.1224, over 3061244.33 frames. ], batch size: 165, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:26:47,157 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 19:27:07,393 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.740e+02 5.248e+02 6.301e+02 7.700e+02 1.780e+03, threshold=1.260e+03, percent-clipped=1.0 2023-04-27 19:27:52,990 INFO [train.py:904] (5/8) Epoch 2, batch 7200, loss[loss=0.2401, simple_loss=0.3138, pruned_loss=0.08322, over 16456.00 frames. ], tot_loss[loss=0.3013, simple_loss=0.3651, pruned_loss=0.1187, over 3076480.33 frames. ], batch size: 68, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:28:10,422 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:28:18,229 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4083, 4.1874, 3.8350, 1.7937, 2.8474, 2.3565, 3.7588, 4.1073], device='cuda:5'), covar=tensor([0.0247, 0.0371, 0.0405, 0.1740, 0.0828, 0.1035, 0.0576, 0.0396], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0107, 0.0151, 0.0155, 0.0147, 0.0138, 0.0150, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 19:28:45,020 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:29:13,599 INFO [train.py:904] (5/8) Epoch 2, batch 7250, loss[loss=0.2965, simple_loss=0.3527, pruned_loss=0.1201, over 11400.00 frames. ], tot_loss[loss=0.2982, simple_loss=0.3624, pruned_loss=0.117, over 3067572.25 frames. ], batch size: 247, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:29:42,540 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.618e+02 4.777e+02 5.726e+02 7.027e+02 1.873e+03, threshold=1.145e+03, percent-clipped=4.0 2023-04-27 19:29:47,301 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:30:19,121 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:30:29,437 INFO [train.py:904] (5/8) Epoch 2, batch 7300, loss[loss=0.2736, simple_loss=0.3502, pruned_loss=0.09854, over 16665.00 frames. ], tot_loss[loss=0.2974, simple_loss=0.3617, pruned_loss=0.1165, over 3067542.31 frames. ], batch size: 76, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:31:04,460 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5291, 3.3943, 2.9570, 1.7672, 2.6192, 1.9872, 3.0420, 3.4147], device='cuda:5'), covar=tensor([0.0326, 0.0386, 0.0462, 0.1629, 0.0698, 0.0953, 0.0653, 0.0367], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0109, 0.0153, 0.0154, 0.0146, 0.0137, 0.0154, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 19:31:46,484 INFO [train.py:904] (5/8) Epoch 2, batch 7350, loss[loss=0.2911, simple_loss=0.3632, pruned_loss=0.1095, over 16678.00 frames. ], tot_loss[loss=0.2966, simple_loss=0.3608, pruned_loss=0.1162, over 3049564.80 frames. ], batch size: 89, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:32:16,423 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.647e+02 4.932e+02 5.634e+02 7.018e+02 1.459e+03, threshold=1.127e+03, percent-clipped=2.0 2023-04-27 19:32:22,419 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:32:57,430 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:33:06,160 INFO [train.py:904] (5/8) Epoch 2, batch 7400, loss[loss=0.2599, simple_loss=0.3331, pruned_loss=0.09335, over 16481.00 frames. ], tot_loss[loss=0.2985, simple_loss=0.3624, pruned_loss=0.1173, over 3053362.76 frames. ], batch size: 68, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:33:25,280 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:33:41,414 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:34:08,113 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8562, 3.0096, 3.1585, 3.2110, 3.2049, 3.0566, 2.7588, 3.2319], device='cuda:5'), covar=tensor([0.0506, 0.0595, 0.0625, 0.0606, 0.0664, 0.0530, 0.1165, 0.0449], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0148, 0.0167, 0.0160, 0.0192, 0.0165, 0.0251, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 19:34:17,444 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6167, 3.3141, 3.1947, 1.2833, 3.4232, 3.4510, 2.7640, 2.5940], device='cuda:5'), covar=tensor([0.1298, 0.0123, 0.0217, 0.1803, 0.0083, 0.0071, 0.0388, 0.0496], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0085, 0.0082, 0.0149, 0.0075, 0.0069, 0.0106, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:34:27,197 INFO [train.py:904] (5/8) Epoch 2, batch 7450, loss[loss=0.28, simple_loss=0.357, pruned_loss=0.1015, over 16940.00 frames. ], tot_loss[loss=0.3021, simple_loss=0.3648, pruned_loss=0.1197, over 3049044.13 frames. ], batch size: 109, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:34:32,901 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-27 19:34:43,811 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17610.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:59,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.824e+02 5.406e+02 6.493e+02 7.703e+02 1.630e+03, threshold=1.299e+03, percent-clipped=5.0 2023-04-27 19:35:18,739 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 19:35:49,122 INFO [train.py:904] (5/8) Epoch 2, batch 7500, loss[loss=0.276, simple_loss=0.3485, pruned_loss=0.1018, over 16862.00 frames. ], tot_loss[loss=0.3029, simple_loss=0.3662, pruned_loss=0.1199, over 3050440.94 frames. ], batch size: 90, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:35:53,385 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3773, 4.0157, 4.0338, 1.8035, 4.2253, 4.2431, 3.2609, 3.4013], device='cuda:5'), covar=tensor([0.0726, 0.0077, 0.0114, 0.1428, 0.0047, 0.0035, 0.0239, 0.0268], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0083, 0.0081, 0.0150, 0.0074, 0.0070, 0.0104, 0.0118], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:36:26,988 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 19:37:05,024 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:05,679 INFO [train.py:904] (5/8) Epoch 2, batch 7550, loss[loss=0.2687, simple_loss=0.354, pruned_loss=0.09171, over 16844.00 frames. ], tot_loss[loss=0.3017, simple_loss=0.3647, pruned_loss=0.1193, over 3055762.53 frames. ], batch size: 102, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:37:32,661 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:34,649 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.677e+02 5.137e+02 6.469e+02 8.298e+02 1.927e+03, threshold=1.294e+03, percent-clipped=3.0 2023-04-27 19:37:51,100 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:38:05,277 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:38:08,148 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 19:38:10,922 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1642, 3.6074, 3.6992, 2.6637, 3.5317, 3.5836, 3.7757, 1.9059], device='cuda:5'), covar=tensor([0.0430, 0.0031, 0.0040, 0.0260, 0.0053, 0.0062, 0.0026, 0.0431], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0055, 0.0060, 0.0113, 0.0055, 0.0060, 0.0060, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 19:38:22,227 INFO [train.py:904] (5/8) Epoch 2, batch 7600, loss[loss=0.2712, simple_loss=0.3346, pruned_loss=0.1039, over 16630.00 frames. ], tot_loss[loss=0.3009, simple_loss=0.3635, pruned_loss=0.1192, over 3061156.08 frames. ], batch size: 62, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:38:29,740 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2494, 3.1590, 3.2697, 3.4686, 3.4301, 3.2046, 3.4310, 3.4409], device='cuda:5'), covar=tensor([0.0455, 0.0451, 0.0750, 0.0310, 0.0376, 0.1047, 0.0373, 0.0327], device='cuda:5'), in_proj_covar=tensor([0.0238, 0.0280, 0.0375, 0.0279, 0.0219, 0.0199, 0.0219, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:38:37,867 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:39:25,145 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:39:28,272 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4049, 3.3836, 3.2488, 3.3587, 3.0052, 3.3165, 3.1761, 3.1061], device='cuda:5'), covar=tensor([0.0286, 0.0139, 0.0185, 0.0132, 0.0574, 0.0172, 0.0609, 0.0227], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0094, 0.0146, 0.0118, 0.0174, 0.0126, 0.0106, 0.0137], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:39:39,708 INFO [train.py:904] (5/8) Epoch 2, batch 7650, loss[loss=0.3207, simple_loss=0.3872, pruned_loss=0.1271, over 15305.00 frames. ], tot_loss[loss=0.304, simple_loss=0.3653, pruned_loss=0.1214, over 3051225.16 frames. ], batch size: 190, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:40:10,731 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.673e+02 5.378e+02 6.384e+02 7.645e+02 1.933e+03, threshold=1.277e+03, percent-clipped=1.0 2023-04-27 19:40:50,407 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:40:58,875 INFO [train.py:904] (5/8) Epoch 2, batch 7700, loss[loss=0.268, simple_loss=0.3432, pruned_loss=0.09646, over 16823.00 frames. ], tot_loss[loss=0.3045, simple_loss=0.3654, pruned_loss=0.1218, over 3061786.55 frames. ], batch size: 102, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:41:59,515 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0708, 3.3242, 3.2808, 1.3733, 3.5521, 3.4972, 2.8397, 2.6398], device='cuda:5'), covar=tensor([0.0878, 0.0116, 0.0199, 0.1630, 0.0074, 0.0053, 0.0322, 0.0438], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0083, 0.0082, 0.0147, 0.0073, 0.0070, 0.0106, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:42:04,332 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:16,830 INFO [train.py:904] (5/8) Epoch 2, batch 7750, loss[loss=0.3573, simple_loss=0.3917, pruned_loss=0.1615, over 11246.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3662, pruned_loss=0.1225, over 3043921.22 frames. ], batch size: 247, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:42:46,604 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.308e+02 4.989e+02 6.319e+02 7.239e+02 1.280e+03, threshold=1.264e+03, percent-clipped=2.0 2023-04-27 19:43:12,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5835, 5.0105, 4.9653, 5.0732, 5.0789, 5.4777, 5.1308, 4.9316], device='cuda:5'), covar=tensor([0.0636, 0.0988, 0.0902, 0.1331, 0.1591, 0.0663, 0.0676, 0.1664], device='cuda:5'), in_proj_covar=tensor([0.0203, 0.0274, 0.0262, 0.0248, 0.0312, 0.0279, 0.0215, 0.0330], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 19:43:22,569 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 19:43:26,469 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8883, 3.8078, 3.7909, 3.2101, 3.7253, 1.7818, 3.5727, 3.7254], device='cuda:5'), covar=tensor([0.0064, 0.0051, 0.0062, 0.0258, 0.0059, 0.1230, 0.0068, 0.0087], device='cuda:5'), in_proj_covar=tensor([0.0061, 0.0051, 0.0077, 0.0098, 0.0059, 0.0109, 0.0070, 0.0074], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:43:31,915 INFO [train.py:904] (5/8) Epoch 2, batch 7800, loss[loss=0.2527, simple_loss=0.333, pruned_loss=0.08614, over 16471.00 frames. ], tot_loss[loss=0.3079, simple_loss=0.368, pruned_loss=0.1239, over 3042465.88 frames. ], batch size: 68, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:43:47,090 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:44:52,442 INFO [train.py:904] (5/8) Epoch 2, batch 7850, loss[loss=0.2517, simple_loss=0.3293, pruned_loss=0.08703, over 16512.00 frames. ], tot_loss[loss=0.3066, simple_loss=0.3681, pruned_loss=0.1225, over 3056952.27 frames. ], batch size: 75, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:45:18,600 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:45:21,253 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.805e+02 5.114e+02 6.121e+02 7.560e+02 1.271e+03, threshold=1.224e+03, percent-clipped=1.0 2023-04-27 19:45:22,905 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:45:48,910 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:46:05,630 INFO [train.py:904] (5/8) Epoch 2, batch 7900, loss[loss=0.3011, simple_loss=0.3741, pruned_loss=0.114, over 16375.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3671, pruned_loss=0.1216, over 3061918.44 frames. ], batch size: 146, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:46:14,085 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:46:29,415 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:03,311 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:05,058 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:47:25,407 INFO [train.py:904] (5/8) Epoch 2, batch 7950, loss[loss=0.2643, simple_loss=0.3312, pruned_loss=0.09868, over 17199.00 frames. ], tot_loss[loss=0.3038, simple_loss=0.3661, pruned_loss=0.1208, over 3074932.17 frames. ], batch size: 46, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:47:56,299 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.877e+02 5.212e+02 6.244e+02 7.914e+02 2.379e+03, threshold=1.249e+03, percent-clipped=3.0 2023-04-27 19:48:42,386 INFO [train.py:904] (5/8) Epoch 2, batch 8000, loss[loss=0.3365, simple_loss=0.3873, pruned_loss=0.1428, over 16728.00 frames. ], tot_loss[loss=0.3044, simple_loss=0.3657, pruned_loss=0.1215, over 3056207.66 frames. ], batch size: 134, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:48:45,288 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-04-27 19:49:16,868 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:18,142 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:56,243 INFO [train.py:904] (5/8) Epoch 2, batch 8050, loss[loss=0.294, simple_loss=0.3594, pruned_loss=0.1143, over 16669.00 frames. ], tot_loss[loss=0.3026, simple_loss=0.365, pruned_loss=0.1201, over 3079676.72 frames. ], batch size: 134, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:50:24,925 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.258e+02 4.729e+02 5.978e+02 7.201e+02 1.227e+03, threshold=1.196e+03, percent-clipped=0.0 2023-04-27 19:50:30,989 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 19:50:47,427 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:48,750 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:51:11,144 INFO [train.py:904] (5/8) Epoch 2, batch 8100, loss[loss=0.3162, simple_loss=0.3596, pruned_loss=0.1364, over 11338.00 frames. ], tot_loss[loss=0.3021, simple_loss=0.3647, pruned_loss=0.1197, over 3085174.74 frames. ], batch size: 246, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:51:33,837 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8979, 5.5940, 5.6423, 5.5907, 5.5587, 6.1282, 5.7691, 5.5683], device='cuda:5'), covar=tensor([0.0488, 0.0903, 0.0864, 0.1111, 0.1700, 0.0621, 0.0535, 0.1295], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0272, 0.0259, 0.0249, 0.0312, 0.0278, 0.0212, 0.0330], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 19:52:05,736 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 19:52:29,028 INFO [train.py:904] (5/8) Epoch 2, batch 8150, loss[loss=0.3064, simple_loss=0.3708, pruned_loss=0.121, over 15397.00 frames. ], tot_loss[loss=0.2985, simple_loss=0.3617, pruned_loss=0.1177, over 3102250.07 frames. ], batch size: 190, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:52:53,008 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:53:00,708 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.225e+02 5.223e+02 6.693e+02 9.034e+02 1.543e+03, threshold=1.339e+03, percent-clipped=7.0 2023-04-27 19:53:48,021 INFO [train.py:904] (5/8) Epoch 2, batch 8200, loss[loss=0.342, simple_loss=0.394, pruned_loss=0.145, over 15314.00 frames. ], tot_loss[loss=0.2959, simple_loss=0.3582, pruned_loss=0.1168, over 3093903.89 frames. ], batch size: 190, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:53:57,215 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:46,794 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18386.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:10,282 INFO [train.py:904] (5/8) Epoch 2, batch 8250, loss[loss=0.2461, simple_loss=0.3343, pruned_loss=0.07896, over 16839.00 frames. ], tot_loss[loss=0.2941, simple_loss=0.358, pruned_loss=0.1151, over 3075849.25 frames. ], batch size: 83, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:55:15,303 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:44,539 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-27 19:55:44,643 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.019e+02 4.548e+02 5.368e+02 6.842e+02 2.128e+03, threshold=1.074e+03, percent-clipped=3.0 2023-04-27 19:55:48,122 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4593, 2.5199, 2.2973, 3.8054, 1.9278, 3.6254, 2.4765, 2.1359], device='cuda:5'), covar=tensor([0.0303, 0.0564, 0.0408, 0.0167, 0.1665, 0.0191, 0.0713, 0.1411], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0188, 0.0158, 0.0216, 0.0268, 0.0168, 0.0191, 0.0246], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:56:05,979 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:33,580 INFO [train.py:904] (5/8) Epoch 2, batch 8300, loss[loss=0.2507, simple_loss=0.3153, pruned_loss=0.09312, over 11822.00 frames. ], tot_loss[loss=0.2857, simple_loss=0.3525, pruned_loss=0.1095, over 3055801.60 frames. ], batch size: 246, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:57:02,916 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5865, 3.7875, 3.0919, 2.6483, 2.8834, 2.1410, 4.0626, 4.4798], device='cuda:5'), covar=tensor([0.1859, 0.0564, 0.1012, 0.0812, 0.1763, 0.1336, 0.0240, 0.0121], device='cuda:5'), in_proj_covar=tensor([0.0253, 0.0221, 0.0236, 0.0178, 0.0261, 0.0185, 0.0195, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:57:37,595 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:57:54,224 INFO [train.py:904] (5/8) Epoch 2, batch 8350, loss[loss=0.2605, simple_loss=0.3418, pruned_loss=0.08958, over 16724.00 frames. ], tot_loss[loss=0.2806, simple_loss=0.3499, pruned_loss=0.1057, over 3045521.77 frames. ], batch size: 124, lr: 2.50e-02, grad_scale: 4.0 2023-04-27 19:58:06,850 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:58:14,612 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7218, 3.6503, 4.2013, 4.1577, 4.2067, 3.8192, 3.8983, 3.9843], device='cuda:5'), covar=tensor([0.0220, 0.0361, 0.0246, 0.0307, 0.0270, 0.0235, 0.0604, 0.0223], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0148, 0.0164, 0.0161, 0.0194, 0.0166, 0.0253, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-27 19:58:28,821 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.204e+02 3.865e+02 4.910e+02 6.166e+02 1.131e+03, threshold=9.820e+02, percent-clipped=1.0 2023-04-27 19:58:42,222 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:58:44,082 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:58:44,174 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2845, 3.1475, 3.3023, 3.5135, 3.4659, 3.1925, 3.4574, 3.4885], device='cuda:5'), covar=tensor([0.0493, 0.0537, 0.0931, 0.0384, 0.0458, 0.1010, 0.0412, 0.0303], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0284, 0.0372, 0.0282, 0.0213, 0.0198, 0.0222, 0.0224], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 19:59:15,562 INFO [train.py:904] (5/8) Epoch 2, batch 8400, loss[loss=0.2768, simple_loss=0.3395, pruned_loss=0.1071, over 15174.00 frames. ], tot_loss[loss=0.2742, simple_loss=0.3456, pruned_loss=0.1014, over 3070401.66 frames. ], batch size: 190, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 19:59:16,157 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:37,138 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2687, 3.0337, 2.8783, 1.8656, 2.6150, 2.0482, 2.8221, 3.0347], device='cuda:5'), covar=tensor([0.0303, 0.0530, 0.0381, 0.1439, 0.0622, 0.0916, 0.0588, 0.0461], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0102, 0.0151, 0.0149, 0.0140, 0.0134, 0.0145, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 19:59:45,648 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 20:00:14,473 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7815, 3.7412, 1.7545, 3.8279, 2.3835, 3.9209, 1.8766, 2.8570], device='cuda:5'), covar=tensor([0.0046, 0.0181, 0.1581, 0.0033, 0.0832, 0.0198, 0.1448, 0.0552], device='cuda:5'), in_proj_covar=tensor([0.0074, 0.0109, 0.0171, 0.0073, 0.0153, 0.0135, 0.0176, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 20:00:35,060 INFO [train.py:904] (5/8) Epoch 2, batch 8450, loss[loss=0.2468, simple_loss=0.3168, pruned_loss=0.0884, over 12267.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3425, pruned_loss=0.0987, over 3067514.77 frames. ], batch size: 247, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 20:00:59,130 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9992, 2.0664, 1.7718, 1.8519, 2.6762, 2.6238, 3.2401, 2.9081], device='cuda:5'), covar=tensor([0.0017, 0.0158, 0.0179, 0.0186, 0.0072, 0.0116, 0.0029, 0.0069], device='cuda:5'), in_proj_covar=tensor([0.0048, 0.0102, 0.0105, 0.0106, 0.0094, 0.0104, 0.0056, 0.0078], device='cuda:5'), out_proj_covar=tensor([6.8910e-05, 1.5674e-04, 1.5757e-04, 1.6412e-04, 1.5010e-04, 1.6460e-04, 8.4764e-05, 1.2580e-04], device='cuda:5') 2023-04-27 20:01:00,314 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:01:04,825 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8366, 4.8474, 4.7892, 2.0008, 3.0858, 2.8270, 4.0846, 4.7977], device='cuda:5'), covar=tensor([0.0232, 0.0347, 0.0215, 0.1600, 0.0714, 0.0952, 0.0682, 0.0509], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0102, 0.0150, 0.0150, 0.0141, 0.0136, 0.0147, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 20:01:09,359 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.749e+02 4.009e+02 5.012e+02 6.279e+02 1.629e+03, threshold=1.002e+03, percent-clipped=8.0 2023-04-27 20:01:55,560 INFO [train.py:904] (5/8) Epoch 2, batch 8500, loss[loss=0.2588, simple_loss=0.3274, pruned_loss=0.09512, over 16203.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3364, pruned_loss=0.09376, over 3072508.18 frames. ], batch size: 165, lr: 2.49e-02, grad_scale: 8.0 2023-04-27 20:02:17,840 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:02:34,562 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:03:19,442 INFO [train.py:904] (5/8) Epoch 2, batch 8550, loss[loss=0.2515, simple_loss=0.3154, pruned_loss=0.09378, over 11909.00 frames. ], tot_loss[loss=0.2592, simple_loss=0.3337, pruned_loss=0.09235, over 3064329.83 frames. ], batch size: 247, lr: 2.49e-02, grad_scale: 4.0 2023-04-27 20:03:35,169 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-27 20:04:03,004 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.327e+02 4.014e+02 4.987e+02 6.688e+02 1.627e+03, threshold=9.973e+02, percent-clipped=3.0 2023-04-27 20:04:26,484 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:04:47,322 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7565, 3.5145, 3.3470, 1.6059, 2.8015, 2.0125, 3.2896, 3.5959], device='cuda:5'), covar=tensor([0.0214, 0.0348, 0.0330, 0.1581, 0.0636, 0.1021, 0.0521, 0.0377], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0102, 0.0152, 0.0151, 0.0141, 0.0137, 0.0146, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 20:04:58,647 INFO [train.py:904] (5/8) Epoch 2, batch 8600, loss[loss=0.3012, simple_loss=0.3672, pruned_loss=0.1176, over 16672.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.3347, pruned_loss=0.0922, over 3049724.63 frames. ], batch size: 89, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:06:37,154 INFO [train.py:904] (5/8) Epoch 2, batch 8650, loss[loss=0.2226, simple_loss=0.3137, pruned_loss=0.06573, over 16185.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3317, pruned_loss=0.08966, over 3037176.27 frames. ], batch size: 165, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:07:29,099 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.199e+02 3.846e+02 4.801e+02 6.827e+02 6.621e+03, threshold=9.601e+02, percent-clipped=16.0 2023-04-27 20:07:44,377 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:46,551 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:14,608 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:23,076 INFO [train.py:904] (5/8) Epoch 2, batch 8700, loss[loss=0.2365, simple_loss=0.3204, pruned_loss=0.07634, over 16327.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3274, pruned_loss=0.08681, over 3029536.92 frames. ], batch size: 146, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:08:48,202 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2216, 4.0544, 3.7283, 1.5493, 3.1097, 2.1222, 3.6149, 3.9547], device='cuda:5'), covar=tensor([0.0186, 0.0274, 0.0362, 0.1840, 0.0635, 0.1107, 0.0571, 0.0414], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0099, 0.0150, 0.0148, 0.0138, 0.0133, 0.0142, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 20:08:49,491 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 20:08:49,845 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 20:09:14,854 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:09:17,251 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18879.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:09:35,944 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9356, 1.4919, 1.8971, 2.5520, 2.6985, 2.5059, 1.6725, 2.6754], device='cuda:5'), covar=tensor([0.0029, 0.0236, 0.0149, 0.0117, 0.0030, 0.0063, 0.0173, 0.0049], device='cuda:5'), in_proj_covar=tensor([0.0070, 0.0109, 0.0095, 0.0084, 0.0062, 0.0056, 0.0096, 0.0054], device='cuda:5'), out_proj_covar=tensor([1.1804e-04, 1.8688e-04, 1.6885e-04, 1.5015e-04, 1.0397e-04, 9.5822e-05, 1.6251e-04, 8.9902e-05], device='cuda:5') 2023-04-27 20:10:01,030 INFO [train.py:904] (5/8) Epoch 2, batch 8750, loss[loss=0.2287, simple_loss=0.3058, pruned_loss=0.0758, over 12158.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3267, pruned_loss=0.08564, over 3042587.57 frames. ], batch size: 246, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:10:21,505 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0032, 3.7403, 3.4800, 1.6471, 2.9146, 2.2475, 3.1749, 3.6659], device='cuda:5'), covar=tensor([0.0246, 0.0454, 0.0345, 0.1590, 0.0668, 0.0848, 0.0850, 0.0454], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0098, 0.0149, 0.0147, 0.0138, 0.0131, 0.0140, 0.0099], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 20:10:57,518 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.382e+02 3.853e+02 4.840e+02 6.218e+02 1.090e+03, threshold=9.680e+02, percent-clipped=4.0 2023-04-27 20:11:10,243 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18930.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:11:53,579 INFO [train.py:904] (5/8) Epoch 2, batch 8800, loss[loss=0.2582, simple_loss=0.3232, pruned_loss=0.09657, over 12618.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3244, pruned_loss=0.08376, over 3049844.66 frames. ], batch size: 248, lr: 2.48e-02, grad_scale: 4.0 2023-04-27 20:12:02,747 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1054, 5.4864, 5.2168, 5.3763, 4.9880, 4.6612, 5.1415, 5.4961], device='cuda:5'), covar=tensor([0.0327, 0.0527, 0.0509, 0.0249, 0.0357, 0.0386, 0.0320, 0.0528], device='cuda:5'), in_proj_covar=tensor([0.0213, 0.0298, 0.0259, 0.0186, 0.0200, 0.0186, 0.0241, 0.0206], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:12:09,039 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 20:13:18,231 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:13:38,738 INFO [train.py:904] (5/8) Epoch 2, batch 8850, loss[loss=0.2295, simple_loss=0.3253, pruned_loss=0.06682, over 16781.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3257, pruned_loss=0.08155, over 3059605.75 frames. ], batch size: 124, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:14:28,282 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.418e+02 3.744e+02 4.791e+02 5.972e+02 1.134e+03, threshold=9.582e+02, percent-clipped=2.0 2023-04-27 20:14:43,728 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:15:27,246 INFO [train.py:904] (5/8) Epoch 2, batch 8900, loss[loss=0.2538, simple_loss=0.3359, pruned_loss=0.08587, over 16643.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3253, pruned_loss=0.08043, over 3065746.29 frames. ], batch size: 89, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:16:37,093 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9958, 4.0269, 4.4729, 4.4830, 4.4280, 3.9615, 4.0836, 4.1415], device='cuda:5'), covar=tensor([0.0213, 0.0289, 0.0321, 0.0360, 0.0396, 0.0274, 0.0668, 0.0261], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0142, 0.0160, 0.0155, 0.0188, 0.0160, 0.0236, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-04-27 20:17:32,232 INFO [train.py:904] (5/8) Epoch 2, batch 8950, loss[loss=0.2321, simple_loss=0.3243, pruned_loss=0.07001, over 16952.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3254, pruned_loss=0.08093, over 3095518.23 frames. ], batch size: 116, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:18:20,859 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.564e+02 4.048e+02 4.879e+02 5.848e+02 1.298e+03, threshold=9.758e+02, percent-clipped=2.0 2023-04-27 20:18:32,702 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:19:11,878 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:20,924 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-04-27 20:19:21,324 INFO [train.py:904] (5/8) Epoch 2, batch 9000, loss[loss=0.2235, simple_loss=0.3129, pruned_loss=0.06708, over 16894.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3214, pruned_loss=0.07865, over 3086483.56 frames. ], batch size: 102, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:19:21,324 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 20:19:31,141 INFO [train.py:938] (5/8) Epoch 2, validation: loss=0.2044, simple_loss=0.3047, pruned_loss=0.05205, over 944034.00 frames. 2023-04-27 20:19:31,141 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-27 20:20:00,889 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:20:52,957 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 20:21:02,823 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:21:13,857 INFO [train.py:904] (5/8) Epoch 2, batch 9050, loss[loss=0.2068, simple_loss=0.2928, pruned_loss=0.06037, over 16810.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3231, pruned_loss=0.08052, over 3083434.33 frames. ], batch size: 90, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:21:37,537 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:21:53,747 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5428, 3.6169, 3.0131, 2.4636, 2.6011, 2.2355, 3.6259, 4.1134], device='cuda:5'), covar=tensor([0.1847, 0.0563, 0.0947, 0.0820, 0.1425, 0.1176, 0.0330, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0245, 0.0215, 0.0235, 0.0178, 0.0210, 0.0182, 0.0194, 0.0139], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:22:01,200 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.755e+02 4.212e+02 4.597e+02 5.989e+02 9.578e+02, threshold=9.193e+02, percent-clipped=0.0 2023-04-27 20:22:58,697 INFO [train.py:904] (5/8) Epoch 2, batch 9100, loss[loss=0.236, simple_loss=0.3243, pruned_loss=0.07384, over 16699.00 frames. ], tot_loss[loss=0.243, simple_loss=0.323, pruned_loss=0.08147, over 3074323.11 frames. ], batch size: 76, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:24:23,837 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:24:39,740 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2803, 3.1332, 3.2197, 3.5081, 3.4269, 3.1753, 3.4294, 3.4476], device='cuda:5'), covar=tensor([0.0409, 0.0490, 0.0862, 0.0332, 0.0369, 0.1060, 0.0457, 0.0336], device='cuda:5'), in_proj_covar=tensor([0.0234, 0.0273, 0.0365, 0.0273, 0.0208, 0.0192, 0.0219, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:24:55,907 INFO [train.py:904] (5/8) Epoch 2, batch 9150, loss[loss=0.2621, simple_loss=0.339, pruned_loss=0.09261, over 12074.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3239, pruned_loss=0.08143, over 3061607.70 frames. ], batch size: 248, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:25:49,060 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-27 20:25:49,334 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.714e+02 4.631e+02 5.471e+02 6.935e+02 2.154e+03, threshold=1.094e+03, percent-clipped=5.0 2023-04-27 20:25:59,131 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:26:16,403 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3207, 3.2605, 2.5430, 2.2374, 2.2506, 1.9685, 3.2113, 3.6021], device='cuda:5'), covar=tensor([0.1746, 0.0669, 0.1040, 0.0821, 0.1474, 0.1293, 0.0346, 0.0180], device='cuda:5'), in_proj_covar=tensor([0.0250, 0.0226, 0.0240, 0.0180, 0.0214, 0.0185, 0.0196, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:26:40,136 INFO [train.py:904] (5/8) Epoch 2, batch 9200, loss[loss=0.2155, simple_loss=0.2979, pruned_loss=0.0666, over 16250.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3186, pruned_loss=0.07972, over 3077606.93 frames. ], batch size: 165, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:27:19,089 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2039, 2.4546, 2.2031, 3.5803, 1.9346, 3.4309, 2.3110, 2.1138], device='cuda:5'), covar=tensor([0.0278, 0.0468, 0.0334, 0.0176, 0.1327, 0.0151, 0.0622, 0.1024], device='cuda:5'), in_proj_covar=tensor([0.0219, 0.0195, 0.0161, 0.0220, 0.0267, 0.0174, 0.0197, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:27:30,641 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:03,621 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:16,043 INFO [train.py:904] (5/8) Epoch 2, batch 9250, loss[loss=0.2201, simple_loss=0.307, pruned_loss=0.06661, over 16444.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3186, pruned_loss=0.07996, over 3061464.25 frames. ], batch size: 146, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:29:05,882 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.723e+02 4.011e+02 4.767e+02 6.782e+02 2.707e+03, threshold=9.534e+02, percent-clipped=7.0 2023-04-27 20:30:06,020 INFO [train.py:904] (5/8) Epoch 2, batch 9300, loss[loss=0.2375, simple_loss=0.3117, pruned_loss=0.08158, over 16426.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3165, pruned_loss=0.07885, over 3064034.39 frames. ], batch size: 146, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:30:14,302 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:31:20,848 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 20:31:30,494 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.20 vs. limit=5.0 2023-04-27 20:31:49,026 INFO [train.py:904] (5/8) Epoch 2, batch 9350, loss[loss=0.2343, simple_loss=0.308, pruned_loss=0.0803, over 12448.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3159, pruned_loss=0.07847, over 3065535.04 frames. ], batch size: 250, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:32:26,141 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 20:32:37,319 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.597e+02 3.883e+02 4.622e+02 6.098e+02 2.142e+03, threshold=9.245e+02, percent-clipped=3.0 2023-04-27 20:33:28,471 INFO [train.py:904] (5/8) Epoch 2, batch 9400, loss[loss=0.1882, simple_loss=0.2706, pruned_loss=0.05287, over 12201.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3158, pruned_loss=0.07786, over 3062016.38 frames. ], batch size: 247, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:34:39,293 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:34:49,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1158, 3.7722, 3.6264, 2.6145, 3.4827, 3.7500, 3.7517, 1.8552], device='cuda:5'), covar=tensor([0.0408, 0.0021, 0.0039, 0.0248, 0.0028, 0.0028, 0.0017, 0.0398], device='cuda:5'), in_proj_covar=tensor([0.0109, 0.0053, 0.0056, 0.0103, 0.0051, 0.0056, 0.0058, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 20:35:08,268 INFO [train.py:904] (5/8) Epoch 2, batch 9450, loss[loss=0.243, simple_loss=0.3184, pruned_loss=0.08378, over 12819.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.318, pruned_loss=0.07836, over 3071115.47 frames. ], batch size: 248, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:35:42,242 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:35:56,424 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.451e+02 4.122e+02 5.164e+02 6.390e+02 1.240e+03, threshold=1.033e+03, percent-clipped=6.0 2023-04-27 20:36:16,207 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:36:48,230 INFO [train.py:904] (5/8) Epoch 2, batch 9500, loss[loss=0.2218, simple_loss=0.306, pruned_loss=0.0688, over 15380.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3172, pruned_loss=0.07764, over 3072524.99 frames. ], batch size: 191, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:37:46,280 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:38:34,562 INFO [train.py:904] (5/8) Epoch 2, batch 9550, loss[loss=0.2096, simple_loss=0.3001, pruned_loss=0.05954, over 16870.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3167, pruned_loss=0.07773, over 3055648.97 frames. ], batch size: 96, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:39:23,659 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.908e+02 4.753e+02 5.956e+02 1.328e+03, threshold=9.507e+02, percent-clipped=2.0 2023-04-27 20:40:12,708 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19750.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:40:13,484 INFO [train.py:904] (5/8) Epoch 2, batch 9600, loss[loss=0.2322, simple_loss=0.3046, pruned_loss=0.07985, over 12563.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3174, pruned_loss=0.07838, over 3050029.28 frames. ], batch size: 250, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:40:21,677 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4551, 2.5169, 2.1674, 3.7571, 1.8036, 3.5799, 2.2666, 2.2060], device='cuda:5'), covar=tensor([0.0278, 0.0532, 0.0387, 0.0183, 0.1479, 0.0139, 0.0741, 0.1165], device='cuda:5'), in_proj_covar=tensor([0.0216, 0.0196, 0.0158, 0.0217, 0.0263, 0.0172, 0.0194, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:40:38,631 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4683, 3.3130, 3.3485, 3.0884, 3.3028, 2.1020, 3.2150, 3.1985], device='cuda:5'), covar=tensor([0.0065, 0.0053, 0.0068, 0.0162, 0.0057, 0.1050, 0.0066, 0.0085], device='cuda:5'), in_proj_covar=tensor([0.0059, 0.0049, 0.0074, 0.0082, 0.0056, 0.0109, 0.0067, 0.0073], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:41:22,062 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:41:37,755 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 20:42:00,145 INFO [train.py:904] (5/8) Epoch 2, batch 9650, loss[loss=0.2521, simple_loss=0.3211, pruned_loss=0.09156, over 12472.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3208, pruned_loss=0.07991, over 3055262.28 frames. ], batch size: 248, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:42:54,374 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.395e+02 4.451e+02 5.149e+02 6.499e+02 1.556e+03, threshold=1.030e+03, percent-clipped=6.0 2023-04-27 20:43:10,778 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:43:48,664 INFO [train.py:904] (5/8) Epoch 2, batch 9700, loss[loss=0.2312, simple_loss=0.3081, pruned_loss=0.07718, over 15347.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3195, pruned_loss=0.07934, over 3060856.54 frames. ], batch size: 190, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:45:10,298 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 20:45:33,550 INFO [train.py:904] (5/8) Epoch 2, batch 9750, loss[loss=0.2334, simple_loss=0.3202, pruned_loss=0.07325, over 16790.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3182, pruned_loss=0.07944, over 3048097.12 frames. ], batch size: 124, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:46:21,258 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.379e+02 3.927e+02 4.879e+02 5.766e+02 1.284e+03, threshold=9.758e+02, percent-clipped=1.0 2023-04-27 20:46:29,857 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8151, 2.7655, 1.4792, 2.7899, 1.9773, 2.7501, 1.8026, 2.3651], device='cuda:5'), covar=tensor([0.0065, 0.0263, 0.1435, 0.0055, 0.0685, 0.0449, 0.1286, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0076, 0.0115, 0.0171, 0.0073, 0.0152, 0.0136, 0.0179, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 20:47:15,375 INFO [train.py:904] (5/8) Epoch 2, batch 9800, loss[loss=0.2191, simple_loss=0.3112, pruned_loss=0.06345, over 16915.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.318, pruned_loss=0.0778, over 3076191.90 frames. ], batch size: 116, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:48:00,442 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:48:00,587 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1381, 3.8403, 3.6845, 2.7141, 3.5256, 3.5455, 3.9729, 1.6514], device='cuda:5'), covar=tensor([0.0387, 0.0018, 0.0035, 0.0196, 0.0028, 0.0044, 0.0012, 0.0410], device='cuda:5'), in_proj_covar=tensor([0.0108, 0.0053, 0.0057, 0.0101, 0.0051, 0.0056, 0.0055, 0.0101], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0001, 0.0003], device='cuda:5') 2023-04-27 20:49:05,350 INFO [train.py:904] (5/8) Epoch 2, batch 9850, loss[loss=0.2455, simple_loss=0.324, pruned_loss=0.08348, over 16135.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3194, pruned_loss=0.07735, over 3091521.55 frames. ], batch size: 165, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:49:46,068 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:49:55,605 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.468e+02 4.069e+02 4.622e+02 5.694e+02 1.429e+03, threshold=9.244e+02, percent-clipped=4.0 2023-04-27 20:50:58,208 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20050.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:50:58,969 INFO [train.py:904] (5/8) Epoch 2, batch 9900, loss[loss=0.2516, simple_loss=0.3191, pruned_loss=0.09203, over 12201.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3195, pruned_loss=0.07704, over 3087520.67 frames. ], batch size: 248, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:51:04,644 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7975, 1.1648, 1.4239, 1.7862, 1.8426, 1.6509, 1.4745, 1.8317], device='cuda:5'), covar=tensor([0.0058, 0.0158, 0.0105, 0.0090, 0.0043, 0.0074, 0.0135, 0.0044], device='cuda:5'), in_proj_covar=tensor([0.0077, 0.0111, 0.0099, 0.0085, 0.0065, 0.0060, 0.0100, 0.0055], device='cuda:5'), out_proj_covar=tensor([1.3101e-04, 1.8759e-04, 1.7258e-04, 1.4814e-04, 1.0582e-04, 9.9604e-05, 1.6579e-04, 8.8637e-05], device='cuda:5') 2023-04-27 20:52:04,581 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8312, 3.7801, 3.6176, 3.7467, 3.3039, 3.7622, 3.6027, 3.4439], device='cuda:5'), covar=tensor([0.0300, 0.0165, 0.0197, 0.0135, 0.0585, 0.0189, 0.0436, 0.0312], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0091, 0.0139, 0.0116, 0.0164, 0.0123, 0.0099, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:52:12,642 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:24,465 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 2023-04-27 20:52:31,333 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:51,148 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:56,920 INFO [train.py:904] (5/8) Epoch 2, batch 9950, loss[loss=0.2049, simple_loss=0.2929, pruned_loss=0.05847, over 16566.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3223, pruned_loss=0.07774, over 3103163.16 frames. ], batch size: 68, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:53:14,425 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1004, 4.1067, 4.5860, 4.6317, 4.6511, 4.1445, 4.2347, 4.1971], device='cuda:5'), covar=tensor([0.0198, 0.0311, 0.0281, 0.0281, 0.0263, 0.0248, 0.0565, 0.0256], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0141, 0.0156, 0.0158, 0.0183, 0.0162, 0.0234, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-04-27 20:53:17,128 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9034, 2.3838, 2.0621, 3.1986, 2.0058, 3.0412, 2.2890, 2.0297], device='cuda:5'), covar=tensor([0.0301, 0.0507, 0.0371, 0.0218, 0.1244, 0.0171, 0.0646, 0.1129], device='cuda:5'), in_proj_covar=tensor([0.0219, 0.0201, 0.0163, 0.0223, 0.0270, 0.0174, 0.0197, 0.0251], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:53:21,558 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2041, 3.9938, 3.9480, 3.3878, 3.8662, 1.5620, 3.6472, 3.8273], device='cuda:5'), covar=tensor([0.0065, 0.0060, 0.0069, 0.0235, 0.0073, 0.1509, 0.0080, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0059, 0.0050, 0.0074, 0.0081, 0.0056, 0.0112, 0.0067, 0.0073], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:53:23,639 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:53:55,579 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 4.201e+02 4.972e+02 6.369e+02 1.989e+03, threshold=9.943e+02, percent-clipped=8.0 2023-04-27 20:54:23,259 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7100, 5.3026, 5.2455, 5.3298, 5.2966, 5.7121, 5.5175, 5.2806], device='cuda:5'), covar=tensor([0.0641, 0.0947, 0.0857, 0.1312, 0.1552, 0.0648, 0.0659, 0.1776], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0257, 0.0237, 0.0231, 0.0284, 0.0259, 0.0195, 0.0297], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 20:54:56,062 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:54:56,745 INFO [train.py:904] (5/8) Epoch 2, batch 10000, loss[loss=0.2223, simple_loss=0.3007, pruned_loss=0.07194, over 12727.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3199, pruned_loss=0.07666, over 3098185.54 frames. ], batch size: 250, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:55:41,437 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:55:47,678 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7880, 1.3489, 1.7509, 2.4957, 2.4413, 2.6488, 1.3062, 2.5343], device='cuda:5'), covar=tensor([0.0039, 0.0268, 0.0172, 0.0110, 0.0058, 0.0070, 0.0232, 0.0065], device='cuda:5'), in_proj_covar=tensor([0.0077, 0.0114, 0.0101, 0.0087, 0.0068, 0.0061, 0.0103, 0.0057], device='cuda:5'), out_proj_covar=tensor([1.3098e-04, 1.9314e-04, 1.7552e-04, 1.5100e-04, 1.1102e-04, 9.9758e-05, 1.7039e-04, 9.1924e-05], device='cuda:5') 2023-04-27 20:56:37,819 INFO [train.py:904] (5/8) Epoch 2, batch 10050, loss[loss=0.2324, simple_loss=0.3177, pruned_loss=0.0735, over 16798.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3195, pruned_loss=0.07612, over 3095239.12 frames. ], batch size: 83, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:56:44,652 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:57:24,816 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.701e+02 4.201e+02 4.991e+02 6.585e+02 1.392e+03, threshold=9.982e+02, percent-clipped=3.0 2023-04-27 20:58:10,201 INFO [train.py:904] (5/8) Epoch 2, batch 10100, loss[loss=0.2172, simple_loss=0.3032, pruned_loss=0.06558, over 17067.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3204, pruned_loss=0.07717, over 3081484.04 frames. ], batch size: 53, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:58:38,034 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:58:58,922 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:59:54,879 INFO [train.py:904] (5/8) Epoch 3, batch 0, loss[loss=0.4685, simple_loss=0.4496, pruned_loss=0.2437, over 16823.00 frames. ], tot_loss[loss=0.4685, simple_loss=0.4496, pruned_loss=0.2437, over 16823.00 frames. ], batch size: 83, lr: 2.28e-02, grad_scale: 8.0 2023-04-27 20:59:54,879 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 21:00:02,300 INFO [train.py:938] (5/8) Epoch 3, validation: loss=0.2012, simple_loss=0.3019, pruned_loss=0.05024, over 944034.00 frames. 2023-04-27 21:00:02,300 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-27 21:00:17,869 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 21:00:31,875 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:00:38,800 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.937e+02 4.805e+02 5.776e+02 7.144e+02 1.042e+03, threshold=1.155e+03, percent-clipped=3.0 2023-04-27 21:01:12,904 INFO [train.py:904] (5/8) Epoch 3, batch 50, loss[loss=0.2728, simple_loss=0.3548, pruned_loss=0.0954, over 17291.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.3495, pruned_loss=0.118, over 748667.46 frames. ], batch size: 52, lr: 2.28e-02, grad_scale: 2.0 2023-04-27 21:01:45,738 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:02,836 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:19,623 INFO [train.py:904] (5/8) Epoch 3, batch 100, loss[loss=0.3166, simple_loss=0.3618, pruned_loss=0.1357, over 16257.00 frames. ], tot_loss[loss=0.2851, simple_loss=0.3438, pruned_loss=0.1132, over 1315764.65 frames. ], batch size: 165, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:02:23,264 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 21:02:42,065 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8561, 3.4007, 3.4421, 2.4551, 3.2710, 3.2622, 3.5484, 1.8555], device='cuda:5'), covar=tensor([0.0391, 0.0029, 0.0045, 0.0232, 0.0047, 0.0079, 0.0024, 0.0361], device='cuda:5'), in_proj_covar=tensor([0.0113, 0.0056, 0.0061, 0.0108, 0.0055, 0.0060, 0.0059, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:02:43,251 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:54,678 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.729e+02 4.335e+02 5.443e+02 6.351e+02 1.481e+03, threshold=1.089e+03, percent-clipped=3.0 2023-04-27 21:03:18,800 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:24,465 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:27,025 INFO [train.py:904] (5/8) Epoch 3, batch 150, loss[loss=0.2362, simple_loss=0.3151, pruned_loss=0.0787, over 17212.00 frames. ], tot_loss[loss=0.2779, simple_loss=0.3384, pruned_loss=0.1087, over 1762874.45 frames. ], batch size: 44, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:03:49,815 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:05,890 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:35,438 INFO [train.py:904] (5/8) Epoch 3, batch 200, loss[loss=0.2948, simple_loss=0.3495, pruned_loss=0.1201, over 16296.00 frames. ], tot_loss[loss=0.278, simple_loss=0.3387, pruned_loss=0.1087, over 2099691.52 frames. ], batch size: 145, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:09,763 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.165e+02 3.891e+02 4.616e+02 5.990e+02 1.220e+03, threshold=9.233e+02, percent-clipped=2.0 2023-04-27 21:05:43,642 INFO [train.py:904] (5/8) Epoch 3, batch 250, loss[loss=0.257, simple_loss=0.3195, pruned_loss=0.0972, over 16853.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3348, pruned_loss=0.1049, over 2361605.34 frames. ], batch size: 116, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:58,368 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:18,617 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6392, 4.7558, 4.7055, 4.8476, 4.8312, 5.2860, 5.1591, 4.8233], device='cuda:5'), covar=tensor([0.0947, 0.1271, 0.1110, 0.1486, 0.2132, 0.0960, 0.0795, 0.1985], device='cuda:5'), in_proj_covar=tensor([0.0202, 0.0301, 0.0282, 0.0270, 0.0344, 0.0297, 0.0229, 0.0352], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:06:35,612 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:54,935 INFO [train.py:904] (5/8) Epoch 3, batch 300, loss[loss=0.2584, simple_loss=0.3288, pruned_loss=0.09396, over 16724.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3304, pruned_loss=0.1014, over 2564672.48 frames. ], batch size: 57, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:07:29,016 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.643e+02 3.919e+02 4.746e+02 5.726e+02 1.058e+03, threshold=9.491e+02, percent-clipped=1.0 2023-04-27 21:07:50,734 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7111, 2.8074, 2.4505, 4.0545, 2.0419, 3.7643, 2.3785, 2.5676], device='cuda:5'), covar=tensor([0.0303, 0.0538, 0.0388, 0.0183, 0.1508, 0.0198, 0.0787, 0.1057], device='cuda:5'), in_proj_covar=tensor([0.0230, 0.0207, 0.0171, 0.0234, 0.0279, 0.0178, 0.0203, 0.0267], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:07:59,123 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20649.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:08:01,110 INFO [train.py:904] (5/8) Epoch 3, batch 350, loss[loss=0.1901, simple_loss=0.2746, pruned_loss=0.05283, over 16951.00 frames. ], tot_loss[loss=0.2604, simple_loss=0.3251, pruned_loss=0.0978, over 2736094.36 frames. ], batch size: 41, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:08:32,086 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5740, 2.9615, 3.2231, 2.1002, 3.1899, 3.0811, 3.2036, 1.7427], device='cuda:5'), covar=tensor([0.0413, 0.0057, 0.0040, 0.0250, 0.0039, 0.0069, 0.0031, 0.0337], device='cuda:5'), in_proj_covar=tensor([0.0112, 0.0057, 0.0060, 0.0108, 0.0055, 0.0062, 0.0060, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:08:36,845 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:09,381 INFO [train.py:904] (5/8) Epoch 3, batch 400, loss[loss=0.2699, simple_loss=0.3276, pruned_loss=0.1061, over 15566.00 frames. ], tot_loss[loss=0.2599, simple_loss=0.3244, pruned_loss=0.09773, over 2863688.34 frames. ], batch size: 190, lr: 2.26e-02, grad_scale: 4.0 2023-04-27 21:09:41,233 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:45,250 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.989e+02 4.408e+02 5.242e+02 6.457e+02 1.269e+03, threshold=1.048e+03, percent-clipped=5.0 2023-04-27 21:10:08,721 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:10,751 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:17,854 INFO [train.py:904] (5/8) Epoch 3, batch 450, loss[loss=0.2844, simple_loss=0.3418, pruned_loss=0.1135, over 16894.00 frames. ], tot_loss[loss=0.2567, simple_loss=0.3215, pruned_loss=0.09597, over 2960495.15 frames. ], batch size: 90, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:10:34,574 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 21:10:41,842 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:49,495 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:14,839 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:24,063 INFO [train.py:904] (5/8) Epoch 3, batch 500, loss[loss=0.2066, simple_loss=0.2864, pruned_loss=0.06343, over 16884.00 frames. ], tot_loss[loss=0.2524, simple_loss=0.3188, pruned_loss=0.09304, over 3040299.70 frames. ], batch size: 42, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:11:46,164 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:00,589 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 4.223e+02 4.982e+02 5.898e+02 1.253e+03, threshold=9.964e+02, percent-clipped=1.0 2023-04-27 21:12:14,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1502, 1.6175, 2.4793, 3.0328, 3.1162, 3.3209, 1.6165, 3.1017], device='cuda:5'), covar=tensor([0.0039, 0.0193, 0.0104, 0.0080, 0.0035, 0.0063, 0.0158, 0.0039], device='cuda:5'), in_proj_covar=tensor([0.0078, 0.0113, 0.0100, 0.0090, 0.0070, 0.0062, 0.0099, 0.0057], device='cuda:5'), out_proj_covar=tensor([1.2999e-04, 1.8814e-04, 1.7271e-04, 1.5561e-04, 1.1261e-04, 1.0386e-04, 1.6156e-04, 9.2662e-05], device='cuda:5') 2023-04-27 21:12:34,374 INFO [train.py:904] (5/8) Epoch 3, batch 550, loss[loss=0.2735, simple_loss=0.3282, pruned_loss=0.1094, over 15572.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3171, pruned_loss=0.0913, over 3103196.40 frames. ], batch size: 190, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:12:45,327 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:40,843 INFO [train.py:904] (5/8) Epoch 3, batch 600, loss[loss=0.2859, simple_loss=0.321, pruned_loss=0.1254, over 16750.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3174, pruned_loss=0.09337, over 3150757.09 frames. ], batch size: 134, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:13:50,653 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:15,689 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.677e+02 4.207e+02 5.314e+02 6.523e+02 1.813e+03, threshold=1.063e+03, percent-clipped=4.0 2023-04-27 21:14:38,954 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:48,040 INFO [train.py:904] (5/8) Epoch 3, batch 650, loss[loss=0.2462, simple_loss=0.306, pruned_loss=0.09325, over 16253.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3144, pruned_loss=0.09194, over 3168148.08 frames. ], batch size: 165, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:14:59,537 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-27 21:15:02,820 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8348, 1.5132, 2.1582, 2.8107, 2.7224, 2.5461, 1.5565, 2.6237], device='cuda:5'), covar=tensor([0.0047, 0.0200, 0.0118, 0.0083, 0.0047, 0.0094, 0.0160, 0.0049], device='cuda:5'), in_proj_covar=tensor([0.0081, 0.0119, 0.0104, 0.0096, 0.0075, 0.0066, 0.0106, 0.0061], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 21:15:57,321 INFO [train.py:904] (5/8) Epoch 3, batch 700, loss[loss=0.1959, simple_loss=0.2806, pruned_loss=0.05554, over 17213.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3139, pruned_loss=0.09099, over 3196667.66 frames. ], batch size: 44, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:16:30,870 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.136e+02 3.749e+02 4.303e+02 5.331e+02 1.033e+03, threshold=8.606e+02, percent-clipped=0.0 2023-04-27 21:16:55,564 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:03,839 INFO [train.py:904] (5/8) Epoch 3, batch 750, loss[loss=0.2412, simple_loss=0.3204, pruned_loss=0.08101, over 17068.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3145, pruned_loss=0.09124, over 3219140.51 frames. ], batch size: 53, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:17:05,974 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:35,089 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:44,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0535, 3.2480, 4.0284, 3.0164, 3.8173, 3.9821, 3.9787, 2.2130], device='cuda:5'), covar=tensor([0.0409, 0.0174, 0.0038, 0.0198, 0.0048, 0.0061, 0.0025, 0.0344], device='cuda:5'), in_proj_covar=tensor([0.0113, 0.0060, 0.0061, 0.0108, 0.0055, 0.0063, 0.0061, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:17:58,574 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:10,817 INFO [train.py:904] (5/8) Epoch 3, batch 800, loss[loss=0.278, simple_loss=0.3307, pruned_loss=0.1127, over 16209.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3141, pruned_loss=0.09049, over 3244872.23 frames. ], batch size: 165, lr: 2.24e-02, grad_scale: 8.0 2023-04-27 21:18:27,378 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:34,927 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:40,094 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:47,525 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.745e+02 4.034e+02 4.669e+02 5.861e+02 1.175e+03, threshold=9.338e+02, percent-clipped=5.0 2023-04-27 21:19:20,173 INFO [train.py:904] (5/8) Epoch 3, batch 850, loss[loss=0.2247, simple_loss=0.3054, pruned_loss=0.072, over 17287.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3123, pruned_loss=0.08881, over 3257449.75 frames. ], batch size: 52, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:19:58,146 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:20:27,609 INFO [train.py:904] (5/8) Epoch 3, batch 900, loss[loss=0.1984, simple_loss=0.2715, pruned_loss=0.06268, over 15978.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3104, pruned_loss=0.08703, over 3273156.47 frames. ], batch size: 35, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:20:46,489 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6373, 4.4344, 4.3589, 4.4871, 4.0175, 4.4380, 4.4363, 4.1391], device='cuda:5'), covar=tensor([0.0382, 0.0275, 0.0208, 0.0142, 0.0821, 0.0228, 0.0301, 0.0299], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0123, 0.0188, 0.0151, 0.0224, 0.0166, 0.0134, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:21:03,015 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.660e+02 3.821e+02 4.761e+02 5.619e+02 1.173e+03, threshold=9.521e+02, percent-clipped=3.0 2023-04-27 21:21:26,707 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:21:35,859 INFO [train.py:904] (5/8) Epoch 3, batch 950, loss[loss=0.2302, simple_loss=0.3067, pruned_loss=0.07681, over 17235.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3111, pruned_loss=0.08707, over 3289185.69 frames. ], batch size: 45, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:21:47,419 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-04-27 21:22:25,581 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0284, 2.5933, 2.5708, 4.1417, 1.7659, 4.0176, 2.0723, 2.2688], device='cuda:5'), covar=tensor([0.0332, 0.0833, 0.0460, 0.0256, 0.2250, 0.0259, 0.1178, 0.2032], device='cuda:5'), in_proj_covar=tensor([0.0237, 0.0212, 0.0177, 0.0239, 0.0279, 0.0185, 0.0205, 0.0274], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:22:34,056 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:22:45,630 INFO [train.py:904] (5/8) Epoch 3, batch 1000, loss[loss=0.1966, simple_loss=0.2715, pruned_loss=0.06089, over 16847.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3095, pruned_loss=0.08634, over 3289898.43 frames. ], batch size: 42, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:22:46,776 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5138, 2.5675, 2.4538, 3.8192, 2.0065, 3.5930, 2.1454, 2.1984], device='cuda:5'), covar=tensor([0.0298, 0.0536, 0.0371, 0.0195, 0.1355, 0.0192, 0.0821, 0.1076], device='cuda:5'), in_proj_covar=tensor([0.0237, 0.0211, 0.0177, 0.0238, 0.0279, 0.0185, 0.0204, 0.0272], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:22:47,821 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4529, 4.4966, 4.4577, 4.5484, 4.3761, 4.9853, 4.7283, 4.4509], device='cuda:5'), covar=tensor([0.0974, 0.1274, 0.1342, 0.1569, 0.2505, 0.0987, 0.0940, 0.2171], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0310, 0.0291, 0.0274, 0.0356, 0.0310, 0.0249, 0.0363], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:23:01,279 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21312.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:21,765 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.770e+02 3.937e+02 4.742e+02 6.159e+02 9.838e+02, threshold=9.485e+02, percent-clipped=2.0 2023-04-27 21:23:23,999 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:55,375 INFO [train.py:904] (5/8) Epoch 3, batch 1050, loss[loss=0.2185, simple_loss=0.2946, pruned_loss=0.07118, over 16798.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.308, pruned_loss=0.08583, over 3290459.00 frames. ], batch size: 57, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:24:25,351 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:24:46,800 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:02,818 INFO [train.py:904] (5/8) Epoch 3, batch 1100, loss[loss=0.2303, simple_loss=0.294, pruned_loss=0.08331, over 15467.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3081, pruned_loss=0.08612, over 3300551.48 frames. ], batch size: 190, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:25:12,314 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:24,871 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 21:25:38,416 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.639e+02 4.233e+02 5.080e+02 6.558e+02 1.311e+03, threshold=1.016e+03, percent-clipped=3.0 2023-04-27 21:25:47,663 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8873, 5.3048, 4.9491, 5.1175, 4.5991, 4.5480, 4.8001, 5.3650], device='cuda:5'), covar=tensor([0.0421, 0.0592, 0.0798, 0.0320, 0.0544, 0.0548, 0.0471, 0.0527], device='cuda:5'), in_proj_covar=tensor([0.0268, 0.0390, 0.0331, 0.0232, 0.0254, 0.0231, 0.0297, 0.0260], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:26:10,403 INFO [train.py:904] (5/8) Epoch 3, batch 1150, loss[loss=0.2547, simple_loss=0.3095, pruned_loss=0.09997, over 16896.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3075, pruned_loss=0.08571, over 3290635.85 frames. ], batch size: 109, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:26:11,485 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5989, 3.5316, 1.7854, 3.6556, 2.4012, 3.6390, 1.8083, 2.8036], device='cuda:5'), covar=tensor([0.0059, 0.0218, 0.1441, 0.0056, 0.0747, 0.0309, 0.1232, 0.0535], device='cuda:5'), in_proj_covar=tensor([0.0084, 0.0130, 0.0170, 0.0082, 0.0158, 0.0160, 0.0180, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 21:26:42,693 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:27:19,792 INFO [train.py:904] (5/8) Epoch 3, batch 1200, loss[loss=0.2832, simple_loss=0.3286, pruned_loss=0.1189, over 12169.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3074, pruned_loss=0.08507, over 3297692.10 frames. ], batch size: 247, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:27:29,208 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9300, 2.3723, 2.1538, 3.1461, 2.0777, 3.0176, 2.3046, 2.1580], device='cuda:5'), covar=tensor([0.0311, 0.0572, 0.0390, 0.0237, 0.1311, 0.0227, 0.0751, 0.1092], device='cuda:5'), in_proj_covar=tensor([0.0244, 0.0216, 0.0181, 0.0246, 0.0283, 0.0187, 0.0209, 0.0275], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:27:56,784 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 3.899e+02 4.622e+02 5.754e+02 9.760e+02, threshold=9.243e+02, percent-clipped=0.0 2023-04-27 21:28:07,525 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3402, 5.6555, 5.3592, 5.5164, 4.8884, 4.8711, 5.2420, 5.7610], device='cuda:5'), covar=tensor([0.0443, 0.0563, 0.0634, 0.0299, 0.0550, 0.0458, 0.0376, 0.0542], device='cuda:5'), in_proj_covar=tensor([0.0262, 0.0381, 0.0326, 0.0231, 0.0252, 0.0229, 0.0294, 0.0259], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:28:27,752 INFO [train.py:904] (5/8) Epoch 3, batch 1250, loss[loss=0.2843, simple_loss=0.321, pruned_loss=0.1238, over 16887.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.308, pruned_loss=0.08588, over 3295184.24 frames. ], batch size: 96, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:29:30,704 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:29:32,881 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 21:29:38,038 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 21:29:39,781 INFO [train.py:904] (5/8) Epoch 3, batch 1300, loss[loss=0.2026, simple_loss=0.277, pruned_loss=0.06409, over 17177.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3074, pruned_loss=0.08499, over 3296893.63 frames. ], batch size: 46, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:29:56,013 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:17,383 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.341e+02 3.774e+02 4.586e+02 5.366e+02 8.286e+02, threshold=9.173e+02, percent-clipped=0.0 2023-04-27 21:30:20,243 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 21:30:49,969 INFO [train.py:904] (5/8) Epoch 3, batch 1350, loss[loss=0.1903, simple_loss=0.2764, pruned_loss=0.05211, over 17229.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3068, pruned_loss=0.08419, over 3300369.03 frames. ], batch size: 45, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:30:56,176 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:13,020 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:20,955 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:35,307 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:58,288 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2912, 3.6853, 3.6492, 1.4681, 3.7981, 3.7692, 3.1106, 2.7812], device='cuda:5'), covar=tensor([0.0808, 0.0088, 0.0137, 0.1481, 0.0070, 0.0054, 0.0326, 0.0426], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0080, 0.0080, 0.0147, 0.0077, 0.0072, 0.0113, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 21:31:58,965 INFO [train.py:904] (5/8) Epoch 3, batch 1400, loss[loss=0.241, simple_loss=0.303, pruned_loss=0.0895, over 16758.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3068, pruned_loss=0.08414, over 3305051.16 frames. ], batch size: 124, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:32:09,320 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:32:35,078 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.932e+02 4.012e+02 4.714e+02 6.047e+02 1.220e+03, threshold=9.428e+02, percent-clipped=2.0 2023-04-27 21:32:54,005 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5803, 4.6410, 4.5346, 4.7614, 4.6107, 5.1802, 4.8723, 4.6023], device='cuda:5'), covar=tensor([0.1079, 0.1136, 0.1152, 0.1557, 0.2239, 0.0798, 0.1030, 0.2088], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0321, 0.0298, 0.0276, 0.0364, 0.0319, 0.0251, 0.0371], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:33:07,177 INFO [train.py:904] (5/8) Epoch 3, batch 1450, loss[loss=0.2641, simple_loss=0.3066, pruned_loss=0.1108, over 16788.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3057, pruned_loss=0.08452, over 3311036.24 frames. ], batch size: 83, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:33:13,877 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:20,860 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6398, 4.4918, 2.0336, 4.7830, 2.7187, 4.6858, 2.2985, 3.1519], device='cuda:5'), covar=tensor([0.0045, 0.0197, 0.1515, 0.0032, 0.0772, 0.0244, 0.1407, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0088, 0.0136, 0.0172, 0.0085, 0.0158, 0.0164, 0.0181, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 21:33:23,981 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:39,327 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:44,592 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 21:34:14,647 INFO [train.py:904] (5/8) Epoch 3, batch 1500, loss[loss=0.2277, simple_loss=0.2842, pruned_loss=0.08556, over 16474.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3053, pruned_loss=0.08428, over 3317362.79 frames. ], batch size: 146, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:34:31,492 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4204, 5.3404, 5.0806, 5.1572, 4.6067, 5.0705, 5.1654, 4.7783], device='cuda:5'), covar=tensor([0.0268, 0.0134, 0.0152, 0.0107, 0.0712, 0.0184, 0.0139, 0.0236], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0123, 0.0189, 0.0154, 0.0223, 0.0167, 0.0134, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:34:43,999 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:47,246 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:52,715 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.241e+02 3.626e+02 4.234e+02 5.507e+02 9.675e+02, threshold=8.468e+02, percent-clipped=1.0 2023-04-27 21:35:12,357 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9221, 2.7916, 2.7421, 4.3140, 2.2255, 4.1401, 2.4141, 2.6599], device='cuda:5'), covar=tensor([0.0244, 0.0523, 0.0319, 0.0138, 0.1299, 0.0153, 0.0750, 0.0962], device='cuda:5'), in_proj_covar=tensor([0.0241, 0.0215, 0.0179, 0.0245, 0.0284, 0.0189, 0.0207, 0.0276], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:35:23,725 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:35:24,400 INFO [train.py:904] (5/8) Epoch 3, batch 1550, loss[loss=0.2689, simple_loss=0.3359, pruned_loss=0.1009, over 16695.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3077, pruned_loss=0.08632, over 3315615.77 frames. ], batch size: 62, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:35:32,817 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 21:35:55,512 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-27 21:36:31,487 INFO [train.py:904] (5/8) Epoch 3, batch 1600, loss[loss=0.2418, simple_loss=0.304, pruned_loss=0.08973, over 16852.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3109, pruned_loss=0.08822, over 3303236.33 frames. ], batch size: 102, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:36:43,088 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3332, 5.1425, 5.0828, 5.0389, 4.6061, 5.1498, 5.0885, 4.8059], device='cuda:5'), covar=tensor([0.0287, 0.0136, 0.0153, 0.0117, 0.0776, 0.0163, 0.0147, 0.0219], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0127, 0.0192, 0.0157, 0.0229, 0.0168, 0.0135, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:36:46,743 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:37:07,921 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.625e+02 3.946e+02 4.752e+02 6.203e+02 1.123e+03, threshold=9.504e+02, percent-clipped=2.0 2023-04-27 21:37:38,901 INFO [train.py:904] (5/8) Epoch 3, batch 1650, loss[loss=0.2148, simple_loss=0.2864, pruned_loss=0.07162, over 16761.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3135, pruned_loss=0.08944, over 3300566.84 frames. ], batch size: 39, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:37:39,926 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:02,228 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:03,298 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:04,773 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:24,353 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:35,833 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-04-27 21:38:50,907 INFO [train.py:904] (5/8) Epoch 3, batch 1700, loss[loss=0.2655, simple_loss=0.325, pruned_loss=0.103, over 16849.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3153, pruned_loss=0.08986, over 3304849.14 frames. ], batch size: 96, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:39:11,658 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:17,268 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0133, 4.3342, 3.0630, 5.3447, 5.2140, 4.5225, 1.8312, 3.5035], device='cuda:5'), covar=tensor([0.1233, 0.0288, 0.0955, 0.0065, 0.0195, 0.0295, 0.1293, 0.0546], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0130, 0.0162, 0.0073, 0.0143, 0.0136, 0.0155, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 21:39:28,012 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.555e+02 4.242e+02 5.121e+02 6.013e+02 1.262e+03, threshold=1.024e+03, percent-clipped=2.0 2023-04-27 21:39:31,576 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:33,270 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:57,287 INFO [train.py:904] (5/8) Epoch 3, batch 1750, loss[loss=0.2423, simple_loss=0.2966, pruned_loss=0.09406, over 16654.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3168, pruned_loss=0.09026, over 3316033.84 frames. ], batch size: 89, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:40:09,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4878, 4.0337, 4.1718, 2.9176, 3.8519, 4.2038, 3.9840, 2.1362], device='cuda:5'), covar=tensor([0.0349, 0.0036, 0.0038, 0.0234, 0.0049, 0.0055, 0.0036, 0.0432], device='cuda:5'), in_proj_covar=tensor([0.0112, 0.0058, 0.0060, 0.0108, 0.0054, 0.0060, 0.0062, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:41:06,921 INFO [train.py:904] (5/8) Epoch 3, batch 1800, loss[loss=0.2502, simple_loss=0.3312, pruned_loss=0.08458, over 17067.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3169, pruned_loss=0.09, over 3311257.52 frames. ], batch size: 53, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:41:25,938 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-27 21:41:30,488 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22119.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:41:44,217 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.299e+02 3.837e+02 4.558e+02 5.944e+02 1.243e+03, threshold=9.115e+02, percent-clipped=4.0 2023-04-27 21:41:45,896 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:42:13,738 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 21:42:14,096 INFO [train.py:904] (5/8) Epoch 3, batch 1850, loss[loss=0.2921, simple_loss=0.359, pruned_loss=0.1126, over 12594.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3179, pruned_loss=0.08999, over 3317465.87 frames. ], batch size: 246, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:42:47,480 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:42:58,767 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 21:43:09,384 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:22,345 INFO [train.py:904] (5/8) Epoch 3, batch 1900, loss[loss=0.2278, simple_loss=0.3008, pruned_loss=0.07736, over 17152.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.316, pruned_loss=0.08781, over 3324213.43 frames. ], batch size: 48, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:43:28,964 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1135, 4.1909, 3.2718, 5.2745, 5.2342, 4.4822, 2.3828, 3.7183], device='cuda:5'), covar=tensor([0.1219, 0.0308, 0.0805, 0.0077, 0.0165, 0.0304, 0.1046, 0.0495], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0129, 0.0161, 0.0073, 0.0141, 0.0135, 0.0152, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 21:43:31,124 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:42,670 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:49,370 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-27 21:44:02,290 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.402e+02 3.901e+02 4.998e+02 5.919e+02 1.840e+03, threshold=9.997e+02, percent-clipped=6.0 2023-04-27 21:44:12,164 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:12,212 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:32,424 INFO [train.py:904] (5/8) Epoch 3, batch 1950, loss[loss=0.2688, simple_loss=0.3279, pruned_loss=0.1048, over 16849.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3154, pruned_loss=0.08704, over 3322929.05 frames. ], batch size: 90, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:44:32,849 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:36,137 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2486, 4.3015, 4.1600, 4.3815, 3.3472, 4.3132, 4.3076, 3.9484], device='cuda:5'), covar=tensor([0.0714, 0.0485, 0.0402, 0.0214, 0.1591, 0.0308, 0.0356, 0.0378], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0131, 0.0189, 0.0158, 0.0227, 0.0170, 0.0139, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 21:44:58,115 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:06,109 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:22,101 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5807, 2.5181, 2.3922, 4.0330, 1.8740, 3.7502, 2.1907, 2.4805], device='cuda:5'), covar=tensor([0.0314, 0.0612, 0.0427, 0.0166, 0.1600, 0.0212, 0.0877, 0.1062], device='cuda:5'), in_proj_covar=tensor([0.0239, 0.0214, 0.0181, 0.0240, 0.0283, 0.0186, 0.0203, 0.0275], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:45:25,706 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 21:45:34,090 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3765, 4.2285, 4.2560, 4.2813, 4.1749, 4.8224, 4.5339, 4.2599], device='cuda:5'), covar=tensor([0.1121, 0.1317, 0.1285, 0.1626, 0.2860, 0.0988, 0.0978, 0.1942], device='cuda:5'), in_proj_covar=tensor([0.0227, 0.0333, 0.0304, 0.0284, 0.0373, 0.0326, 0.0260, 0.0382], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:45:35,321 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:36,282 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:38,312 INFO [train.py:904] (5/8) Epoch 3, batch 2000, loss[loss=0.2621, simple_loss=0.3213, pruned_loss=0.1014, over 16298.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3153, pruned_loss=0.08745, over 3324622.05 frames. ], batch size: 165, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:46:01,055 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:07,217 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-27 21:46:13,083 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:17,541 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.887e+02 3.938e+02 4.847e+02 6.098e+02 1.059e+03, threshold=9.693e+02, percent-clipped=1.0 2023-04-27 21:46:46,970 INFO [train.py:904] (5/8) Epoch 3, batch 2050, loss[loss=0.2615, simple_loss=0.3315, pruned_loss=0.09574, over 17133.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3147, pruned_loss=0.08742, over 3332657.86 frames. ], batch size: 47, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:47:33,616 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:47:53,692 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3667, 4.3446, 1.8715, 4.4285, 2.4686, 4.3461, 2.4547, 3.1958], device='cuda:5'), covar=tensor([0.0039, 0.0170, 0.1325, 0.0026, 0.0761, 0.0269, 0.1114, 0.0450], device='cuda:5'), in_proj_covar=tensor([0.0088, 0.0140, 0.0172, 0.0084, 0.0162, 0.0170, 0.0184, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 21:47:55,916 INFO [train.py:904] (5/8) Epoch 3, batch 2100, loss[loss=0.2831, simple_loss=0.3355, pruned_loss=0.1153, over 16750.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3159, pruned_loss=0.08799, over 3333621.71 frames. ], batch size: 83, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:48:20,723 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:48:21,178 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 21:48:27,470 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4573, 5.8155, 5.5196, 5.6018, 4.9390, 4.8349, 5.2651, 5.9703], device='cuda:5'), covar=tensor([0.0496, 0.0587, 0.0789, 0.0377, 0.0578, 0.0483, 0.0490, 0.0540], device='cuda:5'), in_proj_covar=tensor([0.0269, 0.0384, 0.0335, 0.0232, 0.0249, 0.0231, 0.0301, 0.0263], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:48:31,789 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.621e+02 3.997e+02 4.789e+02 6.119e+02 1.821e+03, threshold=9.579e+02, percent-clipped=3.0 2023-04-27 21:48:48,012 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1257, 5.0272, 4.7907, 4.8971, 4.4244, 4.9434, 4.9279, 4.5928], device='cuda:5'), covar=tensor([0.0270, 0.0128, 0.0171, 0.0128, 0.0712, 0.0152, 0.0179, 0.0270], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0126, 0.0186, 0.0154, 0.0219, 0.0164, 0.0134, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 21:48:55,004 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:00,607 INFO [train.py:904] (5/8) Epoch 3, batch 2150, loss[loss=0.2586, simple_loss=0.3164, pruned_loss=0.1004, over 16832.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3172, pruned_loss=0.08834, over 3336601.22 frames. ], batch size: 102, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:49:22,531 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:47,669 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:51,644 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-27 21:50:08,037 INFO [train.py:904] (5/8) Epoch 3, batch 2200, loss[loss=0.2094, simple_loss=0.2868, pruned_loss=0.06596, over 16749.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3168, pruned_loss=0.08793, over 3341770.84 frames. ], batch size: 39, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:50:14,157 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:46,034 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.515e+02 4.185e+02 4.921e+02 5.938e+02 1.054e+03, threshold=9.842e+02, percent-clipped=2.0 2023-04-27 21:50:48,914 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22532.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:15,031 INFO [train.py:904] (5/8) Epoch 3, batch 2250, loss[loss=0.2331, simple_loss=0.3056, pruned_loss=0.08034, over 17168.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3175, pruned_loss=0.08819, over 3337948.12 frames. ], batch size: 43, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:51:18,574 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:42,423 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:09,989 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:20,029 INFO [train.py:904] (5/8) Epoch 3, batch 2300, loss[loss=0.2361, simple_loss=0.2963, pruned_loss=0.08796, over 16815.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3167, pruned_loss=0.08779, over 3338906.81 frames. ], batch size: 96, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:52:28,601 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3680, 4.0870, 3.4800, 1.8510, 2.7531, 2.3773, 3.6983, 4.0858], device='cuda:5'), covar=tensor([0.0244, 0.0386, 0.0446, 0.1519, 0.0730, 0.0940, 0.0531, 0.0390], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0126, 0.0158, 0.0147, 0.0137, 0.0132, 0.0149, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 21:52:57,430 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:53:01,639 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.100e+02 3.931e+02 4.776e+02 5.724e+02 1.077e+03, threshold=9.553e+02, percent-clipped=1.0 2023-04-27 21:53:12,051 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:53:30,361 INFO [train.py:904] (5/8) Epoch 3, batch 2350, loss[loss=0.2212, simple_loss=0.3035, pruned_loss=0.06947, over 17106.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.318, pruned_loss=0.08966, over 3338510.06 frames. ], batch size: 47, lr: 2.16e-02, grad_scale: 4.0 2023-04-27 21:53:38,512 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 21:54:01,270 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:54:35,095 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 21:54:35,738 INFO [train.py:904] (5/8) Epoch 3, batch 2400, loss[loss=0.2973, simple_loss=0.3537, pruned_loss=0.1204, over 16750.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3191, pruned_loss=0.08956, over 3343435.84 frames. ], batch size: 134, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:55:17,395 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 3.861e+02 4.414e+02 5.506e+02 1.152e+03, threshold=8.829e+02, percent-clipped=5.0 2023-04-27 21:55:32,352 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:55:46,922 INFO [train.py:904] (5/8) Epoch 3, batch 2450, loss[loss=0.2411, simple_loss=0.3055, pruned_loss=0.08839, over 16470.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3201, pruned_loss=0.0892, over 3343117.57 frames. ], batch size: 75, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:56:33,454 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:46,240 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:55,078 INFO [train.py:904] (5/8) Epoch 3, batch 2500, loss[loss=0.2636, simple_loss=0.3116, pruned_loss=0.1078, over 16830.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3205, pruned_loss=0.09022, over 3338111.55 frames. ], batch size: 109, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:57:27,730 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:33,680 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.850e+02 4.203e+02 4.842e+02 6.402e+02 1.699e+03, threshold=9.683e+02, percent-clipped=7.0 2023-04-27 21:57:37,639 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:41,132 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:03,439 INFO [train.py:904] (5/8) Epoch 3, batch 2550, loss[loss=0.2565, simple_loss=0.3334, pruned_loss=0.08981, over 16723.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3193, pruned_loss=0.08976, over 3338487.11 frames. ], batch size: 57, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:58:11,018 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:31,426 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:44,550 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:52,515 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:01,458 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:06,823 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9121, 4.5003, 4.7884, 5.1896, 5.3251, 4.5803, 5.3398, 5.2150], device='cuda:5'), covar=tensor([0.0673, 0.0808, 0.1413, 0.0411, 0.0333, 0.0490, 0.0293, 0.0359], device='cuda:5'), in_proj_covar=tensor([0.0301, 0.0365, 0.0490, 0.0372, 0.0273, 0.0263, 0.0284, 0.0295], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 21:59:13,848 INFO [train.py:904] (5/8) Epoch 3, batch 2600, loss[loss=0.2433, simple_loss=0.3258, pruned_loss=0.08041, over 17064.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3198, pruned_loss=0.0899, over 3326708.64 frames. ], batch size: 53, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:59:39,352 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:53,534 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.483e+02 3.680e+02 4.418e+02 5.171e+02 1.031e+03, threshold=8.837e+02, percent-clipped=1.0 2023-04-27 22:00:09,151 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:00:22,556 INFO [train.py:904] (5/8) Epoch 3, batch 2650, loss[loss=0.2584, simple_loss=0.3174, pruned_loss=0.09975, over 16710.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3194, pruned_loss=0.08902, over 3326148.92 frames. ], batch size: 134, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:00:39,999 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1295, 2.5825, 2.4112, 4.4511, 1.6994, 3.9012, 2.2615, 2.2258], device='cuda:5'), covar=tensor([0.0340, 0.0859, 0.0510, 0.0221, 0.2345, 0.0279, 0.1088, 0.1807], device='cuda:5'), in_proj_covar=tensor([0.0246, 0.0219, 0.0185, 0.0249, 0.0289, 0.0194, 0.0210, 0.0282], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 22:01:22,677 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:01:30,285 INFO [train.py:904] (5/8) Epoch 3, batch 2700, loss[loss=0.2356, simple_loss=0.3244, pruned_loss=0.07336, over 17027.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3202, pruned_loss=0.08877, over 3325316.04 frames. ], batch size: 50, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:02:09,772 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.756e+02 4.226e+02 4.976e+02 6.032e+02 3.495e+03, threshold=9.952e+02, percent-clipped=7.0 2023-04-27 22:02:24,083 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:02:32,360 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 22:02:37,956 INFO [train.py:904] (5/8) Epoch 3, batch 2750, loss[loss=0.2282, simple_loss=0.2989, pruned_loss=0.07872, over 16767.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3189, pruned_loss=0.08757, over 3326476.30 frames. ], batch size: 39, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:03:16,611 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 22:03:23,799 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5278, 4.5410, 4.1215, 1.7840, 2.8121, 2.4921, 3.7924, 4.3008], device='cuda:5'), covar=tensor([0.0269, 0.0372, 0.0386, 0.1588, 0.0775, 0.0936, 0.0719, 0.0889], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0122, 0.0156, 0.0146, 0.0137, 0.0132, 0.0148, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 22:03:28,907 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:03:45,517 INFO [train.py:904] (5/8) Epoch 3, batch 2800, loss[loss=0.2665, simple_loss=0.3245, pruned_loss=0.1042, over 16749.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3178, pruned_loss=0.08573, over 3331085.12 frames. ], batch size: 76, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:03:51,486 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7179, 3.6595, 1.8199, 3.6605, 2.5107, 3.7254, 1.8061, 2.8649], device='cuda:5'), covar=tensor([0.0052, 0.0188, 0.1363, 0.0064, 0.0637, 0.0305, 0.1154, 0.0467], device='cuda:5'), in_proj_covar=tensor([0.0091, 0.0138, 0.0174, 0.0086, 0.0162, 0.0171, 0.0183, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 22:03:55,558 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:14,239 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3737, 3.9927, 3.8275, 1.3538, 4.0234, 3.8622, 3.1421, 3.0492], device='cuda:5'), covar=tensor([0.0904, 0.0072, 0.0154, 0.1490, 0.0085, 0.0077, 0.0355, 0.0414], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0144, 0.0077, 0.0076, 0.0116, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 22:04:25,547 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.403e+02 4.548e+02 5.612e+02 1.011e+03, threshold=9.095e+02, percent-clipped=1.0 2023-04-27 22:04:54,746 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:55,609 INFO [train.py:904] (5/8) Epoch 3, batch 2850, loss[loss=0.2433, simple_loss=0.3034, pruned_loss=0.09161, over 16426.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3171, pruned_loss=0.08569, over 3334620.74 frames. ], batch size: 146, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:05:20,434 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:33,239 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-27 22:05:36,720 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23181.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:03,310 INFO [train.py:904] (5/8) Epoch 3, batch 2900, loss[loss=0.2125, simple_loss=0.3034, pruned_loss=0.0608, over 17031.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3157, pruned_loss=0.08564, over 3340492.52 frames. ], batch size: 50, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:06:25,776 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:42,416 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:43,091 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.672e+02 4.009e+02 4.959e+02 5.927e+02 1.397e+03, threshold=9.918e+02, percent-clipped=5.0 2023-04-27 22:07:12,202 INFO [train.py:904] (5/8) Epoch 3, batch 2950, loss[loss=0.2513, simple_loss=0.3081, pruned_loss=0.09724, over 16732.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3156, pruned_loss=0.08685, over 3331070.71 frames. ], batch size: 134, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:07:49,221 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:07,124 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:12,739 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23295.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:19,858 INFO [train.py:904] (5/8) Epoch 3, batch 3000, loss[loss=0.2556, simple_loss=0.3157, pruned_loss=0.09781, over 16829.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3169, pruned_loss=0.088, over 3333860.11 frames. ], batch size: 96, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:08:19,858 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 22:08:30,493 INFO [train.py:938] (5/8) Epoch 3, validation: loss=0.1721, simple_loss=0.279, pruned_loss=0.03262, over 944034.00 frames. 2023-04-27 22:08:30,493 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-27 22:08:43,296 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6342, 2.5711, 2.1833, 2.4032, 3.0597, 3.0005, 4.0746, 3.4535], device='cuda:5'), covar=tensor([0.0021, 0.0114, 0.0128, 0.0132, 0.0066, 0.0106, 0.0038, 0.0062], device='cuda:5'), in_proj_covar=tensor([0.0060, 0.0116, 0.0113, 0.0114, 0.0107, 0.0116, 0.0077, 0.0093], device='cuda:5'), out_proj_covar=tensor([8.8132e-05, 1.7043e-04, 1.6072e-04, 1.6668e-04, 1.6094e-04, 1.7179e-04, 1.1482e-04, 1.4189e-04], device='cuda:5') 2023-04-27 22:09:10,137 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.558e+02 3.966e+02 4.777e+02 5.754e+02 1.756e+03, threshold=9.554e+02, percent-clipped=1.0 2023-04-27 22:09:25,506 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:37,540 INFO [train.py:904] (5/8) Epoch 3, batch 3050, loss[loss=0.2557, simple_loss=0.3153, pruned_loss=0.09808, over 16828.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3159, pruned_loss=0.088, over 3331377.31 frames. ], batch size: 96, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:09:40,462 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4067, 4.3276, 3.8798, 1.6991, 2.9695, 2.3360, 3.7312, 4.0194], device='cuda:5'), covar=tensor([0.0245, 0.0436, 0.0393, 0.1627, 0.0670, 0.0946, 0.0579, 0.0716], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0123, 0.0155, 0.0145, 0.0136, 0.0131, 0.0147, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 22:10:35,335 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2702, 5.1928, 5.0242, 4.3353, 4.9539, 2.3560, 4.6884, 5.0700], device='cuda:5'), covar=tensor([0.0049, 0.0049, 0.0061, 0.0302, 0.0045, 0.1037, 0.0069, 0.0090], device='cuda:5'), in_proj_covar=tensor([0.0077, 0.0066, 0.0102, 0.0119, 0.0074, 0.0114, 0.0093, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 22:10:44,553 INFO [train.py:904] (5/8) Epoch 3, batch 3100, loss[loss=0.2553, simple_loss=0.3116, pruned_loss=0.09951, over 16913.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3153, pruned_loss=0.08727, over 3333366.95 frames. ], batch size: 109, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:11:28,259 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.500e+02 3.880e+02 4.394e+02 5.487e+02 1.419e+03, threshold=8.788e+02, percent-clipped=5.0 2023-04-27 22:11:53,309 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:11:54,091 INFO [train.py:904] (5/8) Epoch 3, batch 3150, loss[loss=0.2592, simple_loss=0.3295, pruned_loss=0.09441, over 16790.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3146, pruned_loss=0.0862, over 3343481.14 frames. ], batch size: 83, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:12:12,248 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:35,172 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:41,797 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6672, 6.0405, 5.6683, 5.8173, 5.2623, 5.0456, 5.5502, 6.1029], device='cuda:5'), covar=tensor([0.0516, 0.0593, 0.0856, 0.0338, 0.0522, 0.0423, 0.0419, 0.0592], device='cuda:5'), in_proj_covar=tensor([0.0278, 0.0400, 0.0339, 0.0239, 0.0254, 0.0243, 0.0304, 0.0269], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 22:12:44,110 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:58,215 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23498.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:01,361 INFO [train.py:904] (5/8) Epoch 3, batch 3200, loss[loss=0.1972, simple_loss=0.2753, pruned_loss=0.05952, over 16735.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3138, pruned_loss=0.08641, over 3342936.74 frames. ], batch size: 39, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:13:39,252 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:42,226 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.397e+02 3.601e+02 4.370e+02 5.311e+02 9.274e+02, threshold=8.739e+02, percent-clipped=1.0 2023-04-27 22:13:53,251 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0434, 3.2373, 3.7191, 2.8197, 3.7034, 3.7715, 3.6984, 2.0440], device='cuda:5'), covar=tensor([0.0368, 0.0158, 0.0044, 0.0195, 0.0032, 0.0042, 0.0040, 0.0304], device='cuda:5'), in_proj_covar=tensor([0.0109, 0.0052, 0.0058, 0.0105, 0.0052, 0.0060, 0.0059, 0.0101], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 22:14:06,491 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:14:08,952 INFO [train.py:904] (5/8) Epoch 3, batch 3250, loss[loss=0.204, simple_loss=0.2826, pruned_loss=0.06266, over 17191.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3136, pruned_loss=0.0857, over 3345994.02 frames. ], batch size: 46, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:14:38,904 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:56,674 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:18,375 INFO [train.py:904] (5/8) Epoch 3, batch 3300, loss[loss=0.2119, simple_loss=0.2897, pruned_loss=0.06707, over 16828.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.314, pruned_loss=0.08578, over 3341389.48 frames. ], batch size: 42, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:15:53,510 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 22:15:57,198 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.549e+02 3.983e+02 4.910e+02 5.801e+02 1.123e+03, threshold=9.819e+02, percent-clipped=5.0 2023-04-27 22:16:24,700 INFO [train.py:904] (5/8) Epoch 3, batch 3350, loss[loss=0.2573, simple_loss=0.3221, pruned_loss=0.09631, over 16397.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3154, pruned_loss=0.0867, over 3327776.48 frames. ], batch size: 146, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:16:53,557 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2204, 5.0178, 4.9318, 4.9165, 4.3945, 5.0031, 4.9313, 4.5463], device='cuda:5'), covar=tensor([0.0319, 0.0191, 0.0183, 0.0137, 0.0909, 0.0206, 0.0189, 0.0372], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0136, 0.0195, 0.0162, 0.0230, 0.0175, 0.0140, 0.0188], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 22:17:06,755 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 22:17:33,469 INFO [train.py:904] (5/8) Epoch 3, batch 3400, loss[loss=0.2459, simple_loss=0.3143, pruned_loss=0.08872, over 16505.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3157, pruned_loss=0.08638, over 3323833.59 frames. ], batch size: 75, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:18:13,372 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 3.881e+02 4.662e+02 5.457e+02 1.013e+03, threshold=9.324e+02, percent-clipped=1.0 2023-04-27 22:18:33,287 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7567, 3.5188, 3.6552, 3.6501, 3.6084, 4.0938, 3.9703, 3.5764], device='cuda:5'), covar=tensor([0.1838, 0.1754, 0.1167, 0.1925, 0.2439, 0.1263, 0.0913, 0.2232], device='cuda:5'), in_proj_covar=tensor([0.0234, 0.0332, 0.0307, 0.0295, 0.0380, 0.0331, 0.0263, 0.0383], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 22:18:40,217 INFO [train.py:904] (5/8) Epoch 3, batch 3450, loss[loss=0.2811, simple_loss=0.3301, pruned_loss=0.116, over 16733.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3134, pruned_loss=0.08563, over 3325760.27 frames. ], batch size: 134, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:18:59,496 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:32,619 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5661, 5.9181, 5.4930, 5.7568, 5.2401, 4.9111, 5.3580, 5.9768], device='cuda:5'), covar=tensor([0.0504, 0.0621, 0.0792, 0.0322, 0.0458, 0.0443, 0.0497, 0.0560], device='cuda:5'), in_proj_covar=tensor([0.0278, 0.0404, 0.0344, 0.0244, 0.0258, 0.0245, 0.0313, 0.0274], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 22:19:47,201 INFO [train.py:904] (5/8) Epoch 3, batch 3500, loss[loss=0.2151, simple_loss=0.2952, pruned_loss=0.06754, over 17222.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3113, pruned_loss=0.08432, over 3321635.07 frames. ], batch size: 46, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:20:04,595 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23812.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:31,311 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.165e+02 3.876e+02 4.693e+02 5.719e+02 1.171e+03, threshold=9.385e+02, percent-clipped=5.0 2023-04-27 22:20:48,801 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:20:59,089 INFO [train.py:904] (5/8) Epoch 3, batch 3550, loss[loss=0.2058, simple_loss=0.2841, pruned_loss=0.06379, over 16858.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3105, pruned_loss=0.08364, over 3318778.90 frames. ], batch size: 42, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:21:29,311 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:39,719 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 2023-04-27 22:21:45,203 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:05,758 INFO [train.py:904] (5/8) Epoch 3, batch 3600, loss[loss=0.2044, simple_loss=0.2829, pruned_loss=0.06293, over 17231.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3082, pruned_loss=0.08319, over 3317191.62 frames. ], batch size: 43, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:22:33,415 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:47,023 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.441e+02 3.888e+02 4.838e+02 6.309e+02 1.139e+03, threshold=9.677e+02, percent-clipped=4.0 2023-04-27 22:22:50,343 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:14,451 INFO [train.py:904] (5/8) Epoch 3, batch 3650, loss[loss=0.2312, simple_loss=0.2845, pruned_loss=0.089, over 16522.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3071, pruned_loss=0.08363, over 3302952.19 frames. ], batch size: 75, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:24:29,897 INFO [train.py:904] (5/8) Epoch 3, batch 3700, loss[loss=0.2747, simple_loss=0.3235, pruned_loss=0.1129, over 16700.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3059, pruned_loss=0.08552, over 3290287.54 frames. ], batch size: 89, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:25:01,944 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2023-04-27 22:25:13,799 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.162e+02 3.909e+02 4.503e+02 5.530e+02 1.035e+03, threshold=9.006e+02, percent-clipped=1.0 2023-04-27 22:25:22,623 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:22,717 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0987, 2.1311, 1.5026, 1.8184, 2.7321, 2.6584, 2.9637, 2.8466], device='cuda:5'), covar=tensor([0.0042, 0.0137, 0.0195, 0.0172, 0.0069, 0.0102, 0.0050, 0.0070], device='cuda:5'), in_proj_covar=tensor([0.0061, 0.0117, 0.0115, 0.0115, 0.0110, 0.0118, 0.0078, 0.0095], device='cuda:5'), out_proj_covar=tensor([9.0090e-05, 1.6951e-04, 1.6217e-04, 1.6699e-04, 1.6310e-04, 1.7414e-04, 1.1629e-04, 1.4496e-04], device='cuda:5') 2023-04-27 22:25:42,971 INFO [train.py:904] (5/8) Epoch 3, batch 3750, loss[loss=0.2436, simple_loss=0.2977, pruned_loss=0.0947, over 16879.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3077, pruned_loss=0.08797, over 3265780.82 frames. ], batch size: 96, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:25:54,819 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6285, 4.3227, 4.5724, 1.6740, 4.7423, 4.8295, 3.5471, 3.3012], device='cuda:5'), covar=tensor([0.0835, 0.0091, 0.0100, 0.1517, 0.0063, 0.0036, 0.0210, 0.0422], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0082, 0.0080, 0.0142, 0.0077, 0.0073, 0.0112, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 22:26:46,926 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:26:50,664 INFO [train.py:904] (5/8) Epoch 3, batch 3800, loss[loss=0.204, simple_loss=0.2758, pruned_loss=0.06612, over 16792.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3085, pruned_loss=0.08922, over 3270903.49 frames. ], batch size: 102, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:27:34,025 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.430e+02 3.668e+02 4.510e+02 5.659e+02 1.360e+03, threshold=9.019e+02, percent-clipped=5.0 2023-04-27 22:27:52,794 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:28:01,900 INFO [train.py:904] (5/8) Epoch 3, batch 3850, loss[loss=0.227, simple_loss=0.3048, pruned_loss=0.07461, over 17033.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3084, pruned_loss=0.08995, over 3273416.80 frames. ], batch size: 50, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:28:15,853 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 22:28:19,288 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0380, 4.2758, 4.5946, 4.6298, 4.5655, 4.2705, 3.8099, 4.1117], device='cuda:5'), covar=tensor([0.0475, 0.0536, 0.0462, 0.0485, 0.0641, 0.0414, 0.1392, 0.0493], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0185, 0.0197, 0.0199, 0.0241, 0.0202, 0.0299, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 22:28:21,038 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4436, 4.4408, 4.4563, 4.6790, 4.5474, 5.0842, 4.7615, 4.6089], device='cuda:5'), covar=tensor([0.1134, 0.1357, 0.1135, 0.1663, 0.2499, 0.0920, 0.1045, 0.1937], device='cuda:5'), in_proj_covar=tensor([0.0227, 0.0321, 0.0296, 0.0281, 0.0356, 0.0314, 0.0258, 0.0372], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 22:29:01,359 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=24192.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:13,631 INFO [train.py:904] (5/8) Epoch 3, batch 3900, loss[loss=0.2117, simple_loss=0.2916, pruned_loss=0.06595, over 16247.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3078, pruned_loss=0.08987, over 3273570.86 frames. ], batch size: 35, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:29:56,953 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.402e+02 3.854e+02 4.579e+02 5.596e+02 1.788e+03, threshold=9.157e+02, percent-clipped=5.0 2023-04-27 22:30:24,582 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0390, 3.4482, 3.7130, 2.4285, 3.4221, 3.6873, 3.5465, 2.1323], device='cuda:5'), covar=tensor([0.0344, 0.0079, 0.0029, 0.0246, 0.0044, 0.0048, 0.0035, 0.0271], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0054, 0.0058, 0.0110, 0.0054, 0.0061, 0.0059, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 22:30:25,207 INFO [train.py:904] (5/8) Epoch 3, batch 3950, loss[loss=0.2543, simple_loss=0.3075, pruned_loss=0.1006, over 16886.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.307, pruned_loss=0.09004, over 3274107.35 frames. ], batch size: 116, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:31:19,642 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1652, 1.3915, 1.7529, 2.1318, 2.2124, 2.3136, 1.4791, 2.2237], device='cuda:5'), covar=tensor([0.0052, 0.0213, 0.0113, 0.0099, 0.0055, 0.0042, 0.0166, 0.0033], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0123, 0.0111, 0.0103, 0.0086, 0.0070, 0.0114, 0.0065], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 22:31:34,864 INFO [train.py:904] (5/8) Epoch 3, batch 4000, loss[loss=0.2138, simple_loss=0.2906, pruned_loss=0.06847, over 17007.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3065, pruned_loss=0.08963, over 3271926.21 frames. ], batch size: 55, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:32:06,614 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5299, 5.4464, 5.3658, 5.3574, 4.8220, 5.3699, 5.1731, 5.1395], device='cuda:5'), covar=tensor([0.0238, 0.0119, 0.0115, 0.0083, 0.0728, 0.0117, 0.0132, 0.0201], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0127, 0.0177, 0.0148, 0.0209, 0.0161, 0.0128, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 22:32:17,084 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.388e+02 3.319e+02 4.445e+02 5.556e+02 1.324e+03, threshold=8.889e+02, percent-clipped=2.0 2023-04-27 22:32:45,223 INFO [train.py:904] (5/8) Epoch 3, batch 4050, loss[loss=0.2307, simple_loss=0.3079, pruned_loss=0.07674, over 16498.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3053, pruned_loss=0.08709, over 3278920.61 frames. ], batch size: 146, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:33:19,396 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 22:33:46,333 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:33:58,267 INFO [train.py:904] (5/8) Epoch 3, batch 4100, loss[loss=0.2314, simple_loss=0.301, pruned_loss=0.08087, over 16262.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3047, pruned_loss=0.08455, over 3280168.81 frames. ], batch size: 35, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:34:04,849 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0460, 3.3521, 3.4193, 1.5060, 3.6523, 3.6462, 2.9403, 2.7512], device='cuda:5'), covar=tensor([0.0771, 0.0116, 0.0143, 0.1277, 0.0056, 0.0043, 0.0273, 0.0342], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0082, 0.0080, 0.0147, 0.0075, 0.0073, 0.0113, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 22:34:42,996 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.218e+02 2.972e+02 3.823e+02 4.924e+02 9.607e+02, threshold=7.646e+02, percent-clipped=2.0 2023-04-27 22:35:13,096 INFO [train.py:904] (5/8) Epoch 3, batch 4150, loss[loss=0.36, simple_loss=0.3947, pruned_loss=0.1626, over 11406.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3141, pruned_loss=0.0898, over 3224216.23 frames. ], batch size: 246, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:35:37,539 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1638, 2.9265, 2.4997, 4.5480, 1.9173, 4.2042, 2.4876, 2.6215], device='cuda:5'), covar=tensor([0.0264, 0.0668, 0.0457, 0.0162, 0.1885, 0.0222, 0.0864, 0.1390], device='cuda:5'), in_proj_covar=tensor([0.0253, 0.0227, 0.0191, 0.0252, 0.0298, 0.0202, 0.0217, 0.0292], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 22:36:09,464 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-04-27 22:36:22,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7728, 3.6834, 3.6875, 3.7234, 3.6776, 4.0887, 3.9798, 3.7947], device='cuda:5'), covar=tensor([0.1654, 0.1242, 0.1082, 0.1772, 0.2225, 0.1171, 0.0837, 0.2055], device='cuda:5'), in_proj_covar=tensor([0.0227, 0.0307, 0.0284, 0.0272, 0.0350, 0.0304, 0.0250, 0.0365], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 22:36:27,622 INFO [train.py:904] (5/8) Epoch 3, batch 4200, loss[loss=0.2765, simple_loss=0.3628, pruned_loss=0.09507, over 16908.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.3231, pruned_loss=0.09326, over 3191251.40 frames. ], batch size: 102, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:37:10,951 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.707e+02 3.802e+02 4.415e+02 5.397e+02 1.092e+03, threshold=8.829e+02, percent-clipped=9.0 2023-04-27 22:37:35,688 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:37:35,803 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7217, 3.3172, 2.6770, 4.7174, 4.5250, 4.2643, 2.0222, 3.3688], device='cuda:5'), covar=tensor([0.1389, 0.0475, 0.1082, 0.0053, 0.0167, 0.0307, 0.1136, 0.0580], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0129, 0.0160, 0.0071, 0.0130, 0.0132, 0.0150, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 22:37:40,127 INFO [train.py:904] (5/8) Epoch 3, batch 4250, loss[loss=0.2527, simple_loss=0.3312, pruned_loss=0.08707, over 16853.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3248, pruned_loss=0.09204, over 3199210.75 frames. ], batch size: 116, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:38:06,887 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:06,957 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2085, 2.9841, 2.7546, 1.9429, 2.5200, 2.0787, 2.8110, 3.0233], device='cuda:5'), covar=tensor([0.0326, 0.0614, 0.0483, 0.1533, 0.0764, 0.0885, 0.0727, 0.0386], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0117, 0.0156, 0.0148, 0.0139, 0.0132, 0.0147, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 22:38:39,979 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 2023-04-27 22:38:53,685 INFO [train.py:904] (5/8) Epoch 3, batch 4300, loss[loss=0.2969, simple_loss=0.3608, pruned_loss=0.1165, over 15496.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3262, pruned_loss=0.09126, over 3183954.64 frames. ], batch size: 190, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:39:05,053 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:30,415 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6189, 1.1791, 1.4292, 1.6500, 1.6413, 1.7961, 1.4152, 1.8001], device='cuda:5'), covar=tensor([0.0051, 0.0149, 0.0073, 0.0089, 0.0056, 0.0040, 0.0120, 0.0038], device='cuda:5'), in_proj_covar=tensor([0.0085, 0.0121, 0.0107, 0.0100, 0.0086, 0.0065, 0.0111, 0.0063], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 22:39:37,549 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:38,218 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.269e+02 3.576e+02 4.392e+02 5.357e+02 9.860e+02, threshold=8.785e+02, percent-clipped=3.0 2023-04-27 22:40:06,101 INFO [train.py:904] (5/8) Epoch 3, batch 4350, loss[loss=0.2447, simple_loss=0.32, pruned_loss=0.08468, over 17054.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3308, pruned_loss=0.09339, over 3172163.25 frames. ], batch size: 50, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:40:44,470 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 22:40:55,297 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8688, 3.9046, 4.3395, 4.2903, 4.2522, 3.8064, 3.9503, 3.8819], device='cuda:5'), covar=tensor([0.0225, 0.0210, 0.0202, 0.0302, 0.0334, 0.0265, 0.0621, 0.0320], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0171, 0.0184, 0.0180, 0.0224, 0.0185, 0.0283, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-27 22:41:03,951 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0686, 1.2023, 1.5004, 1.9945, 2.1396, 2.1911, 1.3380, 2.2542], device='cuda:5'), covar=tensor([0.0052, 0.0180, 0.0106, 0.0114, 0.0053, 0.0042, 0.0148, 0.0037], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0122, 0.0109, 0.0101, 0.0087, 0.0066, 0.0114, 0.0063], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 22:41:11,417 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24693.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:41:21,564 INFO [train.py:904] (5/8) Epoch 3, batch 4400, loss[loss=0.2725, simple_loss=0.3374, pruned_loss=0.1039, over 16615.00 frames. ], tot_loss[loss=0.2604, simple_loss=0.3324, pruned_loss=0.0942, over 3177431.52 frames. ], batch size: 57, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:42:05,391 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.377e+02 3.698e+02 4.447e+02 5.386e+02 9.920e+02, threshold=8.895e+02, percent-clipped=4.0 2023-04-27 22:42:21,153 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:42:34,828 INFO [train.py:904] (5/8) Epoch 3, batch 4450, loss[loss=0.2687, simple_loss=0.3512, pruned_loss=0.09309, over 16261.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3345, pruned_loss=0.09369, over 3184891.17 frames. ], batch size: 165, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:42:42,276 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9215, 4.5622, 4.2476, 1.5941, 3.2841, 2.6192, 4.0797, 4.3198], device='cuda:5'), covar=tensor([0.0163, 0.0352, 0.0351, 0.1916, 0.0664, 0.0921, 0.0540, 0.0404], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0117, 0.0156, 0.0146, 0.0139, 0.0131, 0.0144, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-27 22:42:55,238 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:43:25,326 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3151, 3.3754, 1.4829, 3.4955, 2.2953, 3.4618, 1.8173, 2.5427], device='cuda:5'), covar=tensor([0.0059, 0.0172, 0.1622, 0.0034, 0.0698, 0.0258, 0.1248, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0088, 0.0127, 0.0171, 0.0076, 0.0159, 0.0157, 0.0179, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 22:43:47,154 INFO [train.py:904] (5/8) Epoch 3, batch 4500, loss[loss=0.2861, simple_loss=0.3388, pruned_loss=0.1167, over 11803.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3338, pruned_loss=0.0927, over 3194139.82 frames. ], batch size: 248, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:44:22,365 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:25,390 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9734, 2.4059, 2.3576, 3.2238, 3.0218, 3.0918, 1.8227, 2.6997], device='cuda:5'), covar=tensor([0.1182, 0.0456, 0.1030, 0.0080, 0.0194, 0.0316, 0.1126, 0.0635], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0130, 0.0163, 0.0070, 0.0130, 0.0134, 0.0154, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 22:44:29,088 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 3.148e+02 3.848e+02 4.477e+02 7.288e+02, threshold=7.696e+02, percent-clipped=0.0 2023-04-27 22:44:52,898 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:56,547 INFO [train.py:904] (5/8) Epoch 3, batch 4550, loss[loss=0.2582, simple_loss=0.3346, pruned_loss=0.09091, over 16534.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.3336, pruned_loss=0.09265, over 3209467.46 frames. ], batch size: 75, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:46:07,318 INFO [train.py:904] (5/8) Epoch 3, batch 4600, loss[loss=0.2553, simple_loss=0.3356, pruned_loss=0.0875, over 16644.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.3343, pruned_loss=0.09236, over 3210804.99 frames. ], batch size: 68, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:46:10,662 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:21,308 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:29,955 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 22:46:37,311 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6611, 3.7897, 2.9229, 2.4305, 3.1182, 2.3080, 4.0879, 4.3729], device='cuda:5'), covar=tensor([0.2138, 0.0662, 0.1178, 0.1108, 0.1914, 0.1237, 0.0355, 0.0244], device='cuda:5'), in_proj_covar=tensor([0.0269, 0.0239, 0.0252, 0.0209, 0.0303, 0.0191, 0.0217, 0.0200], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 22:46:44,269 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:52,478 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 3.022e+02 3.683e+02 4.480e+02 7.885e+02, threshold=7.365e+02, percent-clipped=1.0 2023-04-27 22:47:20,347 INFO [train.py:904] (5/8) Epoch 3, batch 4650, loss[loss=0.2211, simple_loss=0.2983, pruned_loss=0.07196, over 16589.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3326, pruned_loss=0.0918, over 3225707.24 frames. ], batch size: 62, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:47:24,901 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3394, 1.9173, 2.1041, 3.1361, 3.1199, 3.4578, 1.6866, 3.4348], device='cuda:5'), covar=tensor([0.0033, 0.0164, 0.0126, 0.0059, 0.0040, 0.0037, 0.0162, 0.0021], device='cuda:5'), in_proj_covar=tensor([0.0083, 0.0120, 0.0108, 0.0100, 0.0087, 0.0067, 0.0113, 0.0063], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-27 22:48:24,036 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:48:32,757 INFO [train.py:904] (5/8) Epoch 3, batch 4700, loss[loss=0.2221, simple_loss=0.3016, pruned_loss=0.07133, over 16535.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3299, pruned_loss=0.09036, over 3230790.58 frames. ], batch size: 68, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:49:17,307 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.399e+02 3.338e+02 4.097e+02 4.784e+02 1.007e+03, threshold=8.194e+02, percent-clipped=3.0 2023-04-27 22:49:34,312 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0326, 2.7557, 2.6781, 1.7457, 2.8876, 2.8835, 2.4211, 2.3189], device='cuda:5'), covar=tensor([0.0678, 0.0113, 0.0172, 0.0938, 0.0089, 0.0075, 0.0321, 0.0362], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0080, 0.0079, 0.0145, 0.0072, 0.0071, 0.0112, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 22:49:45,111 INFO [train.py:904] (5/8) Epoch 3, batch 4750, loss[loss=0.2129, simple_loss=0.2981, pruned_loss=0.06386, over 16492.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.3266, pruned_loss=0.08889, over 3227787.64 frames. ], batch size: 62, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:49:52,737 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 22:50:04,531 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4868, 3.4643, 1.3848, 3.5981, 2.4044, 3.5035, 1.6245, 2.6624], device='cuda:5'), covar=tensor([0.0048, 0.0132, 0.1460, 0.0029, 0.0550, 0.0205, 0.1161, 0.0429], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0127, 0.0170, 0.0078, 0.0159, 0.0157, 0.0177, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 22:50:58,611 INFO [train.py:904] (5/8) Epoch 3, batch 4800, loss[loss=0.2326, simple_loss=0.3186, pruned_loss=0.07328, over 16849.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3225, pruned_loss=0.08621, over 3233301.43 frames. ], batch size: 102, lr: 2.06e-02, grad_scale: 8.0 2023-04-27 22:51:28,802 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:51:47,435 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 2.955e+02 3.547e+02 4.621e+02 1.014e+03, threshold=7.094e+02, percent-clipped=1.0 2023-04-27 22:52:13,137 INFO [train.py:904] (5/8) Epoch 3, batch 4850, loss[loss=0.2526, simple_loss=0.3321, pruned_loss=0.08655, over 16906.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3242, pruned_loss=0.08624, over 3213380.02 frames. ], batch size: 109, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:52:26,624 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7640, 1.4467, 1.8065, 2.6481, 2.5532, 2.8265, 1.5095, 2.5491], device='cuda:5'), covar=tensor([0.0037, 0.0233, 0.0162, 0.0081, 0.0061, 0.0046, 0.0187, 0.0057], device='cuda:5'), in_proj_covar=tensor([0.0085, 0.0122, 0.0107, 0.0101, 0.0087, 0.0066, 0.0112, 0.0063], device='cuda:5'), out_proj_covar=tensor([1.3277e-04, 1.9428e-04, 1.7738e-04, 1.6782e-04, 1.3833e-04, 1.0377e-04, 1.7602e-04, 9.9907e-05], device='cuda:5') 2023-04-27 22:53:25,042 INFO [train.py:904] (5/8) Epoch 3, batch 4900, loss[loss=0.24, simple_loss=0.3211, pruned_loss=0.0795, over 16855.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3236, pruned_loss=0.08485, over 3210856.77 frames. ], batch size: 102, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:53:28,511 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:29,458 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:58,003 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 22:53:59,417 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:54:10,580 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.158e+02 3.416e+02 4.003e+02 5.207e+02 1.200e+03, threshold=8.007e+02, percent-clipped=10.0 2023-04-27 22:54:35,319 INFO [train.py:904] (5/8) Epoch 3, batch 4950, loss[loss=0.3003, simple_loss=0.3646, pruned_loss=0.118, over 12023.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3237, pruned_loss=0.08505, over 3193977.87 frames. ], batch size: 247, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:54:36,193 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:08,079 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:48,337 INFO [train.py:904] (5/8) Epoch 3, batch 5000, loss[loss=0.261, simple_loss=0.3387, pruned_loss=0.09162, over 16871.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3253, pruned_loss=0.08516, over 3202558.84 frames. ], batch size: 116, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:01,029 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9299, 3.3614, 3.5870, 1.4158, 3.7801, 3.7452, 3.0053, 2.6507], device='cuda:5'), covar=tensor([0.0927, 0.0133, 0.0108, 0.1422, 0.0043, 0.0048, 0.0249, 0.0465], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0081, 0.0076, 0.0143, 0.0071, 0.0072, 0.0111, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 22:56:35,250 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.358e+02 3.469e+02 4.106e+02 5.107e+02 1.084e+03, threshold=8.211e+02, percent-clipped=3.0 2023-04-27 22:56:47,436 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-27 22:56:54,979 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-27 22:56:59,654 INFO [train.py:904] (5/8) Epoch 3, batch 5050, loss[loss=0.2237, simple_loss=0.3038, pruned_loss=0.07182, over 16963.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3251, pruned_loss=0.08465, over 3211669.64 frames. ], batch size: 109, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:59,980 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:58:08,635 INFO [train.py:904] (5/8) Epoch 3, batch 5100, loss[loss=0.2312, simple_loss=0.3155, pruned_loss=0.07347, over 15332.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3233, pruned_loss=0.08374, over 3203209.41 frames. ], batch size: 190, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:58:38,791 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:58:48,396 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:58:57,533 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 3.320e+02 3.999e+02 5.208e+02 7.552e+02, threshold=7.999e+02, percent-clipped=0.0 2023-04-27 22:59:16,178 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2756, 3.1957, 1.5544, 3.2740, 2.2766, 3.2710, 1.6786, 2.5555], device='cuda:5'), covar=tensor([0.0062, 0.0195, 0.1509, 0.0045, 0.0746, 0.0410, 0.1467, 0.0616], device='cuda:5'), in_proj_covar=tensor([0.0084, 0.0129, 0.0168, 0.0078, 0.0158, 0.0155, 0.0176, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 22:59:23,214 INFO [train.py:904] (5/8) Epoch 3, batch 5150, loss[loss=0.2396, simple_loss=0.3208, pruned_loss=0.07925, over 17305.00 frames. ], tot_loss[loss=0.245, simple_loss=0.3241, pruned_loss=0.08292, over 3203290.79 frames. ], batch size: 52, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:59:48,253 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 22:59:50,322 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:00:19,399 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:00:35,992 INFO [train.py:904] (5/8) Epoch 3, batch 5200, loss[loss=0.2139, simple_loss=0.2957, pruned_loss=0.06611, over 16795.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3231, pruned_loss=0.08278, over 3195863.01 frames. ], batch size: 83, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:00:40,360 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:01:05,063 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9202, 1.6468, 1.4576, 1.3274, 1.8138, 1.7704, 1.7994, 1.9391], device='cuda:5'), covar=tensor([0.0017, 0.0098, 0.0138, 0.0148, 0.0077, 0.0102, 0.0045, 0.0065], device='cuda:5'), in_proj_covar=tensor([0.0053, 0.0118, 0.0123, 0.0121, 0.0115, 0.0125, 0.0075, 0.0097], device='cuda:5'), out_proj_covar=tensor([7.5587e-05, 1.7005e-04, 1.7208e-04, 1.7427e-04, 1.6841e-04, 1.8255e-04, 1.0702e-04, 1.4671e-04], device='cuda:5') 2023-04-27 23:01:22,143 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.363e+02 3.525e+02 4.137e+02 5.123e+02 8.624e+02, threshold=8.275e+02, percent-clipped=4.0 2023-04-27 23:01:45,792 INFO [train.py:904] (5/8) Epoch 3, batch 5250, loss[loss=0.2215, simple_loss=0.2966, pruned_loss=0.07323, over 16615.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3204, pruned_loss=0.08257, over 3187648.57 frames. ], batch size: 62, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:01:47,881 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:02:56,111 INFO [train.py:904] (5/8) Epoch 3, batch 5300, loss[loss=0.218, simple_loss=0.2929, pruned_loss=0.07159, over 16915.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3165, pruned_loss=0.08122, over 3196450.74 frames. ], batch size: 109, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:03:05,666 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-27 23:03:43,230 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 3.089e+02 3.787e+02 4.480e+02 7.662e+02, threshold=7.574e+02, percent-clipped=0.0 2023-04-27 23:04:08,024 INFO [train.py:904] (5/8) Epoch 3, batch 5350, loss[loss=0.2596, simple_loss=0.3389, pruned_loss=0.09013, over 15415.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3141, pruned_loss=0.08019, over 3207084.90 frames. ], batch size: 190, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:04:08,406 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:05:17,186 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:05:19,291 INFO [train.py:904] (5/8) Epoch 3, batch 5400, loss[loss=0.2267, simple_loss=0.3141, pruned_loss=0.06963, over 16804.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.3176, pruned_loss=0.08161, over 3200006.59 frames. ], batch size: 83, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:06:07,803 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.583e+02 3.733e+02 4.548e+02 5.512e+02 9.876e+02, threshold=9.097e+02, percent-clipped=3.0 2023-04-27 23:06:34,330 INFO [train.py:904] (5/8) Epoch 3, batch 5450, loss[loss=0.2905, simple_loss=0.3627, pruned_loss=0.1092, over 16728.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.322, pruned_loss=0.08461, over 3180307.85 frames. ], batch size: 89, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:07:24,731 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:07:49,291 INFO [train.py:904] (5/8) Epoch 3, batch 5500, loss[loss=0.3152, simple_loss=0.3773, pruned_loss=0.1265, over 17104.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.3322, pruned_loss=0.09302, over 3143895.30 frames. ], batch size: 49, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:08:39,224 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.596e+02 5.216e+02 6.145e+02 8.695e+02 2.860e+03, threshold=1.229e+03, percent-clipped=22.0 2023-04-27 23:08:41,248 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5763, 3.9563, 4.1437, 1.6590, 4.3452, 4.3317, 3.1752, 3.1825], device='cuda:5'), covar=tensor([0.0664, 0.0108, 0.0168, 0.1181, 0.0047, 0.0032, 0.0246, 0.0350], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0081, 0.0079, 0.0145, 0.0075, 0.0073, 0.0113, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 23:09:06,207 INFO [train.py:904] (5/8) Epoch 3, batch 5550, loss[loss=0.4574, simple_loss=0.452, pruned_loss=0.2314, over 10877.00 frames. ], tot_loss[loss=0.272, simple_loss=0.3418, pruned_loss=0.1011, over 3116200.47 frames. ], batch size: 246, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:09:06,778 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3636, 3.3650, 3.1187, 3.2069, 2.9819, 3.2347, 3.1453, 3.1186], device='cuda:5'), covar=tensor([0.0343, 0.0199, 0.0170, 0.0135, 0.0528, 0.0179, 0.0622, 0.0284], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0129, 0.0175, 0.0146, 0.0205, 0.0165, 0.0128, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 23:09:18,170 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:09:56,488 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8699, 3.0657, 2.1869, 4.0261, 3.7870, 3.8784, 1.5567, 2.7876], device='cuda:5'), covar=tensor([0.1272, 0.0365, 0.1290, 0.0067, 0.0210, 0.0277, 0.1232, 0.0706], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0128, 0.0164, 0.0071, 0.0134, 0.0141, 0.0154, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-27 23:10:25,187 INFO [train.py:904] (5/8) Epoch 3, batch 5600, loss[loss=0.3698, simple_loss=0.4165, pruned_loss=0.1616, over 15233.00 frames. ], tot_loss[loss=0.2823, simple_loss=0.3491, pruned_loss=0.1077, over 3093875.48 frames. ], batch size: 191, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:10:56,250 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:11:16,985 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8193, 3.7211, 1.5787, 3.8076, 2.4127, 3.8789, 1.8670, 2.7437], device='cuda:5'), covar=tensor([0.0048, 0.0190, 0.1730, 0.0039, 0.0764, 0.0309, 0.1373, 0.0588], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0131, 0.0174, 0.0080, 0.0161, 0.0163, 0.0178, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 23:11:21,473 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.427e+02 5.585e+02 6.873e+02 8.559e+02 2.132e+03, threshold=1.375e+03, percent-clipped=5.0 2023-04-27 23:11:48,515 INFO [train.py:904] (5/8) Epoch 3, batch 5650, loss[loss=0.298, simple_loss=0.3677, pruned_loss=0.1142, over 16601.00 frames. ], tot_loss[loss=0.2898, simple_loss=0.3553, pruned_loss=0.1121, over 3088130.42 frames. ], batch size: 57, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:13:10,054 INFO [train.py:904] (5/8) Epoch 3, batch 5700, loss[loss=0.323, simple_loss=0.3884, pruned_loss=0.1288, over 15442.00 frames. ], tot_loss[loss=0.2953, simple_loss=0.359, pruned_loss=0.1158, over 3060544.76 frames. ], batch size: 190, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:13:29,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8439, 3.7570, 4.3665, 4.2860, 4.3253, 3.8619, 3.9646, 3.9030], device='cuda:5'), covar=tensor([0.0255, 0.0336, 0.0322, 0.0416, 0.0369, 0.0336, 0.0751, 0.0354], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0164, 0.0185, 0.0183, 0.0220, 0.0188, 0.0285, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-27 23:13:44,372 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5994, 3.9722, 3.5975, 3.7440, 2.6447, 3.8139, 3.6477, 3.4500], device='cuda:5'), covar=tensor([0.0645, 0.0249, 0.0390, 0.0309, 0.1588, 0.0416, 0.0716, 0.0442], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0127, 0.0170, 0.0142, 0.0201, 0.0160, 0.0125, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 23:14:00,538 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.120e+02 5.064e+02 6.111e+02 7.661e+02 1.195e+03, threshold=1.222e+03, percent-clipped=0.0 2023-04-27 23:14:27,441 INFO [train.py:904] (5/8) Epoch 3, batch 5750, loss[loss=0.4042, simple_loss=0.4267, pruned_loss=0.1908, over 11680.00 frames. ], tot_loss[loss=0.2979, simple_loss=0.3618, pruned_loss=0.117, over 3053438.98 frames. ], batch size: 246, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:14:39,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7835, 2.6204, 2.3154, 4.2723, 1.9343, 3.9410, 2.4252, 2.4410], device='cuda:5'), covar=tensor([0.0350, 0.0791, 0.0557, 0.0173, 0.1927, 0.0220, 0.1015, 0.1546], device='cuda:5'), in_proj_covar=tensor([0.0252, 0.0230, 0.0195, 0.0251, 0.0308, 0.0204, 0.0220, 0.0294], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:14:47,785 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6819, 4.1144, 4.0782, 1.8757, 4.4049, 4.3930, 3.3077, 3.1714], device='cuda:5'), covar=tensor([0.0748, 0.0106, 0.0147, 0.1343, 0.0034, 0.0025, 0.0250, 0.0398], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0079, 0.0076, 0.0142, 0.0072, 0.0069, 0.0111, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-27 23:15:22,125 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:15:47,185 INFO [train.py:904] (5/8) Epoch 3, batch 5800, loss[loss=0.2908, simple_loss=0.3468, pruned_loss=0.1175, over 11844.00 frames. ], tot_loss[loss=0.2949, simple_loss=0.3606, pruned_loss=0.1146, over 3068515.69 frames. ], batch size: 248, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:48,971 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:16:36,165 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:16:38,591 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.751e+02 5.049e+02 6.109e+02 8.114e+02 1.629e+03, threshold=1.222e+03, percent-clipped=2.0 2023-04-27 23:17:05,086 INFO [train.py:904] (5/8) Epoch 3, batch 5850, loss[loss=0.3245, simple_loss=0.3858, pruned_loss=0.1316, over 16659.00 frames. ], tot_loss[loss=0.2908, simple_loss=0.3576, pruned_loss=0.112, over 3071376.52 frames. ], batch size: 62, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:17:24,138 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:17:31,395 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:17:34,371 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4811, 4.7217, 4.4214, 4.5086, 4.1621, 4.1009, 4.2944, 4.7703], device='cuda:5'), covar=tensor([0.0463, 0.0631, 0.0920, 0.0379, 0.0507, 0.0794, 0.0588, 0.0587], device='cuda:5'), in_proj_covar=tensor([0.0272, 0.0377, 0.0333, 0.0240, 0.0241, 0.0241, 0.0302, 0.0264], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:17:39,110 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1237, 4.2540, 3.2598, 2.8793, 3.6312, 2.8112, 4.6755, 4.8009], device='cuda:5'), covar=tensor([0.1905, 0.0635, 0.1213, 0.1050, 0.1538, 0.1015, 0.0341, 0.0297], device='cuda:5'), in_proj_covar=tensor([0.0265, 0.0237, 0.0253, 0.0209, 0.0292, 0.0190, 0.0215, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:18:26,965 INFO [train.py:904] (5/8) Epoch 3, batch 5900, loss[loss=0.2684, simple_loss=0.3494, pruned_loss=0.09372, over 16915.00 frames. ], tot_loss[loss=0.2881, simple_loss=0.3552, pruned_loss=0.1105, over 3081819.64 frames. ], batch size: 109, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:18:51,856 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:19:15,311 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:19:22,628 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.765e+02 4.492e+02 5.249e+02 6.969e+02 1.603e+03, threshold=1.050e+03, percent-clipped=2.0 2023-04-27 23:19:47,800 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:19:48,589 INFO [train.py:904] (5/8) Epoch 3, batch 5950, loss[loss=0.2489, simple_loss=0.3247, pruned_loss=0.08655, over 16997.00 frames. ], tot_loss[loss=0.2868, simple_loss=0.3557, pruned_loss=0.109, over 3088649.78 frames. ], batch size: 53, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:21:07,942 INFO [train.py:904] (5/8) Epoch 3, batch 6000, loss[loss=0.3184, simple_loss=0.3618, pruned_loss=0.1375, over 11478.00 frames. ], tot_loss[loss=0.2856, simple_loss=0.3545, pruned_loss=0.1084, over 3105039.71 frames. ], batch size: 248, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:21:07,943 INFO [train.py:929] (5/8) Computing validation loss 2023-04-27 23:21:18,891 INFO [train.py:938] (5/8) Epoch 3, validation: loss=0.2097, simple_loss=0.3184, pruned_loss=0.05055, over 944034.00 frames. 2023-04-27 23:21:18,892 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-27 23:21:34,610 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:22:06,061 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7398, 5.0574, 5.1543, 5.1311, 5.0846, 5.6298, 5.2260, 5.0466], device='cuda:5'), covar=tensor([0.0630, 0.2586, 0.1047, 0.1253, 0.1865, 0.0724, 0.1159, 0.1811], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0309, 0.0289, 0.0270, 0.0357, 0.0313, 0.0252, 0.0361], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 23:22:07,349 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.939e+02 4.548e+02 5.519e+02 7.491e+02 1.813e+03, threshold=1.104e+03, percent-clipped=4.0 2023-04-27 23:22:36,232 INFO [train.py:904] (5/8) Epoch 3, batch 6050, loss[loss=0.2788, simple_loss=0.357, pruned_loss=0.1003, over 16992.00 frames. ], tot_loss[loss=0.2844, simple_loss=0.3533, pruned_loss=0.1078, over 3102414.34 frames. ], batch size: 41, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:23:51,472 INFO [train.py:904] (5/8) Epoch 3, batch 6100, loss[loss=0.2386, simple_loss=0.3218, pruned_loss=0.07764, over 16895.00 frames. ], tot_loss[loss=0.2809, simple_loss=0.3513, pruned_loss=0.1053, over 3123244.68 frames. ], batch size: 96, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:23:54,171 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-27 23:24:12,791 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:24:42,456 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.827e+02 4.226e+02 5.750e+02 6.713e+02 1.587e+03, threshold=1.150e+03, percent-clipped=3.0 2023-04-27 23:25:11,819 INFO [train.py:904] (5/8) Epoch 3, batch 6150, loss[loss=0.3424, simple_loss=0.3809, pruned_loss=0.1519, over 11502.00 frames. ], tot_loss[loss=0.2786, simple_loss=0.3488, pruned_loss=0.1042, over 3123637.87 frames. ], batch size: 247, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:25:23,349 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:25:50,287 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:26:23,298 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9181, 3.8448, 3.7222, 3.8127, 3.4227, 3.8886, 3.6242, 3.6344], device='cuda:5'), covar=tensor([0.0338, 0.0201, 0.0188, 0.0136, 0.0634, 0.0196, 0.0619, 0.0309], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0130, 0.0171, 0.0144, 0.0200, 0.0161, 0.0127, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 23:26:28,221 INFO [train.py:904] (5/8) Epoch 3, batch 6200, loss[loss=0.2939, simple_loss=0.3653, pruned_loss=0.1113, over 16732.00 frames. ], tot_loss[loss=0.2772, simple_loss=0.347, pruned_loss=0.1037, over 3123697.82 frames. ], batch size: 134, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:26:48,688 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:27:03,501 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:27:06,548 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 23:27:18,521 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.570e+02 4.817e+02 5.994e+02 8.159e+02 2.733e+03, threshold=1.199e+03, percent-clipped=9.0 2023-04-27 23:27:31,934 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6409, 4.9404, 5.0710, 5.0344, 5.0042, 5.5412, 5.1947, 4.9232], device='cuda:5'), covar=tensor([0.0688, 0.1228, 0.1018, 0.1342, 0.2017, 0.0727, 0.0881, 0.1924], device='cuda:5'), in_proj_covar=tensor([0.0229, 0.0314, 0.0289, 0.0278, 0.0364, 0.0314, 0.0254, 0.0368], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 23:27:36,728 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-27 23:27:41,897 INFO [train.py:904] (5/8) Epoch 3, batch 6250, loss[loss=0.2301, simple_loss=0.322, pruned_loss=0.06907, over 16813.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3467, pruned_loss=0.103, over 3143372.75 frames. ], batch size: 102, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:27:58,190 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:28:26,377 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-27 23:28:44,060 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2218, 3.9857, 3.8564, 4.4291, 4.4979, 4.1281, 4.4879, 4.3858], device='cuda:5'), covar=tensor([0.0734, 0.0748, 0.1740, 0.0566, 0.0541, 0.0640, 0.0616, 0.0592], device='cuda:5'), in_proj_covar=tensor([0.0288, 0.0349, 0.0453, 0.0348, 0.0260, 0.0246, 0.0289, 0.0285], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:28:54,867 INFO [train.py:904] (5/8) Epoch 3, batch 6300, loss[loss=0.3064, simple_loss=0.3688, pruned_loss=0.122, over 16722.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3461, pruned_loss=0.102, over 3154989.26 frames. ], batch size: 134, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:29:02,922 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:29:48,252 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.782e+02 4.322e+02 5.254e+02 6.822e+02 1.499e+03, threshold=1.051e+03, percent-clipped=4.0 2023-04-27 23:30:05,364 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0918, 4.7903, 4.8693, 4.8343, 4.3542, 4.8485, 4.8969, 4.5479], device='cuda:5'), covar=tensor([0.0277, 0.0192, 0.0156, 0.0113, 0.0719, 0.0207, 0.0134, 0.0277], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0128, 0.0166, 0.0138, 0.0196, 0.0159, 0.0124, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-27 23:30:11,918 INFO [train.py:904] (5/8) Epoch 3, batch 6350, loss[loss=0.2709, simple_loss=0.3426, pruned_loss=0.09955, over 16694.00 frames. ], tot_loss[loss=0.2797, simple_loss=0.3485, pruned_loss=0.1054, over 3117004.98 frames. ], batch size: 134, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:30:14,341 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7101, 1.3730, 1.4353, 1.6548, 1.7213, 1.7997, 1.4954, 1.7536], device='cuda:5'), covar=tensor([0.0054, 0.0120, 0.0067, 0.0090, 0.0054, 0.0042, 0.0103, 0.0034], device='cuda:5'), in_proj_covar=tensor([0.0085, 0.0121, 0.0106, 0.0102, 0.0092, 0.0068, 0.0114, 0.0062], device='cuda:5'), out_proj_covar=tensor([1.3013e-04, 1.9102e-04, 1.7128e-04, 1.6517e-04, 1.4443e-04, 1.0501e-04, 1.7663e-04, 9.5567e-05], device='cuda:5') 2023-04-27 23:30:27,966 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0199, 3.7968, 3.9666, 4.2285, 4.2759, 3.8958, 4.2712, 4.2151], device='cuda:5'), covar=tensor([0.0587, 0.0582, 0.1035, 0.0386, 0.0383, 0.0670, 0.0361, 0.0391], device='cuda:5'), in_proj_covar=tensor([0.0289, 0.0355, 0.0461, 0.0352, 0.0262, 0.0248, 0.0289, 0.0288], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:31:25,039 INFO [train.py:904] (5/8) Epoch 3, batch 6400, loss[loss=0.2543, simple_loss=0.3334, pruned_loss=0.08765, over 16232.00 frames. ], tot_loss[loss=0.2823, simple_loss=0.3496, pruned_loss=0.1075, over 3089577.71 frames. ], batch size: 35, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:31:49,432 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8061, 1.2705, 1.4428, 1.7024, 1.7060, 1.8271, 1.4848, 1.7994], device='cuda:5'), covar=tensor([0.0057, 0.0131, 0.0070, 0.0094, 0.0059, 0.0038, 0.0121, 0.0030], device='cuda:5'), in_proj_covar=tensor([0.0085, 0.0121, 0.0106, 0.0102, 0.0093, 0.0068, 0.0115, 0.0061], device='cuda:5'), out_proj_covar=tensor([1.3107e-04, 1.9164e-04, 1.7061e-04, 1.6540e-04, 1.4501e-04, 1.0439e-04, 1.7833e-04, 9.4621e-05], device='cuda:5') 2023-04-27 23:32:12,499 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.974e+02 5.048e+02 6.146e+02 7.949e+02 2.287e+03, threshold=1.229e+03, percent-clipped=11.0 2023-04-27 23:32:35,287 INFO [train.py:904] (5/8) Epoch 3, batch 6450, loss[loss=0.2558, simple_loss=0.3121, pruned_loss=0.09977, over 11461.00 frames. ], tot_loss[loss=0.2806, simple_loss=0.3486, pruned_loss=0.1063, over 3081320.71 frames. ], batch size: 247, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:32:41,258 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7631, 5.0319, 5.1296, 5.0547, 4.9690, 5.5370, 5.2024, 4.9314], device='cuda:5'), covar=tensor([0.0622, 0.1181, 0.1085, 0.1153, 0.1827, 0.0695, 0.0817, 0.1792], device='cuda:5'), in_proj_covar=tensor([0.0231, 0.0313, 0.0291, 0.0277, 0.0359, 0.0313, 0.0256, 0.0370], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 23:32:47,299 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:05,729 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:14,386 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 23:33:30,274 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8645, 2.8006, 2.3137, 3.7822, 3.5598, 3.5921, 1.5698, 2.7437], device='cuda:5'), covar=tensor([0.1333, 0.0479, 0.1188, 0.0073, 0.0249, 0.0325, 0.1354, 0.0717], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0132, 0.0165, 0.0071, 0.0137, 0.0144, 0.0154, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-04-27 23:33:53,763 INFO [train.py:904] (5/8) Epoch 3, batch 6500, loss[loss=0.2567, simple_loss=0.3205, pruned_loss=0.09644, over 16443.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3461, pruned_loss=0.105, over 3095213.30 frames. ], batch size: 35, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:34:01,036 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:29,505 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:44,551 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.621e+02 5.604e+02 6.997e+02 1.424e+03, threshold=1.121e+03, percent-clipped=2.0 2023-04-27 23:34:50,017 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-27 23:35:05,294 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7707, 1.2006, 1.4837, 1.7529, 1.8088, 1.8392, 1.4271, 1.8152], device='cuda:5'), covar=tensor([0.0049, 0.0125, 0.0068, 0.0078, 0.0052, 0.0039, 0.0124, 0.0028], device='cuda:5'), in_proj_covar=tensor([0.0085, 0.0124, 0.0107, 0.0100, 0.0092, 0.0068, 0.0115, 0.0062], device='cuda:5'), out_proj_covar=tensor([1.3028e-04, 1.9491e-04, 1.7336e-04, 1.6283e-04, 1.4280e-04, 1.0394e-04, 1.7883e-04, 9.4845e-05], device='cuda:5') 2023-04-27 23:35:12,122 INFO [train.py:904] (5/8) Epoch 3, batch 6550, loss[loss=0.2547, simple_loss=0.3411, pruned_loss=0.08413, over 17015.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.3498, pruned_loss=0.1069, over 3081021.58 frames. ], batch size: 50, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:35:45,523 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:36:28,913 INFO [train.py:904] (5/8) Epoch 3, batch 6600, loss[loss=0.2528, simple_loss=0.3336, pruned_loss=0.08598, over 16822.00 frames. ], tot_loss[loss=0.2844, simple_loss=0.3531, pruned_loss=0.1078, over 3088098.40 frames. ], batch size: 96, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:36:33,585 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:36:37,699 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:19,525 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 23:37:20,964 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.349e+02 4.809e+02 5.887e+02 7.395e+02 2.123e+03, threshold=1.177e+03, percent-clipped=7.0 2023-04-27 23:37:45,819 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:46,399 INFO [train.py:904] (5/8) Epoch 3, batch 6650, loss[loss=0.2846, simple_loss=0.3497, pruned_loss=0.1098, over 16699.00 frames. ], tot_loss[loss=0.2861, simple_loss=0.3537, pruned_loss=0.1092, over 3086855.03 frames. ], batch size: 124, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:37:51,225 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:38:09,010 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:38:45,560 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3526, 3.3404, 1.6830, 3.4236, 2.2936, 3.4268, 1.7657, 2.5520], device='cuda:5'), covar=tensor([0.0067, 0.0265, 0.1415, 0.0042, 0.0723, 0.0379, 0.1324, 0.0617], device='cuda:5'), in_proj_covar=tensor([0.0088, 0.0133, 0.0175, 0.0081, 0.0162, 0.0163, 0.0181, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-27 23:39:03,595 INFO [train.py:904] (5/8) Epoch 3, batch 6700, loss[loss=0.3411, simple_loss=0.3774, pruned_loss=0.1524, over 11383.00 frames. ], tot_loss[loss=0.2854, simple_loss=0.3526, pruned_loss=0.1091, over 3090022.86 frames. ], batch size: 246, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:39:19,069 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:57,112 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.121e+02 4.872e+02 5.906e+02 7.141e+02 1.703e+03, threshold=1.181e+03, percent-clipped=1.0 2023-04-27 23:40:21,196 INFO [train.py:904] (5/8) Epoch 3, batch 6750, loss[loss=0.2774, simple_loss=0.3427, pruned_loss=0.106, over 17081.00 frames. ], tot_loss[loss=0.2838, simple_loss=0.3509, pruned_loss=0.1084, over 3112113.65 frames. ], batch size: 55, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:40:51,280 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:41:37,198 INFO [train.py:904] (5/8) Epoch 3, batch 6800, loss[loss=0.302, simple_loss=0.3667, pruned_loss=0.1186, over 15372.00 frames. ], tot_loss[loss=0.2823, simple_loss=0.3503, pruned_loss=0.1072, over 3132718.60 frames. ], batch size: 190, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:42:03,929 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:19,014 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 23:42:22,325 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3009, 4.1218, 4.2984, 4.5637, 4.5722, 4.1742, 4.5909, 4.5669], device='cuda:5'), covar=tensor([0.0658, 0.0578, 0.1043, 0.0383, 0.0443, 0.0564, 0.0407, 0.0333], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0349, 0.0451, 0.0343, 0.0262, 0.0250, 0.0282, 0.0283], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:42:31,276 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.734e+02 4.279e+02 5.417e+02 7.296e+02 1.152e+03, threshold=1.083e+03, percent-clipped=0.0 2023-04-27 23:42:44,347 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7422, 1.4107, 1.8081, 2.3855, 2.5491, 2.7597, 1.5153, 2.7461], device='cuda:5'), covar=tensor([0.0043, 0.0212, 0.0133, 0.0106, 0.0060, 0.0050, 0.0186, 0.0040], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0125, 0.0108, 0.0101, 0.0093, 0.0070, 0.0116, 0.0062], device='cuda:5'), out_proj_covar=tensor([1.3246e-04, 1.9666e-04, 1.7491e-04, 1.6269e-04, 1.4469e-04, 1.0664e-04, 1.7999e-04, 9.4501e-05], device='cuda:5') 2023-04-27 23:42:51,772 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:55,403 INFO [train.py:904] (5/8) Epoch 3, batch 6850, loss[loss=0.2924, simple_loss=0.3603, pruned_loss=0.1123, over 15388.00 frames. ], tot_loss[loss=0.283, simple_loss=0.3512, pruned_loss=0.1074, over 3133502.90 frames. ], batch size: 191, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:43:26,037 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4880, 3.0418, 2.5392, 2.3856, 2.2418, 2.0593, 2.9882, 3.1464], device='cuda:5'), covar=tensor([0.1325, 0.0505, 0.0982, 0.0972, 0.1727, 0.1191, 0.0392, 0.0384], device='cuda:5'), in_proj_covar=tensor([0.0268, 0.0238, 0.0255, 0.0213, 0.0298, 0.0191, 0.0220, 0.0205], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:44:10,173 INFO [train.py:904] (5/8) Epoch 3, batch 6900, loss[loss=0.2761, simple_loss=0.3531, pruned_loss=0.09961, over 16199.00 frames. ], tot_loss[loss=0.2838, simple_loss=0.3538, pruned_loss=0.1069, over 3143663.07 frames. ], batch size: 165, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:44:22,748 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:44:23,230 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 23:45:02,580 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.150e+02 4.514e+02 5.203e+02 6.904e+02 1.170e+03, threshold=1.041e+03, percent-clipped=2.0 2023-04-27 23:45:06,810 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2398, 3.1298, 3.1568, 3.3398, 3.3601, 3.1110, 3.3361, 3.4203], device='cuda:5'), covar=tensor([0.0525, 0.0532, 0.0986, 0.0444, 0.0507, 0.1517, 0.0621, 0.0439], device='cuda:5'), in_proj_covar=tensor([0.0287, 0.0357, 0.0458, 0.0350, 0.0268, 0.0254, 0.0289, 0.0288], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:45:28,470 INFO [train.py:904] (5/8) Epoch 3, batch 6950, loss[loss=0.2886, simple_loss=0.3539, pruned_loss=0.1117, over 16728.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.3576, pruned_loss=0.1113, over 3097063.99 frames. ], batch size: 57, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:45:41,661 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:45:56,966 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 23:46:43,379 INFO [train.py:904] (5/8) Epoch 3, batch 7000, loss[loss=0.2762, simple_loss=0.3552, pruned_loss=0.09865, over 16302.00 frames. ], tot_loss[loss=0.2891, simple_loss=0.3575, pruned_loss=0.1104, over 3096548.97 frames. ], batch size: 165, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:46:51,661 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:19,257 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4842, 3.8917, 3.2763, 3.7332, 3.3438, 3.5646, 3.5776, 3.7567], device='cuda:5'), covar=tensor([0.1442, 0.1181, 0.2033, 0.0874, 0.1220, 0.1469, 0.0971, 0.1260], device='cuda:5'), in_proj_covar=tensor([0.0274, 0.0378, 0.0337, 0.0242, 0.0243, 0.0241, 0.0300, 0.0265], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:47:35,872 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.986e+02 5.041e+02 6.145e+02 8.860e+02 1.565e+03, threshold=1.229e+03, percent-clipped=7.0 2023-04-27 23:47:45,528 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-04-27 23:48:01,115 INFO [train.py:904] (5/8) Epoch 3, batch 7050, loss[loss=0.3072, simple_loss=0.3658, pruned_loss=0.1243, over 15200.00 frames. ], tot_loss[loss=0.2875, simple_loss=0.357, pruned_loss=0.109, over 3105516.66 frames. ], batch size: 190, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:49:19,645 INFO [train.py:904] (5/8) Epoch 3, batch 7100, loss[loss=0.2694, simple_loss=0.346, pruned_loss=0.09635, over 16692.00 frames. ], tot_loss[loss=0.2851, simple_loss=0.3546, pruned_loss=0.1078, over 3110903.40 frames. ], batch size: 89, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:49:37,736 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9322, 3.0779, 3.4757, 3.4673, 3.4122, 3.1369, 3.2591, 3.3076], device='cuda:5'), covar=tensor([0.0279, 0.0465, 0.0285, 0.0357, 0.0410, 0.0354, 0.0557, 0.0320], device='cuda:5'), in_proj_covar=tensor([0.0184, 0.0172, 0.0186, 0.0181, 0.0225, 0.0193, 0.0289, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-27 23:49:46,875 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-27 23:50:12,103 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.055e+02 4.788e+02 6.087e+02 7.690e+02 2.114e+03, threshold=1.217e+03, percent-clipped=3.0 2023-04-27 23:50:36,348 INFO [train.py:904] (5/8) Epoch 3, batch 7150, loss[loss=0.3241, simple_loss=0.3697, pruned_loss=0.1392, over 11413.00 frames. ], tot_loss[loss=0.284, simple_loss=0.3527, pruned_loss=0.1077, over 3093994.66 frames. ], batch size: 247, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:51:51,152 INFO [train.py:904] (5/8) Epoch 3, batch 7200, loss[loss=0.2259, simple_loss=0.3155, pruned_loss=0.06808, over 16842.00 frames. ], tot_loss[loss=0.279, simple_loss=0.349, pruned_loss=0.1045, over 3104239.28 frames. ], batch size: 96, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:51:55,807 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:52:45,516 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.403e+02 3.669e+02 4.524e+02 6.083e+02 1.066e+03, threshold=9.047e+02, percent-clipped=1.0 2023-04-27 23:53:12,425 INFO [train.py:904] (5/8) Epoch 3, batch 7250, loss[loss=0.2801, simple_loss=0.3378, pruned_loss=0.1112, over 11307.00 frames. ], tot_loss[loss=0.2751, simple_loss=0.3456, pruned_loss=0.1023, over 3105222.98 frames. ], batch size: 246, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:53:24,786 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3350, 4.2295, 4.2155, 4.1948, 3.7358, 4.2536, 4.1087, 4.0158], device='cuda:5'), covar=tensor([0.0301, 0.0179, 0.0158, 0.0129, 0.0637, 0.0195, 0.0257, 0.0288], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0127, 0.0166, 0.0139, 0.0195, 0.0157, 0.0122, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-27 23:53:26,551 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:54:26,530 INFO [train.py:904] (5/8) Epoch 3, batch 7300, loss[loss=0.2741, simple_loss=0.3583, pruned_loss=0.09493, over 16860.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3446, pruned_loss=0.102, over 3095800.41 frames. ], batch size: 96, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:54:30,892 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9494, 2.3690, 2.2931, 3.1985, 2.9739, 3.1406, 1.8451, 2.5668], device='cuda:5'), covar=tensor([0.1379, 0.0573, 0.1191, 0.0141, 0.0316, 0.0322, 0.1366, 0.0803], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0136, 0.0163, 0.0071, 0.0137, 0.0146, 0.0155, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-27 23:54:35,535 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:54:38,062 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:55:17,371 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.622e+02 4.400e+02 5.943e+02 7.946e+02 1.369e+03, threshold=1.189e+03, percent-clipped=13.0 2023-04-27 23:55:40,980 INFO [train.py:904] (5/8) Epoch 3, batch 7350, loss[loss=0.3152, simple_loss=0.3512, pruned_loss=0.1396, over 11087.00 frames. ], tot_loss[loss=0.274, simple_loss=0.3441, pruned_loss=0.1019, over 3086718.23 frames. ], batch size: 246, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:55:46,793 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:56:59,434 INFO [train.py:904] (5/8) Epoch 3, batch 7400, loss[loss=0.2569, simple_loss=0.3332, pruned_loss=0.09031, over 16996.00 frames. ], tot_loss[loss=0.2751, simple_loss=0.3453, pruned_loss=0.1024, over 3116195.10 frames. ], batch size: 53, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:57:00,283 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 23:57:20,321 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27714.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:57:55,130 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.894e+02 4.696e+02 5.728e+02 6.981e+02 1.392e+03, threshold=1.146e+03, percent-clipped=2.0 2023-04-27 23:58:18,295 INFO [train.py:904] (5/8) Epoch 3, batch 7450, loss[loss=0.3115, simple_loss=0.3525, pruned_loss=0.1352, over 11546.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.3467, pruned_loss=0.1043, over 3103176.78 frames. ], batch size: 248, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:58:58,419 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:58:59,933 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 23:59:26,846 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2676, 4.3213, 4.5057, 2.9839, 4.1931, 4.3130, 4.4109, 1.9978], device='cuda:5'), covar=tensor([0.0332, 0.0011, 0.0017, 0.0215, 0.0021, 0.0039, 0.0015, 0.0332], device='cuda:5'), in_proj_covar=tensor([0.0110, 0.0051, 0.0056, 0.0109, 0.0051, 0.0061, 0.0056, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-27 23:59:39,387 INFO [train.py:904] (5/8) Epoch 3, batch 7500, loss[loss=0.2689, simple_loss=0.3378, pruned_loss=0.09997, over 17209.00 frames. ], tot_loss[loss=0.2774, simple_loss=0.3471, pruned_loss=0.1039, over 3088832.25 frames. ], batch size: 44, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:59:44,203 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:00:33,137 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.883e+02 4.957e+02 5.924e+02 7.629e+02 1.550e+03, threshold=1.185e+03, percent-clipped=6.0 2023-04-28 00:00:55,954 INFO [train.py:904] (5/8) Epoch 3, batch 7550, loss[loss=0.2361, simple_loss=0.3038, pruned_loss=0.08422, over 16575.00 frames. ], tot_loss[loss=0.2786, simple_loss=0.3467, pruned_loss=0.1053, over 3052968.18 frames. ], batch size: 62, lr: 1.96e-02, grad_scale: 4.0 2023-04-28 00:00:58,759 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:34,061 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9629, 3.0673, 3.4569, 3.4187, 3.4041, 3.1385, 3.1742, 3.3248], device='cuda:5'), covar=tensor([0.0298, 0.0430, 0.0309, 0.0438, 0.0388, 0.0344, 0.0710, 0.0291], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0169, 0.0184, 0.0188, 0.0225, 0.0195, 0.0292, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 00:02:02,972 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7546, 3.7164, 2.4187, 4.7269, 4.5786, 4.1662, 1.9077, 3.2052], device='cuda:5'), covar=tensor([0.1308, 0.0332, 0.1082, 0.0060, 0.0155, 0.0261, 0.1077, 0.0624], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0131, 0.0160, 0.0070, 0.0135, 0.0143, 0.0151, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-04-28 00:02:13,372 INFO [train.py:904] (5/8) Epoch 3, batch 7600, loss[loss=0.3319, simple_loss=0.3993, pruned_loss=0.1322, over 16269.00 frames. ], tot_loss[loss=0.2786, simple_loss=0.3463, pruned_loss=0.1055, over 3052280.42 frames. ], batch size: 165, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:02:29,921 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1084, 2.1673, 1.8017, 2.0333, 2.7259, 2.4627, 3.3541, 3.1226], device='cuda:5'), covar=tensor([0.0017, 0.0134, 0.0178, 0.0149, 0.0074, 0.0133, 0.0030, 0.0060], device='cuda:5'), in_proj_covar=tensor([0.0054, 0.0120, 0.0125, 0.0123, 0.0115, 0.0124, 0.0079, 0.0100], device='cuda:5'), out_proj_covar=tensor([7.3196e-05, 1.6882e-04, 1.7100e-04, 1.7258e-04, 1.6567e-04, 1.7593e-04, 1.1114e-04, 1.4420e-04], device='cuda:5') 2023-04-28 00:02:41,705 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-28 00:03:08,898 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.035e+02 4.816e+02 6.068e+02 7.688e+02 1.978e+03, threshold=1.214e+03, percent-clipped=6.0 2023-04-28 00:03:31,746 INFO [train.py:904] (5/8) Epoch 3, batch 7650, loss[loss=0.2892, simple_loss=0.3557, pruned_loss=0.1114, over 15275.00 frames. ], tot_loss[loss=0.279, simple_loss=0.3468, pruned_loss=0.1056, over 3083495.77 frames. ], batch size: 190, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:03:38,522 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9281, 3.2030, 3.4554, 3.4389, 3.4082, 3.1343, 3.2007, 3.3169], device='cuda:5'), covar=tensor([0.0293, 0.0343, 0.0298, 0.0340, 0.0376, 0.0329, 0.0697, 0.0289], device='cuda:5'), in_proj_covar=tensor([0.0190, 0.0174, 0.0187, 0.0189, 0.0228, 0.0197, 0.0297, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 00:04:45,625 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4562, 3.3044, 2.9512, 1.7547, 2.6715, 1.9817, 3.0657, 3.3432], device='cuda:5'), covar=tensor([0.0305, 0.0542, 0.0642, 0.1923, 0.0800, 0.1118, 0.0554, 0.0716], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0121, 0.0158, 0.0148, 0.0139, 0.0133, 0.0147, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 00:04:52,687 INFO [train.py:904] (5/8) Epoch 3, batch 7700, loss[loss=0.3165, simple_loss=0.3625, pruned_loss=0.1353, over 11537.00 frames. ], tot_loss[loss=0.2792, simple_loss=0.3465, pruned_loss=0.1059, over 3073288.63 frames. ], batch size: 246, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:05:46,978 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.821e+02 4.680e+02 5.750e+02 6.748e+02 1.214e+03, threshold=1.150e+03, percent-clipped=1.0 2023-04-28 00:06:11,248 INFO [train.py:904] (5/8) Epoch 3, batch 7750, loss[loss=0.2639, simple_loss=0.3477, pruned_loss=0.09006, over 16734.00 frames. ], tot_loss[loss=0.2794, simple_loss=0.347, pruned_loss=0.1059, over 3060514.67 frames. ], batch size: 134, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:06:40,896 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:07:28,347 INFO [train.py:904] (5/8) Epoch 3, batch 7800, loss[loss=0.2864, simple_loss=0.3491, pruned_loss=0.1118, over 16415.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3473, pruned_loss=0.1059, over 3076568.61 frames. ], batch size: 146, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:08:22,851 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.157e+02 4.949e+02 5.902e+02 7.570e+02 1.555e+03, threshold=1.180e+03, percent-clipped=4.0 2023-04-28 00:08:45,074 INFO [train.py:904] (5/8) Epoch 3, batch 7850, loss[loss=0.3015, simple_loss=0.3647, pruned_loss=0.1191, over 15357.00 frames. ], tot_loss[loss=0.2801, simple_loss=0.3486, pruned_loss=0.1058, over 3090566.15 frames. ], batch size: 190, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:09:03,298 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 00:10:00,841 INFO [train.py:904] (5/8) Epoch 3, batch 7900, loss[loss=0.2495, simple_loss=0.3221, pruned_loss=0.08838, over 16580.00 frames. ], tot_loss[loss=0.278, simple_loss=0.3471, pruned_loss=0.1045, over 3095730.80 frames. ], batch size: 62, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:10:45,268 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:55,799 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.334e+02 4.942e+02 5.999e+02 2.073e+03, threshold=9.884e+02, percent-clipped=3.0 2023-04-28 00:11:18,513 INFO [train.py:904] (5/8) Epoch 3, batch 7950, loss[loss=0.2918, simple_loss=0.3643, pruned_loss=0.1096, over 16772.00 frames. ], tot_loss[loss=0.2787, simple_loss=0.3473, pruned_loss=0.105, over 3095078.32 frames. ], batch size: 124, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:11:26,026 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:12:13,426 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:18,744 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:33,532 INFO [train.py:904] (5/8) Epoch 3, batch 8000, loss[loss=0.2451, simple_loss=0.3247, pruned_loss=0.08273, over 16891.00 frames. ], tot_loss[loss=0.2792, simple_loss=0.3477, pruned_loss=0.1053, over 3093834.17 frames. ], batch size: 96, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:12:56,265 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 00:13:20,265 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 00:13:27,063 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.455e+02 4.065e+02 5.202e+02 6.886e+02 1.574e+03, threshold=1.040e+03, percent-clipped=4.0 2023-04-28 00:13:46,431 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:49,306 INFO [train.py:904] (5/8) Epoch 3, batch 8050, loss[loss=0.2736, simple_loss=0.3488, pruned_loss=0.09924, over 16543.00 frames. ], tot_loss[loss=0.2784, simple_loss=0.347, pruned_loss=0.1049, over 3083945.16 frames. ], batch size: 68, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:14:08,508 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6662, 3.3582, 3.4850, 2.2684, 3.3781, 3.4635, 3.4854, 1.7184], device='cuda:5'), covar=tensor([0.0382, 0.0031, 0.0034, 0.0230, 0.0032, 0.0051, 0.0025, 0.0322], device='cuda:5'), in_proj_covar=tensor([0.0112, 0.0054, 0.0057, 0.0110, 0.0053, 0.0063, 0.0059, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 00:14:18,844 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:14:23,366 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 00:15:05,405 INFO [train.py:904] (5/8) Epoch 3, batch 8100, loss[loss=0.2795, simple_loss=0.3505, pruned_loss=0.1043, over 16421.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3463, pruned_loss=0.1037, over 3091553.17 frames. ], batch size: 146, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:15:30,658 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:15:38,637 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1282, 4.7298, 4.9605, 5.3295, 5.4241, 4.6997, 5.4541, 5.3886], device='cuda:5'), covar=tensor([0.0588, 0.0677, 0.1153, 0.0388, 0.0338, 0.0530, 0.0378, 0.0305], device='cuda:5'), in_proj_covar=tensor([0.0297, 0.0373, 0.0476, 0.0367, 0.0275, 0.0265, 0.0299, 0.0301], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:15:57,044 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.864e+02 4.688e+02 5.875e+02 7.578e+02 1.406e+03, threshold=1.175e+03, percent-clipped=5.0 2023-04-28 00:16:20,015 INFO [train.py:904] (5/8) Epoch 3, batch 8150, loss[loss=0.2447, simple_loss=0.3056, pruned_loss=0.0919, over 16673.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.3443, pruned_loss=0.1032, over 3093671.71 frames. ], batch size: 57, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:16:59,704 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 00:17:15,230 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:23,262 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:36,829 INFO [train.py:904] (5/8) Epoch 3, batch 8200, loss[loss=0.2751, simple_loss=0.3486, pruned_loss=0.1008, over 15075.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3423, pruned_loss=0.1027, over 3097353.70 frames. ], batch size: 190, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:18:27,911 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8717, 5.1055, 4.8506, 4.8765, 4.5798, 4.3415, 4.7106, 5.1263], device='cuda:5'), covar=tensor([0.0453, 0.0567, 0.0783, 0.0368, 0.0448, 0.0662, 0.0456, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0278, 0.0383, 0.0334, 0.0249, 0.0247, 0.0254, 0.0304, 0.0263], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:18:31,951 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.657e+02 4.728e+02 5.816e+02 7.305e+02 1.524e+03, threshold=1.163e+03, percent-clipped=3.0 2023-04-28 00:18:50,560 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:55,865 INFO [train.py:904] (5/8) Epoch 3, batch 8250, loss[loss=0.2362, simple_loss=0.3248, pruned_loss=0.07378, over 16855.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.341, pruned_loss=0.1001, over 3087086.72 frames. ], batch size: 96, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:19:00,270 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28553.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:19:34,954 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2044, 3.2758, 1.5879, 3.2425, 2.3788, 3.3288, 1.7767, 2.5426], device='cuda:5'), covar=tensor([0.0086, 0.0147, 0.1302, 0.0046, 0.0644, 0.0285, 0.1312, 0.0539], device='cuda:5'), in_proj_covar=tensor([0.0088, 0.0131, 0.0168, 0.0077, 0.0154, 0.0158, 0.0177, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 00:19:51,783 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:16,990 INFO [train.py:904] (5/8) Epoch 3, batch 8300, loss[loss=0.2436, simple_loss=0.3154, pruned_loss=0.08588, over 12059.00 frames. ], tot_loss[loss=0.2644, simple_loss=0.3372, pruned_loss=0.09576, over 3079058.71 frames. ], batch size: 248, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:20:33,929 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:20:48,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0004, 3.8143, 3.9419, 2.8030, 3.7764, 1.4450, 3.5529, 3.6359], device='cuda:5'), covar=tensor([0.0109, 0.0081, 0.0100, 0.0494, 0.0099, 0.2270, 0.0118, 0.0248], device='cuda:5'), in_proj_covar=tensor([0.0073, 0.0061, 0.0096, 0.0105, 0.0071, 0.0122, 0.0083, 0.0093], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:21:14,649 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.491e+02 3.724e+02 4.538e+02 5.868e+02 1.330e+03, threshold=9.076e+02, percent-clipped=2.0 2023-04-28 00:21:27,077 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:39,715 INFO [train.py:904] (5/8) Epoch 3, batch 8350, loss[loss=0.2548, simple_loss=0.3376, pruned_loss=0.086, over 15432.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3352, pruned_loss=0.09265, over 3072687.29 frames. ], batch size: 191, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:22:41,973 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5563, 4.2666, 4.5497, 4.7815, 4.8634, 4.2684, 4.9201, 4.8512], device='cuda:5'), covar=tensor([0.0549, 0.0637, 0.0969, 0.0382, 0.0364, 0.0606, 0.0285, 0.0290], device='cuda:5'), in_proj_covar=tensor([0.0282, 0.0352, 0.0446, 0.0346, 0.0264, 0.0254, 0.0281, 0.0284], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:23:00,538 INFO [train.py:904] (5/8) Epoch 3, batch 8400, loss[loss=0.2255, simple_loss=0.3111, pruned_loss=0.07001, over 16711.00 frames. ], tot_loss[loss=0.2542, simple_loss=0.3307, pruned_loss=0.08888, over 3075657.34 frames. ], batch size: 124, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:23:18,395 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1519, 4.0428, 3.7513, 1.8921, 2.8673, 2.5348, 3.3572, 3.8454], device='cuda:5'), covar=tensor([0.0250, 0.0365, 0.0420, 0.1595, 0.0721, 0.0884, 0.0682, 0.0564], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0116, 0.0152, 0.0145, 0.0135, 0.0130, 0.0143, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 00:23:44,584 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0160, 2.6095, 2.4055, 3.4066, 3.1801, 3.3662, 1.7743, 2.8108], device='cuda:5'), covar=tensor([0.1113, 0.0355, 0.0861, 0.0071, 0.0180, 0.0290, 0.1065, 0.0584], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0134, 0.0161, 0.0070, 0.0134, 0.0143, 0.0154, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-04-28 00:23:58,231 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.191e+02 3.936e+02 4.649e+02 5.360e+02 9.184e+02, threshold=9.299e+02, percent-clipped=1.0 2023-04-28 00:24:07,906 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 00:24:20,124 INFO [train.py:904] (5/8) Epoch 3, batch 8450, loss[loss=0.2573, simple_loss=0.3236, pruned_loss=0.09547, over 12258.00 frames. ], tot_loss[loss=0.2511, simple_loss=0.3282, pruned_loss=0.08704, over 3047981.23 frames. ], batch size: 246, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:24:43,225 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.59 vs. limit=5.0 2023-04-28 00:25:39,680 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9619, 3.2820, 3.1883, 2.2032, 3.1762, 3.1918, 3.0458, 1.6984], device='cuda:5'), covar=tensor([0.0325, 0.0020, 0.0032, 0.0233, 0.0028, 0.0048, 0.0032, 0.0368], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0053, 0.0056, 0.0111, 0.0052, 0.0062, 0.0058, 0.0108], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 00:25:42,014 INFO [train.py:904] (5/8) Epoch 3, batch 8500, loss[loss=0.218, simple_loss=0.2976, pruned_loss=0.06919, over 16237.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3225, pruned_loss=0.08293, over 3048640.31 frames. ], batch size: 165, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:26:05,698 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5039, 4.5436, 4.5869, 4.6167, 4.6238, 5.1005, 4.9009, 4.5340], device='cuda:5'), covar=tensor([0.0735, 0.1215, 0.1111, 0.1263, 0.2003, 0.0805, 0.0831, 0.1928], device='cuda:5'), in_proj_covar=tensor([0.0224, 0.0312, 0.0288, 0.0267, 0.0345, 0.0321, 0.0255, 0.0358], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:26:40,696 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.668e+02 4.634e+02 5.862e+02 2.485e+03, threshold=9.268e+02, percent-clipped=6.0 2023-04-28 00:26:50,004 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:26:50,211 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1857, 2.3125, 1.8898, 2.1258, 2.7710, 2.6669, 3.2960, 3.0898], device='cuda:5'), covar=tensor([0.0018, 0.0115, 0.0151, 0.0150, 0.0075, 0.0098, 0.0036, 0.0051], device='cuda:5'), in_proj_covar=tensor([0.0054, 0.0116, 0.0117, 0.0120, 0.0109, 0.0118, 0.0074, 0.0094], device='cuda:5'), out_proj_covar=tensor([7.1183e-05, 1.6166e-04, 1.5924e-04, 1.6681e-04, 1.5521e-04, 1.6537e-04, 1.0251e-04, 1.3292e-04], device='cuda:5') 2023-04-28 00:27:01,014 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:05,709 INFO [train.py:904] (5/8) Epoch 3, batch 8550, loss[loss=0.2406, simple_loss=0.3248, pruned_loss=0.07823, over 15267.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3192, pruned_loss=0.0811, over 3034637.36 frames. ], batch size: 191, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:27:26,033 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:12,990 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:46,482 INFO [train.py:904] (5/8) Epoch 3, batch 8600, loss[loss=0.2158, simple_loss=0.313, pruned_loss=0.05928, over 16698.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3202, pruned_loss=0.08046, over 3035230.36 frames. ], batch size: 76, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:29:07,501 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 00:29:10,117 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-04-28 00:29:30,925 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:50,790 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:57,564 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.629e+02 3.755e+02 4.530e+02 5.746e+02 8.888e+02, threshold=9.061e+02, percent-clipped=0.0 2023-04-28 00:30:11,248 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:30:14,392 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-28 00:30:25,455 INFO [train.py:904] (5/8) Epoch 3, batch 8650, loss[loss=0.2052, simple_loss=0.2997, pruned_loss=0.05532, over 15310.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3174, pruned_loss=0.07805, over 3039093.46 frames. ], batch size: 190, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:30:45,827 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:31:52,872 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:12,032 INFO [train.py:904] (5/8) Epoch 3, batch 8700, loss[loss=0.2183, simple_loss=0.3012, pruned_loss=0.06774, over 16850.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3144, pruned_loss=0.07601, over 3053056.34 frames. ], batch size: 116, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:32:15,166 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:33:18,500 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 3.379e+02 3.929e+02 4.901e+02 7.493e+02, threshold=7.858e+02, percent-clipped=0.0 2023-04-28 00:33:26,052 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8292, 3.6180, 3.5975, 2.4428, 3.3145, 3.4742, 3.4265, 1.9060], device='cuda:5'), covar=tensor([0.0322, 0.0014, 0.0024, 0.0210, 0.0029, 0.0033, 0.0023, 0.0299], device='cuda:5'), in_proj_covar=tensor([0.0107, 0.0050, 0.0054, 0.0106, 0.0051, 0.0058, 0.0055, 0.0104], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 00:33:46,778 INFO [train.py:904] (5/8) Epoch 3, batch 8750, loss[loss=0.2371, simple_loss=0.3266, pruned_loss=0.07385, over 16418.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3135, pruned_loss=0.07534, over 3048097.12 frames. ], batch size: 68, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:34:19,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8111, 3.9189, 1.7183, 3.9087, 2.4957, 3.8281, 1.8969, 2.8537], device='cuda:5'), covar=tensor([0.0066, 0.0163, 0.1591, 0.0030, 0.0885, 0.0363, 0.1603, 0.0593], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0127, 0.0170, 0.0075, 0.0156, 0.0149, 0.0178, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 00:34:19,085 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:35:38,459 INFO [train.py:904] (5/8) Epoch 3, batch 8800, loss[loss=0.2616, simple_loss=0.3309, pruned_loss=0.09616, over 12436.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3111, pruned_loss=0.07324, over 3058174.32 frames. ], batch size: 247, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:36:52,046 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.420e+02 3.673e+02 4.669e+02 5.868e+02 1.100e+03, threshold=9.339e+02, percent-clipped=8.0 2023-04-28 00:37:04,042 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:17,017 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:21,276 INFO [train.py:904] (5/8) Epoch 3, batch 8850, loss[loss=0.2149, simple_loss=0.3125, pruned_loss=0.05859, over 15287.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3132, pruned_loss=0.07237, over 3067793.20 frames. ], batch size: 191, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:38:41,600 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4760, 4.2409, 3.9624, 4.6230, 4.7410, 4.2350, 4.7437, 4.6907], device='cuda:5'), covar=tensor([0.0555, 0.0544, 0.1665, 0.0656, 0.0625, 0.0589, 0.0611, 0.0628], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0346, 0.0437, 0.0342, 0.0260, 0.0250, 0.0275, 0.0282], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:38:45,754 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:38:57,575 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:39:06,873 INFO [train.py:904] (5/8) Epoch 3, batch 8900, loss[loss=0.2227, simple_loss=0.3089, pruned_loss=0.06824, over 16437.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.313, pruned_loss=0.07127, over 3077527.70 frames. ], batch size: 147, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:39:40,945 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:00,168 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7188, 3.5749, 3.7258, 3.9812, 3.9909, 3.5445, 3.9727, 3.9966], device='cuda:5'), covar=tensor([0.0614, 0.0601, 0.1102, 0.0397, 0.0393, 0.1051, 0.0454, 0.0360], device='cuda:5'), in_proj_covar=tensor([0.0282, 0.0348, 0.0438, 0.0339, 0.0257, 0.0249, 0.0275, 0.0281], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:40:35,449 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29236.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:36,247 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 3.766e+02 4.506e+02 5.745e+02 1.186e+03, threshold=9.012e+02, percent-clipped=1.0 2023-04-28 00:41:11,861 INFO [train.py:904] (5/8) Epoch 3, batch 8950, loss[loss=0.239, simple_loss=0.3108, pruned_loss=0.08364, over 12843.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3137, pruned_loss=0.07242, over 3073454.08 frames. ], batch size: 248, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:42:51,678 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-28 00:42:54,185 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:42:58,070 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8762, 2.7438, 2.7110, 1.7664, 2.9115, 2.8732, 2.5361, 2.4266], device='cuda:5'), covar=tensor([0.0712, 0.0124, 0.0137, 0.0979, 0.0070, 0.0083, 0.0289, 0.0356], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0081, 0.0076, 0.0145, 0.0069, 0.0070, 0.0110, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 00:43:00,676 INFO [train.py:904] (5/8) Epoch 3, batch 9000, loss[loss=0.238, simple_loss=0.3071, pruned_loss=0.08447, over 11866.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3091, pruned_loss=0.06985, over 3077417.61 frames. ], batch size: 248, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:43:00,676 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 00:43:11,839 INFO [train.py:938] (5/8) Epoch 3, validation: loss=0.1886, simple_loss=0.2904, pruned_loss=0.04341, over 944034.00 frames. 2023-04-28 00:43:11,840 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 00:44:05,779 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4042, 4.3643, 1.7406, 4.4759, 2.9650, 4.4182, 2.0615, 3.3309], device='cuda:5'), covar=tensor([0.0039, 0.0120, 0.1655, 0.0022, 0.0649, 0.0265, 0.1480, 0.0477], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0126, 0.0173, 0.0077, 0.0153, 0.0152, 0.0181, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 00:44:26,935 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.245e+02 3.595e+02 4.170e+02 5.160e+02 8.461e+02, threshold=8.339e+02, percent-clipped=0.0 2023-04-28 00:44:57,088 INFO [train.py:904] (5/8) Epoch 3, batch 9050, loss[loss=0.2023, simple_loss=0.286, pruned_loss=0.05928, over 16762.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3102, pruned_loss=0.07077, over 3089968.86 frames. ], batch size: 76, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:45:12,883 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:45:25,802 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-04-28 00:46:12,771 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3088, 3.1484, 3.0937, 3.4416, 3.4670, 3.1293, 3.3892, 3.4382], device='cuda:5'), covar=tensor([0.0586, 0.0666, 0.1433, 0.0566, 0.0550, 0.2078, 0.0839, 0.0640], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0354, 0.0444, 0.0348, 0.0261, 0.0249, 0.0277, 0.0287], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:46:41,635 INFO [train.py:904] (5/8) Epoch 3, batch 9100, loss[loss=0.2391, simple_loss=0.3239, pruned_loss=0.07713, over 16705.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.311, pruned_loss=0.07224, over 3080834.34 frames. ], batch size: 134, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:48:08,510 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.097e+02 4.371e+02 5.293e+02 6.986e+02 1.215e+03, threshold=1.059e+03, percent-clipped=12.0 2023-04-28 00:48:40,651 INFO [train.py:904] (5/8) Epoch 3, batch 9150, loss[loss=0.2211, simple_loss=0.3026, pruned_loss=0.06977, over 16551.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3111, pruned_loss=0.07174, over 3071370.77 frames. ], batch size: 68, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:49:44,915 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 00:50:24,287 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1134, 2.0581, 1.6144, 1.9039, 2.7461, 2.4224, 3.1469, 2.9315], device='cuda:5'), covar=tensor([0.0016, 0.0167, 0.0227, 0.0187, 0.0093, 0.0138, 0.0033, 0.0070], device='cuda:5'), in_proj_covar=tensor([0.0053, 0.0121, 0.0122, 0.0123, 0.0112, 0.0121, 0.0075, 0.0097], device='cuda:5'), out_proj_covar=tensor([6.7971e-05, 1.6846e-04, 1.6484e-04, 1.6822e-04, 1.5671e-04, 1.6929e-04, 1.0163e-04, 1.3634e-04], device='cuda:5') 2023-04-28 00:50:24,880 INFO [train.py:904] (5/8) Epoch 3, batch 9200, loss[loss=0.2086, simple_loss=0.2868, pruned_loss=0.06518, over 16707.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3058, pruned_loss=0.06994, over 3082735.83 frames. ], batch size: 76, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:50:54,998 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:32,424 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.365e+02 4.043e+02 4.899e+02 6.737e+02 1.438e+03, threshold=9.797e+02, percent-clipped=1.0 2023-04-28 00:51:55,107 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4531, 3.4125, 3.3849, 3.0928, 3.3081, 2.0710, 3.2027, 3.0456], device='cuda:5'), covar=tensor([0.0080, 0.0057, 0.0083, 0.0184, 0.0063, 0.1300, 0.0092, 0.0109], device='cuda:5'), in_proj_covar=tensor([0.0071, 0.0060, 0.0097, 0.0097, 0.0070, 0.0123, 0.0084, 0.0091], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 00:52:01,393 INFO [train.py:904] (5/8) Epoch 3, batch 9250, loss[loss=0.2159, simple_loss=0.306, pruned_loss=0.06288, over 15493.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.306, pruned_loss=0.07078, over 3051688.40 frames. ], batch size: 191, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:52:02,567 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:29,837 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:59,577 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9232, 2.9447, 2.4480, 3.8891, 3.7288, 3.7851, 1.5886, 2.8509], device='cuda:5'), covar=tensor([0.1256, 0.0428, 0.1043, 0.0065, 0.0194, 0.0301, 0.1303, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0131, 0.0159, 0.0068, 0.0129, 0.0140, 0.0152, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-04-28 00:53:31,912 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:53:50,325 INFO [train.py:904] (5/8) Epoch 3, batch 9300, loss[loss=0.206, simple_loss=0.292, pruned_loss=0.05999, over 15449.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3044, pruned_loss=0.06993, over 3049422.89 frames. ], batch size: 191, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:54:02,331 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9538, 3.6390, 2.9360, 4.7636, 4.6665, 4.6276, 2.1476, 3.3650], device='cuda:5'), covar=tensor([0.1615, 0.0465, 0.1053, 0.0094, 0.0162, 0.0231, 0.1433, 0.0680], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0132, 0.0161, 0.0068, 0.0129, 0.0141, 0.0154, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-04-28 00:54:16,021 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:55:02,829 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 00:55:11,309 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.186e+02 3.534e+02 4.207e+02 4.979e+02 1.086e+03, threshold=8.414e+02, percent-clipped=1.0 2023-04-28 00:55:35,085 INFO [train.py:904] (5/8) Epoch 3, batch 9350, loss[loss=0.2545, simple_loss=0.335, pruned_loss=0.08699, over 16243.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.304, pruned_loss=0.06991, over 3034531.07 frames. ], batch size: 165, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:55:51,520 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:56:00,152 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:57:16,972 INFO [train.py:904] (5/8) Epoch 3, batch 9400, loss[loss=0.1986, simple_loss=0.277, pruned_loss=0.06007, over 12295.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.304, pruned_loss=0.06958, over 3031214.33 frames. ], batch size: 250, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:57:24,330 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9637, 2.8104, 2.6489, 1.6958, 2.8209, 2.8155, 2.5674, 2.3962], device='cuda:5'), covar=tensor([0.0703, 0.0097, 0.0116, 0.0974, 0.0091, 0.0088, 0.0319, 0.0373], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0077, 0.0073, 0.0140, 0.0069, 0.0068, 0.0107, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 00:57:27,968 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:01,721 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:58:32,712 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.403e+02 3.993e+02 5.001e+02 6.201e+02 1.324e+03, threshold=1.000e+03, percent-clipped=7.0 2023-04-28 00:58:58,200 INFO [train.py:904] (5/8) Epoch 3, batch 9450, loss[loss=0.2277, simple_loss=0.3117, pruned_loss=0.07191, over 15242.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.306, pruned_loss=0.06992, over 3039139.32 frames. ], batch size: 191, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:59:07,619 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8110, 1.2261, 1.5318, 1.6908, 1.7381, 1.8089, 1.3177, 1.7851], device='cuda:5'), covar=tensor([0.0067, 0.0177, 0.0106, 0.0105, 0.0097, 0.0069, 0.0152, 0.0056], device='cuda:5'), in_proj_covar=tensor([0.0092, 0.0126, 0.0112, 0.0102, 0.0099, 0.0072, 0.0117, 0.0064], device='cuda:5'), out_proj_covar=tensor([1.3732e-04, 1.9257e-04, 1.7541e-04, 1.5796e-04, 1.5104e-04, 1.0728e-04, 1.7746e-04, 9.6519e-05], device='cuda:5') 2023-04-28 00:59:39,717 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4827, 3.7111, 3.9292, 2.9398, 3.6890, 3.8590, 3.9299, 2.4949], device='cuda:5'), covar=tensor([0.0270, 0.0017, 0.0024, 0.0179, 0.0031, 0.0027, 0.0016, 0.0249], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0052, 0.0055, 0.0109, 0.0053, 0.0059, 0.0055, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 01:00:37,296 INFO [train.py:904] (5/8) Epoch 3, batch 9500, loss[loss=0.2076, simple_loss=0.2979, pruned_loss=0.05868, over 16828.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3043, pruned_loss=0.06879, over 3044011.95 frames. ], batch size: 83, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 01:00:41,506 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:00:59,316 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0283, 2.6295, 2.3670, 3.3568, 3.1383, 3.4023, 1.7203, 2.7766], device='cuda:5'), covar=tensor([0.1055, 0.0384, 0.0911, 0.0080, 0.0199, 0.0319, 0.1108, 0.0595], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0130, 0.0159, 0.0066, 0.0128, 0.0139, 0.0151, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-28 01:01:15,166 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2927, 3.2244, 3.2984, 3.4918, 3.5092, 3.1288, 3.4927, 3.5330], device='cuda:5'), covar=tensor([0.0570, 0.0513, 0.0982, 0.0418, 0.0381, 0.1444, 0.0480, 0.0354], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0344, 0.0436, 0.0338, 0.0258, 0.0237, 0.0275, 0.0277], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 01:01:15,353 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9298, 2.2750, 2.1837, 3.1127, 1.9165, 2.9444, 2.2604, 1.9488], device='cuda:5'), covar=tensor([0.0379, 0.0974, 0.0553, 0.0267, 0.2011, 0.0331, 0.1085, 0.1611], device='cuda:5'), in_proj_covar=tensor([0.0259, 0.0246, 0.0205, 0.0263, 0.0310, 0.0216, 0.0234, 0.0286], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 01:01:50,898 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.435e+02 3.682e+02 4.369e+02 5.544e+02 1.182e+03, threshold=8.738e+02, percent-clipped=2.0 2023-04-28 01:02:22,330 INFO [train.py:904] (5/8) Epoch 3, batch 9550, loss[loss=0.2307, simple_loss=0.3086, pruned_loss=0.07639, over 16522.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3038, pruned_loss=0.06841, over 3057783.06 frames. ], batch size: 75, lr: 1.89e-02, grad_scale: 4.0 2023-04-28 01:02:49,814 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:03:05,327 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 01:03:48,949 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:04,160 INFO [train.py:904] (5/8) Epoch 3, batch 9600, loss[loss=0.276, simple_loss=0.3541, pruned_loss=0.09891, over 16384.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3061, pruned_loss=0.06988, over 3036495.08 frames. ], batch size: 146, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:04:16,862 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:18,578 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.324e+02 3.589e+02 4.321e+02 5.133e+02 1.002e+03, threshold=8.641e+02, percent-clipped=3.0 2023-04-28 01:05:23,702 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:51,218 INFO [train.py:904] (5/8) Epoch 3, batch 9650, loss[loss=0.231, simple_loss=0.3048, pruned_loss=0.07864, over 12369.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3088, pruned_loss=0.07073, over 3038923.30 frames. ], batch size: 247, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:07:41,653 INFO [train.py:904] (5/8) Epoch 3, batch 9700, loss[loss=0.2194, simple_loss=0.2978, pruned_loss=0.07047, over 12162.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3073, pruned_loss=0.06999, over 3045769.33 frames. ], batch size: 248, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:08:16,063 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:08:59,980 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.947e+02 3.446e+02 4.381e+02 5.437e+02 9.929e+02, threshold=8.763e+02, percent-clipped=3.0 2023-04-28 01:09:24,296 INFO [train.py:904] (5/8) Epoch 3, batch 9750, loss[loss=0.211, simple_loss=0.2842, pruned_loss=0.06887, over 12249.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3062, pruned_loss=0.07032, over 3041627.27 frames. ], batch size: 247, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:09:47,256 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 01:09:51,601 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4532, 2.9849, 2.7021, 2.3217, 2.1756, 2.0409, 3.0645, 3.1476], device='cuda:5'), covar=tensor([0.1453, 0.0584, 0.0903, 0.0953, 0.1676, 0.1250, 0.0337, 0.0433], device='cuda:5'), in_proj_covar=tensor([0.0259, 0.0236, 0.0254, 0.0212, 0.0225, 0.0192, 0.0210, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 01:11:02,299 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1383, 1.7019, 2.1497, 3.0211, 2.9518, 3.1035, 2.0358, 3.2186], device='cuda:5'), covar=tensor([0.0042, 0.0217, 0.0159, 0.0072, 0.0060, 0.0074, 0.0168, 0.0042], device='cuda:5'), in_proj_covar=tensor([0.0093, 0.0125, 0.0111, 0.0101, 0.0097, 0.0073, 0.0116, 0.0063], device='cuda:5'), out_proj_covar=tensor([1.3806e-04, 1.9061e-04, 1.7305e-04, 1.5652e-04, 1.4704e-04, 1.0728e-04, 1.7564e-04, 9.3993e-05], device='cuda:5') 2023-04-28 01:11:02,945 INFO [train.py:904] (5/8) Epoch 3, batch 9800, loss[loss=0.2035, simple_loss=0.2998, pruned_loss=0.05362, over 16555.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3056, pruned_loss=0.06884, over 3048585.13 frames. ], batch size: 57, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:11:48,282 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:12:14,774 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.212e+02 3.552e+02 4.452e+02 5.972e+02 1.192e+03, threshold=8.904e+02, percent-clipped=4.0 2023-04-28 01:12:47,470 INFO [train.py:904] (5/8) Epoch 3, batch 9850, loss[loss=0.2169, simple_loss=0.2902, pruned_loss=0.07183, over 12369.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3067, pruned_loss=0.06862, over 3037664.77 frames. ], batch size: 248, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:13:02,429 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:06,748 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:39,083 INFO [train.py:904] (5/8) Epoch 3, batch 9900, loss[loss=0.225, simple_loss=0.3171, pruned_loss=0.06649, over 16391.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3076, pruned_loss=0.06869, over 3036983.46 frames. ], batch size: 146, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:14:54,559 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:59,650 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:16:05,828 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.521e+02 3.778e+02 4.995e+02 6.247e+02 1.716e+03, threshold=9.990e+02, percent-clipped=10.0 2023-04-28 01:16:11,187 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 01:16:35,578 INFO [train.py:904] (5/8) Epoch 3, batch 9950, loss[loss=0.247, simple_loss=0.3184, pruned_loss=0.08775, over 12644.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3094, pruned_loss=0.06965, over 3020002.05 frames. ], batch size: 248, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:16:47,151 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:17:23,161 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:18:37,076 INFO [train.py:904] (5/8) Epoch 3, batch 10000, loss[loss=0.2202, simple_loss=0.3094, pruned_loss=0.06553, over 16751.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3069, pruned_loss=0.06792, over 3062110.85 frames. ], batch size: 134, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:18:42,291 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6620, 2.6083, 2.5735, 1.8382, 2.4797, 2.5390, 2.5781, 1.6333], device='cuda:5'), covar=tensor([0.0300, 0.0025, 0.0045, 0.0197, 0.0041, 0.0041, 0.0033, 0.0300], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0053, 0.0056, 0.0109, 0.0055, 0.0060, 0.0056, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 01:18:42,657 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 01:19:12,402 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:19:47,501 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-28 01:19:55,685 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.113e+02 3.684e+02 4.714e+02 5.759e+02 1.067e+03, threshold=9.428e+02, percent-clipped=2.0 2023-04-28 01:20:19,880 INFO [train.py:904] (5/8) Epoch 3, batch 10050, loss[loss=0.2113, simple_loss=0.3007, pruned_loss=0.06089, over 16731.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3069, pruned_loss=0.06735, over 3072775.62 frames. ], batch size: 76, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:20:38,348 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:20:50,750 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:29,343 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0499, 4.1014, 3.8901, 3.9517, 3.5566, 4.0000, 3.8354, 3.7706], device='cuda:5'), covar=tensor([0.0366, 0.0206, 0.0175, 0.0130, 0.0591, 0.0248, 0.0407, 0.0309], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0124, 0.0162, 0.0134, 0.0188, 0.0151, 0.0115, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 01:21:54,401 INFO [train.py:904] (5/8) Epoch 3, batch 10100, loss[loss=0.2143, simple_loss=0.2942, pruned_loss=0.0672, over 16650.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3077, pruned_loss=0.06825, over 3071745.82 frames. ], batch size: 134, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:22:37,222 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:22:55,013 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 01:22:56,720 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3812, 4.0443, 4.0122, 1.8705, 4.2777, 4.2726, 3.0290, 3.2914], device='cuda:5'), covar=tensor([0.0827, 0.0095, 0.0147, 0.1258, 0.0041, 0.0042, 0.0375, 0.0369], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0082, 0.0079, 0.0149, 0.0073, 0.0072, 0.0114, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 01:23:00,424 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.870e+02 3.919e+02 4.872e+02 6.072e+02 1.469e+03, threshold=9.744e+02, percent-clipped=6.0 2023-04-28 01:23:38,635 INFO [train.py:904] (5/8) Epoch 4, batch 0, loss[loss=0.2446, simple_loss=0.314, pruned_loss=0.08757, over 15800.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.314, pruned_loss=0.08757, over 15800.00 frames. ], batch size: 35, lr: 1.75e-02, grad_scale: 8.0 2023-04-28 01:23:38,635 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 01:23:46,516 INFO [train.py:938] (5/8) Epoch 4, validation: loss=0.188, simple_loss=0.2904, pruned_loss=0.04284, over 944034.00 frames. 2023-04-28 01:23:46,518 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 01:23:57,015 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7277, 4.4333, 4.1139, 2.0080, 3.2822, 2.5944, 3.9182, 4.3869], device='cuda:5'), covar=tensor([0.0199, 0.0432, 0.0416, 0.1499, 0.0606, 0.0934, 0.0622, 0.0588], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0109, 0.0152, 0.0143, 0.0133, 0.0129, 0.0136, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 01:23:58,009 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:12,762 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1895, 4.6474, 4.6222, 2.5491, 4.9509, 4.9688, 3.6390, 3.8848], device='cuda:5'), covar=tensor([0.0596, 0.0090, 0.0106, 0.1028, 0.0030, 0.0025, 0.0255, 0.0311], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0081, 0.0079, 0.0147, 0.0072, 0.0072, 0.0112, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 01:24:18,679 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 01:24:29,165 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:56,028 INFO [train.py:904] (5/8) Epoch 4, batch 50, loss[loss=0.221, simple_loss=0.2877, pruned_loss=0.07712, over 16787.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3315, pruned_loss=0.1007, over 746646.03 frames. ], batch size: 39, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:25:02,246 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:49,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2357, 4.3956, 2.2691, 4.8112, 2.8293, 4.6568, 2.1053, 3.2016], device='cuda:5'), covar=tensor([0.0098, 0.0241, 0.1496, 0.0024, 0.0780, 0.0270, 0.1585, 0.0620], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0138, 0.0173, 0.0081, 0.0160, 0.0162, 0.0181, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 01:25:49,880 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.617e+02 4.176e+02 5.221e+02 6.159e+02 1.340e+03, threshold=1.044e+03, percent-clipped=3.0 2023-04-28 01:26:02,171 INFO [train.py:904] (5/8) Epoch 4, batch 100, loss[loss=0.2628, simple_loss=0.3235, pruned_loss=0.101, over 16475.00 frames. ], tot_loss[loss=0.2578, simple_loss=0.3256, pruned_loss=0.09495, over 1318963.40 frames. ], batch size: 75, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:26:23,720 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:27:11,709 INFO [train.py:904] (5/8) Epoch 4, batch 150, loss[loss=0.2348, simple_loss=0.3073, pruned_loss=0.08118, over 17173.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3224, pruned_loss=0.09409, over 1748892.41 frames. ], batch size: 46, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:28:04,892 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 3.830e+02 4.808e+02 5.813e+02 1.621e+03, threshold=9.616e+02, percent-clipped=3.0 2023-04-28 01:28:19,174 INFO [train.py:904] (5/8) Epoch 4, batch 200, loss[loss=0.2467, simple_loss=0.3113, pruned_loss=0.091, over 16811.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3212, pruned_loss=0.0935, over 2096763.31 frames. ], batch size: 102, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:28:23,995 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:19,566 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-04-28 01:29:26,989 INFO [train.py:904] (5/8) Epoch 4, batch 250, loss[loss=0.2602, simple_loss=0.3169, pruned_loss=0.1018, over 16417.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3177, pruned_loss=0.09055, over 2366331.86 frames. ], batch size: 68, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:29:48,588 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:48,754 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:30:21,644 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.145e+02 3.686e+02 4.713e+02 5.844e+02 8.594e+02, threshold=9.427e+02, percent-clipped=0.0 2023-04-28 01:30:28,389 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 01:30:36,166 INFO [train.py:904] (5/8) Epoch 4, batch 300, loss[loss=0.2561, simple_loss=0.3087, pruned_loss=0.1017, over 12311.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3134, pruned_loss=0.08798, over 2575804.13 frames. ], batch size: 248, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:31:16,645 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:31:43,239 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 01:31:43,572 INFO [train.py:904] (5/8) Epoch 4, batch 350, loss[loss=0.2162, simple_loss=0.2953, pruned_loss=0.06857, over 17099.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3098, pruned_loss=0.08482, over 2744041.23 frames. ], batch size: 47, lr: 1.74e-02, grad_scale: 1.0 2023-04-28 01:32:20,667 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30829.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:38,099 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0459, 2.4925, 2.4115, 4.4673, 1.9352, 3.9411, 2.2270, 2.3575], device='cuda:5'), covar=tensor([0.0350, 0.1220, 0.0646, 0.0195, 0.2315, 0.0381, 0.1310, 0.2093], device='cuda:5'), in_proj_covar=tensor([0.0276, 0.0264, 0.0218, 0.0279, 0.0326, 0.0233, 0.0245, 0.0327], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 01:32:38,673 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.414e+02 3.784e+02 4.816e+02 5.903e+02 1.455e+03, threshold=9.632e+02, percent-clipped=4.0 2023-04-28 01:32:51,067 INFO [train.py:904] (5/8) Epoch 4, batch 400, loss[loss=0.2735, simple_loss=0.3235, pruned_loss=0.1118, over 16880.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3073, pruned_loss=0.08288, over 2873558.26 frames. ], batch size: 116, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:33:11,946 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:33:47,455 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 01:34:01,541 INFO [train.py:904] (5/8) Epoch 4, batch 450, loss[loss=0.2405, simple_loss=0.2958, pruned_loss=0.09263, over 16415.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3051, pruned_loss=0.08197, over 2962970.23 frames. ], batch size: 146, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:34:02,004 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6840, 4.6052, 4.5214, 4.4723, 4.0220, 4.5573, 4.5505, 4.3406], device='cuda:5'), covar=tensor([0.0373, 0.0259, 0.0175, 0.0149, 0.0744, 0.0210, 0.0252, 0.0340], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0151, 0.0196, 0.0163, 0.0231, 0.0184, 0.0138, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 01:34:05,008 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30903.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:18,183 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:35,359 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:45,917 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:56,473 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.382e+02 3.455e+02 4.330e+02 5.569e+02 1.727e+03, threshold=8.660e+02, percent-clipped=4.0 2023-04-28 01:35:00,030 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5968, 2.4584, 2.3417, 3.9411, 1.9261, 3.4132, 2.1482, 2.2878], device='cuda:5'), covar=tensor([0.0389, 0.1010, 0.0628, 0.0252, 0.1987, 0.0415, 0.1291, 0.1530], device='cuda:5'), in_proj_covar=tensor([0.0278, 0.0264, 0.0216, 0.0279, 0.0326, 0.0234, 0.0244, 0.0329], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 01:35:00,186 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-28 01:35:08,471 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8164, 5.4401, 5.3427, 5.3762, 5.2353, 5.8852, 5.6594, 5.3274], device='cuda:5'), covar=tensor([0.0699, 0.1394, 0.1301, 0.1618, 0.2679, 0.1045, 0.0863, 0.1916], device='cuda:5'), in_proj_covar=tensor([0.0242, 0.0347, 0.0319, 0.0289, 0.0387, 0.0351, 0.0274, 0.0394], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 01:35:09,303 INFO [train.py:904] (5/8) Epoch 4, batch 500, loss[loss=0.2323, simple_loss=0.3095, pruned_loss=0.07754, over 17088.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.304, pruned_loss=0.08143, over 3031963.30 frames. ], batch size: 53, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:35:28,329 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:37,234 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 01:35:59,014 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:09,097 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:17,825 INFO [train.py:904] (5/8) Epoch 4, batch 550, loss[loss=0.23, simple_loss=0.2958, pruned_loss=0.08212, over 15545.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3032, pruned_loss=0.08075, over 3101489.64 frames. ], batch size: 190, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:36:33,159 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:36:33,263 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:36:39,375 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:13,483 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.455e+02 3.696e+02 4.486e+02 5.421e+02 1.042e+03, threshold=8.971e+02, percent-clipped=3.0 2023-04-28 01:37:28,717 INFO [train.py:904] (5/8) Epoch 4, batch 600, loss[loss=0.2194, simple_loss=0.3084, pruned_loss=0.06523, over 17262.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3027, pruned_loss=0.08137, over 3150799.38 frames. ], batch size: 52, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:37:46,830 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:58,252 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:38:36,823 INFO [train.py:904] (5/8) Epoch 4, batch 650, loss[loss=0.2293, simple_loss=0.2946, pruned_loss=0.08207, over 16756.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3009, pruned_loss=0.08076, over 3185576.04 frames. ], batch size: 124, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:38:46,561 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6331, 4.1497, 4.3439, 3.5501, 4.1586, 4.4109, 4.2120, 2.6257], device='cuda:5'), covar=tensor([0.0289, 0.0025, 0.0028, 0.0160, 0.0029, 0.0044, 0.0025, 0.0256], device='cuda:5'), in_proj_covar=tensor([0.0110, 0.0057, 0.0058, 0.0110, 0.0057, 0.0063, 0.0056, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 01:39:30,540 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 3.618e+02 4.311e+02 5.742e+02 1.399e+03, threshold=8.622e+02, percent-clipped=6.0 2023-04-28 01:39:45,330 INFO [train.py:904] (5/8) Epoch 4, batch 700, loss[loss=0.2135, simple_loss=0.29, pruned_loss=0.06851, over 16580.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.2994, pruned_loss=0.07934, over 3215808.04 frames. ], batch size: 75, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:39:55,109 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 01:40:53,739 INFO [train.py:904] (5/8) Epoch 4, batch 750, loss[loss=0.1834, simple_loss=0.2584, pruned_loss=0.05423, over 16980.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.2997, pruned_loss=0.0788, over 3241547.40 frames. ], batch size: 41, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:40:55,439 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5120, 4.3255, 3.7748, 1.8035, 2.8424, 2.2974, 3.7278, 4.1364], device='cuda:5'), covar=tensor([0.0250, 0.0472, 0.0492, 0.1689, 0.0802, 0.1087, 0.0573, 0.0681], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0126, 0.0156, 0.0144, 0.0136, 0.0128, 0.0139, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 01:41:14,008 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3246, 1.7218, 2.3458, 2.9994, 2.8990, 3.3425, 1.9975, 3.1965], device='cuda:5'), covar=tensor([0.0055, 0.0204, 0.0126, 0.0084, 0.0074, 0.0065, 0.0157, 0.0060], device='cuda:5'), in_proj_covar=tensor([0.0105, 0.0133, 0.0119, 0.0113, 0.0108, 0.0083, 0.0126, 0.0070], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 01:41:31,299 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2733, 5.2146, 5.0691, 5.1047, 4.5205, 5.1196, 5.1936, 4.7112], device='cuda:5'), covar=tensor([0.0359, 0.0172, 0.0166, 0.0113, 0.0853, 0.0201, 0.0186, 0.0324], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0153, 0.0200, 0.0165, 0.0237, 0.0187, 0.0143, 0.0203], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 01:41:48,309 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.232e+02 3.368e+02 4.224e+02 5.033e+02 8.537e+02, threshold=8.448e+02, percent-clipped=0.0 2023-04-28 01:42:01,639 INFO [train.py:904] (5/8) Epoch 4, batch 800, loss[loss=0.1921, simple_loss=0.2774, pruned_loss=0.05341, over 17173.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.2995, pruned_loss=0.07881, over 3263262.79 frames. ], batch size: 46, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:42:02,639 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5983, 4.6615, 5.1587, 5.2010, 5.1603, 4.7449, 4.7852, 4.4633], device='cuda:5'), covar=tensor([0.0261, 0.0380, 0.0335, 0.0343, 0.0325, 0.0268, 0.0645, 0.0393], device='cuda:5'), in_proj_covar=tensor([0.0210, 0.0209, 0.0215, 0.0212, 0.0251, 0.0225, 0.0316, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 01:42:14,293 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:44,521 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:56,258 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:57,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8111, 3.7057, 2.9747, 5.0622, 4.8198, 4.3241, 1.9188, 3.5888], device='cuda:5'), covar=tensor([0.1338, 0.0427, 0.0986, 0.0080, 0.0235, 0.0334, 0.1239, 0.0569], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0136, 0.0165, 0.0072, 0.0159, 0.0154, 0.0155, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 01:43:11,605 INFO [train.py:904] (5/8) Epoch 4, batch 850, loss[loss=0.1983, simple_loss=0.2767, pruned_loss=0.05994, over 17006.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.2988, pruned_loss=0.07774, over 3277419.01 frames. ], batch size: 41, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:43:24,667 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:44:07,365 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.752e+02 3.699e+02 4.562e+02 5.706e+02 1.340e+03, threshold=9.123e+02, percent-clipped=5.0 2023-04-28 01:44:19,713 INFO [train.py:904] (5/8) Epoch 4, batch 900, loss[loss=0.2197, simple_loss=0.3029, pruned_loss=0.06829, over 16602.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2974, pruned_loss=0.07738, over 3278941.71 frames. ], batch size: 57, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:44:32,830 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:44:42,772 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:45:31,312 INFO [train.py:904] (5/8) Epoch 4, batch 950, loss[loss=0.2293, simple_loss=0.2906, pruned_loss=0.08398, over 16783.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2977, pruned_loss=0.0773, over 3278844.72 frames. ], batch size: 124, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:45:41,805 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:45:54,424 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9830, 3.8174, 2.8816, 5.3025, 5.0549, 4.4946, 2.1516, 3.4609], device='cuda:5'), covar=tensor([0.1335, 0.0439, 0.1109, 0.0070, 0.0176, 0.0272, 0.1206, 0.0603], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0137, 0.0166, 0.0074, 0.0161, 0.0155, 0.0155, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 01:46:26,070 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.766e+02 3.401e+02 4.040e+02 4.698e+02 9.697e+02, threshold=8.080e+02, percent-clipped=2.0 2023-04-28 01:46:38,675 INFO [train.py:904] (5/8) Epoch 4, batch 1000, loss[loss=0.2205, simple_loss=0.2848, pruned_loss=0.07808, over 15418.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.2962, pruned_loss=0.07721, over 3280705.45 frames. ], batch size: 191, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:47:05,153 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0730, 1.3424, 1.8566, 2.1770, 2.2413, 2.2736, 1.4252, 2.1541], device='cuda:5'), covar=tensor([0.0068, 0.0196, 0.0099, 0.0106, 0.0069, 0.0068, 0.0161, 0.0041], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0128, 0.0114, 0.0109, 0.0103, 0.0079, 0.0122, 0.0067], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 01:47:06,264 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:36,264 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:47:42,086 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6151, 1.5637, 2.0956, 2.5111, 2.5866, 2.6373, 1.5497, 2.7219], device='cuda:5'), covar=tensor([0.0054, 0.0191, 0.0131, 0.0099, 0.0065, 0.0077, 0.0175, 0.0041], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0129, 0.0113, 0.0109, 0.0104, 0.0079, 0.0122, 0.0067], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 01:47:49,010 INFO [train.py:904] (5/8) Epoch 4, batch 1050, loss[loss=0.209, simple_loss=0.2822, pruned_loss=0.06785, over 17023.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.2956, pruned_loss=0.07634, over 3300190.01 frames. ], batch size: 41, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:48:00,198 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9181, 5.2459, 5.2681, 5.3092, 5.1932, 5.8054, 5.5708, 5.2500], device='cuda:5'), covar=tensor([0.0756, 0.1462, 0.1290, 0.1435, 0.2493, 0.0808, 0.0777, 0.1795], device='cuda:5'), in_proj_covar=tensor([0.0256, 0.0363, 0.0335, 0.0304, 0.0409, 0.0364, 0.0282, 0.0408], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 01:48:32,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0231, 4.7774, 4.8740, 5.3014, 5.3865, 4.6648, 5.3905, 5.3221], device='cuda:5'), covar=tensor([0.0670, 0.0609, 0.1330, 0.0408, 0.0355, 0.0554, 0.0374, 0.0312], device='cuda:5'), in_proj_covar=tensor([0.0351, 0.0421, 0.0566, 0.0437, 0.0328, 0.0314, 0.0342, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 01:48:37,523 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5628, 4.6075, 5.2311, 5.2078, 5.1521, 4.7364, 4.7310, 4.4822], device='cuda:5'), covar=tensor([0.0247, 0.0341, 0.0317, 0.0391, 0.0342, 0.0283, 0.0609, 0.0297], device='cuda:5'), in_proj_covar=tensor([0.0212, 0.0211, 0.0219, 0.0216, 0.0255, 0.0231, 0.0322, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 01:48:45,343 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 3.437e+02 4.188e+02 5.032e+02 1.579e+03, threshold=8.377e+02, percent-clipped=2.0 2023-04-28 01:49:00,854 INFO [train.py:904] (5/8) Epoch 4, batch 1100, loss[loss=0.2602, simple_loss=0.3172, pruned_loss=0.1016, over 12317.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2948, pruned_loss=0.0753, over 3303710.07 frames. ], batch size: 248, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:49:02,435 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:49:11,910 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:41,948 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:55,694 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31590.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:09,618 INFO [train.py:904] (5/8) Epoch 4, batch 1150, loss[loss=0.1641, simple_loss=0.2486, pruned_loss=0.03984, over 16968.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2937, pruned_loss=0.07393, over 3316925.57 frames. ], batch size: 41, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:50:18,999 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:22,336 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3437, 2.2496, 2.2128, 3.6823, 1.8619, 3.2917, 2.1987, 2.2101], device='cuda:5'), covar=tensor([0.0433, 0.1182, 0.0661, 0.0254, 0.2040, 0.0404, 0.1277, 0.1589], device='cuda:5'), in_proj_covar=tensor([0.0287, 0.0270, 0.0222, 0.0286, 0.0332, 0.0243, 0.0250, 0.0340], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 01:50:36,033 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6925, 4.4153, 4.4101, 2.0701, 3.0024, 2.5632, 3.9192, 4.2625], device='cuda:5'), covar=tensor([0.0234, 0.0390, 0.0313, 0.1502, 0.0718, 0.0926, 0.0566, 0.0732], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0127, 0.0153, 0.0143, 0.0135, 0.0128, 0.0141, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 01:50:49,047 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:51,386 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 01:51:00,571 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:04,259 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.256e+02 3.893e+02 4.649e+02 5.891e+02 1.405e+03, threshold=9.297e+02, percent-clipped=5.0 2023-04-28 01:51:19,080 INFO [train.py:904] (5/8) Epoch 4, batch 1200, loss[loss=0.2092, simple_loss=0.2945, pruned_loss=0.06198, over 17118.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2928, pruned_loss=0.07368, over 3321791.91 frames. ], batch size: 49, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:51:40,966 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:52:19,330 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3011, 1.8089, 2.2990, 2.9761, 2.9105, 3.3806, 2.1046, 3.2848], device='cuda:5'), covar=tensor([0.0053, 0.0189, 0.0136, 0.0102, 0.0075, 0.0071, 0.0155, 0.0044], device='cuda:5'), in_proj_covar=tensor([0.0102, 0.0130, 0.0116, 0.0111, 0.0106, 0.0081, 0.0123, 0.0068], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 01:52:27,381 INFO [train.py:904] (5/8) Epoch 4, batch 1250, loss[loss=0.2373, simple_loss=0.3187, pruned_loss=0.07794, over 17104.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2939, pruned_loss=0.07446, over 3320205.15 frames. ], batch size: 47, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:52:31,897 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:47,964 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:53:22,525 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.448e+02 3.415e+02 4.132e+02 5.466e+02 7.512e+02, threshold=8.265e+02, percent-clipped=0.0 2023-04-28 01:53:31,098 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 01:53:35,329 INFO [train.py:904] (5/8) Epoch 4, batch 1300, loss[loss=0.2159, simple_loss=0.2795, pruned_loss=0.07616, over 16453.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2939, pruned_loss=0.07453, over 3322278.16 frames. ], batch size: 68, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:53:46,240 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2222, 5.4827, 5.0554, 5.1653, 4.1399, 5.2534, 5.2663, 4.8380], device='cuda:5'), covar=tensor([0.0469, 0.0137, 0.0235, 0.0181, 0.1348, 0.0191, 0.0144, 0.0291], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0158, 0.0208, 0.0175, 0.0245, 0.0190, 0.0148, 0.0208], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 01:53:54,675 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:53:54,846 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:26,668 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:44,575 INFO [train.py:904] (5/8) Epoch 4, batch 1350, loss[loss=0.2111, simple_loss=0.2969, pruned_loss=0.0627, over 17096.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2935, pruned_loss=0.07441, over 3319312.14 frames. ], batch size: 49, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:55:00,317 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4804, 1.8367, 2.5826, 3.0310, 3.0519, 3.4879, 2.3739, 3.3772], device='cuda:5'), covar=tensor([0.0047, 0.0204, 0.0126, 0.0098, 0.0076, 0.0070, 0.0154, 0.0044], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0130, 0.0117, 0.0111, 0.0106, 0.0081, 0.0125, 0.0069], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 01:55:14,609 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:39,670 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.154e+02 3.282e+02 4.037e+02 5.139e+02 9.544e+02, threshold=8.075e+02, percent-clipped=2.0 2023-04-28 01:55:48,227 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:55:50,789 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:52,774 INFO [train.py:904] (5/8) Epoch 4, batch 1400, loss[loss=0.2073, simple_loss=0.2906, pruned_loss=0.06199, over 17184.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2931, pruned_loss=0.07383, over 3316147.12 frames. ], batch size: 44, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:56:13,003 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 01:56:32,794 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:56:40,516 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:02,742 INFO [train.py:904] (5/8) Epoch 4, batch 1450, loss[loss=0.2024, simple_loss=0.2766, pruned_loss=0.06411, over 17062.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2928, pruned_loss=0.07425, over 3308003.66 frames. ], batch size: 41, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:57:22,505 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:56,650 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:00,354 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.224e+02 3.023e+02 3.906e+02 5.046e+02 9.760e+02, threshold=7.812e+02, percent-clipped=3.0 2023-04-28 01:58:12,687 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:13,470 INFO [train.py:904] (5/8) Epoch 4, batch 1500, loss[loss=0.2234, simple_loss=0.2835, pruned_loss=0.0817, over 16807.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2928, pruned_loss=0.07494, over 3318193.39 frames. ], batch size: 83, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:58:15,464 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-28 01:58:19,447 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-28 01:58:47,452 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:56,794 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-28 01:59:26,626 INFO [train.py:904] (5/8) Epoch 4, batch 1550, loss[loss=0.2183, simple_loss=0.3036, pruned_loss=0.06652, over 17158.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2946, pruned_loss=0.07624, over 3315160.42 frames. ], batch size: 48, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:59:40,574 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:21,781 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.322e+02 3.491e+02 4.515e+02 5.293e+02 8.780e+02, threshold=9.030e+02, percent-clipped=2.0 2023-04-28 02:00:34,388 INFO [train.py:904] (5/8) Epoch 4, batch 1600, loss[loss=0.2099, simple_loss=0.2893, pruned_loss=0.06524, over 17237.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.2966, pruned_loss=0.07708, over 3318515.85 frames. ], batch size: 45, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:00:46,086 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:52,744 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:01:05,748 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.21 vs. limit=2.0 2023-04-28 02:01:41,687 INFO [train.py:904] (5/8) Epoch 4, batch 1650, loss[loss=0.221, simple_loss=0.3021, pruned_loss=0.06994, over 17119.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.2982, pruned_loss=0.07778, over 3320589.16 frames. ], batch size: 48, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:01:58,772 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:37,465 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.433e+02 3.425e+02 4.282e+02 5.685e+02 1.161e+03, threshold=8.563e+02, percent-clipped=3.0 2023-04-28 02:02:40,720 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:44,922 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:02:50,063 INFO [train.py:904] (5/8) Epoch 4, batch 1700, loss[loss=0.2408, simple_loss=0.3085, pruned_loss=0.08654, over 15910.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3011, pruned_loss=0.0792, over 3326056.18 frames. ], batch size: 35, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:02:52,832 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2125, 1.6502, 2.3706, 2.9187, 2.7750, 3.3609, 2.1592, 3.1414], device='cuda:5'), covar=tensor([0.0048, 0.0201, 0.0123, 0.0094, 0.0095, 0.0061, 0.0154, 0.0061], device='cuda:5'), in_proj_covar=tensor([0.0104, 0.0132, 0.0119, 0.0115, 0.0110, 0.0083, 0.0127, 0.0073], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 02:03:29,696 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32179.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:03:53,201 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:04:01,147 INFO [train.py:904] (5/8) Epoch 4, batch 1750, loss[loss=0.2005, simple_loss=0.2856, pruned_loss=0.05767, over 17181.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3011, pruned_loss=0.07804, over 3328148.36 frames. ], batch size: 46, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:04:48,321 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:58,763 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.044e+02 3.415e+02 4.164e+02 5.127e+02 8.958e+02, threshold=8.329e+02, percent-clipped=1.0 2023-04-28 02:05:11,466 INFO [train.py:904] (5/8) Epoch 4, batch 1800, loss[loss=0.2127, simple_loss=0.2851, pruned_loss=0.07015, over 16812.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3031, pruned_loss=0.07871, over 3315965.86 frames. ], batch size: 42, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:05:39,581 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:18,951 INFO [train.py:904] (5/8) Epoch 4, batch 1850, loss[loss=0.2895, simple_loss=0.3527, pruned_loss=0.1131, over 12036.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3041, pruned_loss=0.07896, over 3301940.06 frames. ], batch size: 246, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:06:26,592 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:07:18,153 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.462e+02 3.616e+02 4.272e+02 5.393e+02 1.345e+03, threshold=8.544e+02, percent-clipped=8.0 2023-04-28 02:07:29,377 INFO [train.py:904] (5/8) Epoch 4, batch 1900, loss[loss=0.2072, simple_loss=0.2944, pruned_loss=0.06002, over 17071.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3037, pruned_loss=0.07841, over 3311102.14 frames. ], batch size: 49, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:07:43,022 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:41,350 INFO [train.py:904] (5/8) Epoch 4, batch 1950, loss[loss=0.2055, simple_loss=0.2792, pruned_loss=0.06585, over 16003.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3019, pruned_loss=0.0769, over 3304888.10 frames. ], batch size: 35, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:08:50,794 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:16,855 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9532, 4.8461, 4.9052, 3.4879, 4.7918, 1.9977, 4.5235, 4.9162], device='cuda:5'), covar=tensor([0.0099, 0.0093, 0.0095, 0.0603, 0.0096, 0.1838, 0.0114, 0.0167], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0078, 0.0115, 0.0127, 0.0085, 0.0129, 0.0104, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 02:09:37,987 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.345e+02 3.206e+02 3.800e+02 4.937e+02 1.195e+03, threshold=7.601e+02, percent-clipped=2.0 2023-04-28 02:09:39,455 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:47,992 INFO [train.py:904] (5/8) Epoch 4, batch 2000, loss[loss=0.2525, simple_loss=0.3225, pruned_loss=0.09123, over 17162.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3023, pruned_loss=0.07702, over 3301049.18 frames. ], batch size: 47, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:10:13,152 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2668, 4.9840, 5.1496, 5.5216, 5.6034, 4.6837, 5.5438, 5.5090], device='cuda:5'), covar=tensor([0.0776, 0.0641, 0.1415, 0.0405, 0.0435, 0.0545, 0.0472, 0.0350], device='cuda:5'), in_proj_covar=tensor([0.0357, 0.0430, 0.0577, 0.0445, 0.0331, 0.0325, 0.0352, 0.0362], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 02:10:28,180 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:43,493 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9340, 4.1869, 3.3944, 2.4600, 3.1418, 2.3619, 4.5937, 4.3418], device='cuda:5'), covar=tensor([0.1987, 0.0662, 0.1079, 0.1326, 0.2273, 0.1321, 0.0283, 0.0545], device='cuda:5'), in_proj_covar=tensor([0.0272, 0.0251, 0.0259, 0.0226, 0.0297, 0.0198, 0.0228, 0.0246], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 02:10:45,039 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:58,284 INFO [train.py:904] (5/8) Epoch 4, batch 2050, loss[loss=0.2237, simple_loss=0.314, pruned_loss=0.06666, over 17285.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3025, pruned_loss=0.07671, over 3299901.82 frames. ], batch size: 52, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:11:28,297 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9463, 3.4291, 2.7489, 4.7970, 4.6019, 4.1979, 1.8816, 3.2032], device='cuda:5'), covar=tensor([0.1159, 0.0431, 0.0963, 0.0072, 0.0271, 0.0292, 0.1137, 0.0629], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0136, 0.0162, 0.0078, 0.0171, 0.0156, 0.0153, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 02:11:34,834 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:44,265 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:57,878 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.075e+02 3.479e+02 4.226e+02 5.349e+02 1.056e+03, threshold=8.453e+02, percent-clipped=7.0 2023-04-28 02:12:07,694 INFO [train.py:904] (5/8) Epoch 4, batch 2100, loss[loss=0.2329, simple_loss=0.307, pruned_loss=0.07942, over 16497.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3029, pruned_loss=0.07732, over 3304706.84 frames. ], batch size: 75, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:12:36,408 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:51,052 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:17,666 INFO [train.py:904] (5/8) Epoch 4, batch 2150, loss[loss=0.2263, simple_loss=0.3088, pruned_loss=0.07185, over 16724.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3046, pruned_loss=0.0787, over 3311091.46 frames. ], batch size: 57, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:13:25,184 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:27,059 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6279, 4.5003, 4.4799, 4.4660, 4.1057, 4.5257, 4.3650, 4.2843], device='cuda:5'), covar=tensor([0.0361, 0.0233, 0.0160, 0.0128, 0.0645, 0.0211, 0.0257, 0.0294], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0163, 0.0210, 0.0176, 0.0248, 0.0196, 0.0149, 0.0211], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 02:13:42,054 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:49,759 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 02:14:08,615 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 02:14:15,067 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.513e+02 3.522e+02 4.278e+02 5.461e+02 9.146e+02, threshold=8.556e+02, percent-clipped=3.0 2023-04-28 02:14:26,833 INFO [train.py:904] (5/8) Epoch 4, batch 2200, loss[loss=0.2125, simple_loss=0.295, pruned_loss=0.06497, over 17214.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3052, pruned_loss=0.07922, over 3316140.38 frames. ], batch size: 45, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:14:30,433 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:15:34,039 INFO [train.py:904] (5/8) Epoch 4, batch 2250, loss[loss=0.2256, simple_loss=0.2966, pruned_loss=0.07726, over 16449.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3057, pruned_loss=0.0794, over 3322986.27 frames. ], batch size: 75, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:15:38,335 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-28 02:16:32,023 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.152e+02 3.558e+02 4.554e+02 5.446e+02 8.366e+02, threshold=9.108e+02, percent-clipped=0.0 2023-04-28 02:16:42,219 INFO [train.py:904] (5/8) Epoch 4, batch 2300, loss[loss=0.2004, simple_loss=0.2839, pruned_loss=0.05845, over 17245.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3058, pruned_loss=0.07931, over 3327123.95 frames. ], batch size: 44, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:16:57,820 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 02:17:51,372 INFO [train.py:904] (5/8) Epoch 4, batch 2350, loss[loss=0.2337, simple_loss=0.3146, pruned_loss=0.07641, over 17260.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3069, pruned_loss=0.08002, over 3318221.02 frames. ], batch size: 52, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:18:48,986 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.340e+02 3.303e+02 3.984e+02 5.110e+02 1.375e+03, threshold=7.969e+02, percent-clipped=3.0 2023-04-28 02:18:57,207 INFO [train.py:904] (5/8) Epoch 4, batch 2400, loss[loss=0.2133, simple_loss=0.2966, pruned_loss=0.06503, over 17130.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3073, pruned_loss=0.07985, over 3319160.20 frames. ], batch size: 47, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:19:21,656 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3930, 5.6696, 5.4286, 5.5099, 4.9440, 4.5678, 5.1829, 5.7855], device='cuda:5'), covar=tensor([0.0524, 0.0619, 0.0810, 0.0396, 0.0511, 0.0639, 0.0496, 0.0653], device='cuda:5'), in_proj_covar=tensor([0.0321, 0.0449, 0.0380, 0.0280, 0.0284, 0.0280, 0.0349, 0.0308], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 02:19:49,158 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-04-28 02:20:06,099 INFO [train.py:904] (5/8) Epoch 4, batch 2450, loss[loss=0.2597, simple_loss=0.3191, pruned_loss=0.1002, over 16473.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3075, pruned_loss=0.07943, over 3320274.80 frames. ], batch size: 146, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:20:52,788 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7063, 4.7579, 5.3162, 5.2436, 5.2820, 4.8902, 4.8400, 4.4964], device='cuda:5'), covar=tensor([0.0215, 0.0298, 0.0271, 0.0373, 0.0347, 0.0224, 0.0643, 0.0289], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0221, 0.0227, 0.0228, 0.0271, 0.0239, 0.0344, 0.0203], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 02:21:03,697 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.564e+02 3.644e+02 4.364e+02 5.421e+02 9.049e+02, threshold=8.728e+02, percent-clipped=5.0 2023-04-28 02:21:13,298 INFO [train.py:904] (5/8) Epoch 4, batch 2500, loss[loss=0.2153, simple_loss=0.2934, pruned_loss=0.06862, over 17222.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3076, pruned_loss=0.07934, over 3311405.14 frames. ], batch size: 44, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:21:34,413 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:21,146 INFO [train.py:904] (5/8) Epoch 4, batch 2550, loss[loss=0.2182, simple_loss=0.2876, pruned_loss=0.07437, over 16517.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3075, pruned_loss=0.07974, over 3310322.18 frames. ], batch size: 75, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:22:57,676 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33027.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:23:20,059 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.536e+02 3.626e+02 4.489e+02 5.750e+02 1.218e+03, threshold=8.978e+02, percent-clipped=4.0 2023-04-28 02:23:30,993 INFO [train.py:904] (5/8) Epoch 4, batch 2600, loss[loss=0.2171, simple_loss=0.3107, pruned_loss=0.06178, over 17247.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3077, pruned_loss=0.07953, over 3312270.99 frames. ], batch size: 52, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:23:55,332 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:24:39,068 INFO [train.py:904] (5/8) Epoch 4, batch 2650, loss[loss=0.2275, simple_loss=0.3078, pruned_loss=0.07356, over 17177.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3072, pruned_loss=0.07825, over 3321865.81 frames. ], batch size: 46, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:19,573 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:25:23,049 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7639, 2.9772, 2.3900, 4.4441, 4.1617, 4.1131, 1.6043, 2.7635], device='cuda:5'), covar=tensor([0.1331, 0.0516, 0.1158, 0.0074, 0.0290, 0.0261, 0.1289, 0.0777], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0136, 0.0163, 0.0079, 0.0170, 0.0158, 0.0155, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 02:25:38,561 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 3.285e+02 4.083e+02 4.956e+02 1.136e+03, threshold=8.166e+02, percent-clipped=3.0 2023-04-28 02:25:46,698 INFO [train.py:904] (5/8) Epoch 4, batch 2700, loss[loss=0.2258, simple_loss=0.3046, pruned_loss=0.07344, over 17036.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3067, pruned_loss=0.07794, over 3320823.62 frames. ], batch size: 50, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:54,233 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33155.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:26:56,822 INFO [train.py:904] (5/8) Epoch 4, batch 2750, loss[loss=0.2048, simple_loss=0.2962, pruned_loss=0.05666, over 17071.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3057, pruned_loss=0.07615, over 3323180.23 frames. ], batch size: 53, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:27:16,289 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:27:50,346 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1231, 5.1208, 4.8757, 4.9110, 4.2429, 4.9537, 4.9656, 4.5484], device='cuda:5'), covar=tensor([0.0442, 0.0240, 0.0246, 0.0166, 0.1177, 0.0244, 0.0251, 0.0487], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0165, 0.0212, 0.0177, 0.0248, 0.0197, 0.0152, 0.0217], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 02:27:54,084 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.476e+02 3.299e+02 3.913e+02 4.982e+02 9.564e+02, threshold=7.826e+02, percent-clipped=3.0 2023-04-28 02:28:04,367 INFO [train.py:904] (5/8) Epoch 4, batch 2800, loss[loss=0.2288, simple_loss=0.2978, pruned_loss=0.07992, over 16915.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3055, pruned_loss=0.07601, over 3328281.95 frames. ], batch size: 96, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:28:13,390 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 02:29:06,950 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 02:29:10,599 INFO [train.py:904] (5/8) Epoch 4, batch 2850, loss[loss=0.2532, simple_loss=0.3078, pruned_loss=0.09929, over 16243.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3054, pruned_loss=0.07665, over 3334947.16 frames. ], batch size: 165, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:29:39,686 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:30:09,824 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 3.527e+02 4.279e+02 5.315e+02 1.559e+03, threshold=8.559e+02, percent-clipped=4.0 2023-04-28 02:30:20,497 INFO [train.py:904] (5/8) Epoch 4, batch 2900, loss[loss=0.2322, simple_loss=0.3149, pruned_loss=0.07469, over 16676.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3047, pruned_loss=0.07738, over 3321943.74 frames. ], batch size: 57, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:30:55,846 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6295, 3.5507, 3.6941, 3.5018, 3.4916, 4.0023, 3.8701, 3.5822], device='cuda:5'), covar=tensor([0.2166, 0.1969, 0.1604, 0.2764, 0.3310, 0.1696, 0.1411, 0.2883], device='cuda:5'), in_proj_covar=tensor([0.0269, 0.0372, 0.0351, 0.0315, 0.0423, 0.0373, 0.0296, 0.0428], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 02:31:29,005 INFO [train.py:904] (5/8) Epoch 4, batch 2950, loss[loss=0.2552, simple_loss=0.3302, pruned_loss=0.09012, over 16719.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3051, pruned_loss=0.07909, over 3310557.24 frames. ], batch size: 57, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:01,400 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:32:27,165 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.112e+02 3.646e+02 4.580e+02 6.032e+02 1.054e+03, threshold=9.160e+02, percent-clipped=6.0 2023-04-28 02:32:35,895 INFO [train.py:904] (5/8) Epoch 4, batch 3000, loss[loss=0.269, simple_loss=0.3232, pruned_loss=0.1074, over 16869.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3049, pruned_loss=0.07933, over 3317702.99 frames. ], batch size: 109, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:35,895 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 02:32:45,551 INFO [train.py:938] (5/8) Epoch 4, validation: loss=0.1627, simple_loss=0.2694, pruned_loss=0.02796, over 944034.00 frames. 2023-04-28 02:32:45,551 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 02:32:50,970 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1635, 4.5664, 4.6008, 1.8927, 4.8648, 4.9358, 3.2765, 3.9239], device='cuda:5'), covar=tensor([0.0526, 0.0088, 0.0145, 0.1151, 0.0055, 0.0041, 0.0312, 0.0235], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0084, 0.0084, 0.0144, 0.0075, 0.0077, 0.0116, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 02:32:54,637 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0554, 4.4075, 3.4329, 2.5549, 3.2816, 2.4283, 4.6684, 4.6839], device='cuda:5'), covar=tensor([0.1687, 0.0443, 0.0871, 0.1160, 0.1879, 0.1173, 0.0232, 0.0343], device='cuda:5'), in_proj_covar=tensor([0.0265, 0.0246, 0.0256, 0.0224, 0.0296, 0.0192, 0.0225, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 02:33:12,905 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:54,596 INFO [train.py:904] (5/8) Epoch 4, batch 3050, loss[loss=0.2215, simple_loss=0.3046, pruned_loss=0.06918, over 17055.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3049, pruned_loss=0.0799, over 3311518.02 frames. ], batch size: 53, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:34:07,911 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:38,110 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:41,705 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8740, 5.1675, 4.8199, 4.9354, 4.5763, 4.4565, 4.6367, 5.2003], device='cuda:5'), covar=tensor([0.0594, 0.0541, 0.0816, 0.0389, 0.0522, 0.0630, 0.0613, 0.0609], device='cuda:5'), in_proj_covar=tensor([0.0334, 0.0457, 0.0392, 0.0287, 0.0291, 0.0286, 0.0362, 0.0321], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 02:34:43,963 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-28 02:34:44,952 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 02:34:54,166 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 3.585e+02 4.279e+02 5.215e+02 1.682e+03, threshold=8.559e+02, percent-clipped=1.0 2023-04-28 02:35:02,725 INFO [train.py:904] (5/8) Epoch 4, batch 3100, loss[loss=0.2242, simple_loss=0.303, pruned_loss=0.07274, over 17178.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3049, pruned_loss=0.08008, over 3315863.49 frames. ], batch size: 46, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:36:03,238 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2336, 4.5378, 1.9566, 4.7401, 2.7623, 4.6933, 2.2961, 3.1240], device='cuda:5'), covar=tensor([0.0093, 0.0127, 0.1317, 0.0021, 0.0632, 0.0207, 0.1155, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0108, 0.0151, 0.0175, 0.0086, 0.0159, 0.0183, 0.0183, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 02:36:10,687 INFO [train.py:904] (5/8) Epoch 4, batch 3150, loss[loss=0.2415, simple_loss=0.3214, pruned_loss=0.08082, over 17110.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3041, pruned_loss=0.07939, over 3311840.14 frames. ], batch size: 55, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:36:37,243 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7754, 2.5427, 2.4232, 3.9873, 3.6960, 3.8202, 1.5580, 2.8074], device='cuda:5'), covar=tensor([0.1280, 0.0634, 0.1125, 0.0091, 0.0264, 0.0266, 0.1285, 0.0728], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0138, 0.0165, 0.0080, 0.0171, 0.0158, 0.0156, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 02:36:39,478 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:11,334 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.300e+02 3.211e+02 3.811e+02 4.570e+02 1.076e+03, threshold=7.622e+02, percent-clipped=2.0 2023-04-28 02:37:18,487 INFO [train.py:904] (5/8) Epoch 4, batch 3200, loss[loss=0.2156, simple_loss=0.2856, pruned_loss=0.07277, over 16451.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3033, pruned_loss=0.07926, over 3316343.82 frames. ], batch size: 75, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:37:20,132 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0458, 4.6653, 4.7141, 3.7753, 4.3286, 4.7919, 4.2952, 2.9349], device='cuda:5'), covar=tensor([0.0220, 0.0028, 0.0029, 0.0128, 0.0025, 0.0025, 0.0021, 0.0212], device='cuda:5'), in_proj_covar=tensor([0.0107, 0.0053, 0.0056, 0.0104, 0.0057, 0.0062, 0.0057, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 02:37:44,876 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:54,521 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:38:00,434 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 02:38:25,342 INFO [train.py:904] (5/8) Epoch 4, batch 3250, loss[loss=0.252, simple_loss=0.3285, pruned_loss=0.08772, over 16708.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3029, pruned_loss=0.07851, over 3317878.99 frames. ], batch size: 62, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:38:45,464 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 02:38:58,309 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:16,475 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 02:39:26,584 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.718e+02 3.624e+02 4.331e+02 5.289e+02 1.384e+03, threshold=8.662e+02, percent-clipped=5.0 2023-04-28 02:39:36,479 INFO [train.py:904] (5/8) Epoch 4, batch 3300, loss[loss=0.2056, simple_loss=0.2756, pruned_loss=0.06775, over 16803.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3022, pruned_loss=0.0771, over 3326352.99 frames. ], batch size: 39, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:40:07,419 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:40:15,399 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1631, 4.1221, 4.1244, 3.3161, 4.1455, 1.7920, 3.9470, 4.0620], device='cuda:5'), covar=tensor([0.0100, 0.0076, 0.0099, 0.0382, 0.0073, 0.1617, 0.0097, 0.0131], device='cuda:5'), in_proj_covar=tensor([0.0090, 0.0081, 0.0122, 0.0131, 0.0090, 0.0129, 0.0107, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 02:40:45,520 INFO [train.py:904] (5/8) Epoch 4, batch 3350, loss[loss=0.2226, simple_loss=0.2989, pruned_loss=0.07318, over 16798.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.302, pruned_loss=0.07624, over 3328939.90 frames. ], batch size: 124, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:40:58,966 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:22,104 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:44,270 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.269e+02 3.302e+02 3.975e+02 4.780e+02 1.300e+03, threshold=7.950e+02, percent-clipped=1.0 2023-04-28 02:41:52,911 INFO [train.py:904] (5/8) Epoch 4, batch 3400, loss[loss=0.255, simple_loss=0.3076, pruned_loss=0.1012, over 16884.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3025, pruned_loss=0.07668, over 3325806.46 frames. ], batch size: 116, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:42:04,796 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:43:03,532 INFO [train.py:904] (5/8) Epoch 4, batch 3450, loss[loss=0.2589, simple_loss=0.3219, pruned_loss=0.098, over 16456.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3014, pruned_loss=0.0758, over 3331897.43 frames. ], batch size: 68, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:43:58,433 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:44:05,160 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 3.338e+02 4.153e+02 5.120e+02 1.104e+03, threshold=8.306e+02, percent-clipped=4.0 2023-04-28 02:44:13,349 INFO [train.py:904] (5/8) Epoch 4, batch 3500, loss[loss=0.1907, simple_loss=0.2789, pruned_loss=0.05125, over 17143.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2993, pruned_loss=0.0748, over 3334403.89 frames. ], batch size: 49, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:45:26,875 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:45:27,582 INFO [train.py:904] (5/8) Epoch 4, batch 3550, loss[loss=0.222, simple_loss=0.3154, pruned_loss=0.06428, over 17082.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2984, pruned_loss=0.07417, over 3334419.80 frames. ], batch size: 55, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:46:11,051 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:46:27,254 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34044.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:46:27,926 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.116e+02 3.324e+02 4.237e+02 5.025e+02 1.251e+03, threshold=8.474e+02, percent-clipped=3.0 2023-04-28 02:46:35,601 INFO [train.py:904] (5/8) Epoch 4, batch 3600, loss[loss=0.2122, simple_loss=0.3057, pruned_loss=0.05931, over 17110.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2974, pruned_loss=0.0734, over 3341657.87 frames. ], batch size: 47, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:12,046 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 02:47:21,462 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1860, 3.6690, 2.8678, 5.3284, 5.0864, 4.6238, 1.7246, 3.7212], device='cuda:5'), covar=tensor([0.1083, 0.0396, 0.1006, 0.0057, 0.0184, 0.0258, 0.1173, 0.0500], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0136, 0.0160, 0.0079, 0.0170, 0.0157, 0.0152, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 02:47:48,228 INFO [train.py:904] (5/8) Epoch 4, batch 3650, loss[loss=0.2268, simple_loss=0.2856, pruned_loss=0.08397, over 16396.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2966, pruned_loss=0.07434, over 3332777.24 frames. ], batch size: 75, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:54,687 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:29,190 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:53,745 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.336e+02 3.266e+02 3.802e+02 4.883e+02 8.817e+02, threshold=7.604e+02, percent-clipped=1.0 2023-04-28 02:48:54,325 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:02,651 INFO [train.py:904] (5/8) Epoch 4, batch 3700, loss[loss=0.2046, simple_loss=0.2677, pruned_loss=0.07078, over 16397.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.2949, pruned_loss=0.07621, over 3303763.46 frames. ], batch size: 75, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:49:12,452 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 02:49:41,508 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:51,218 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0732, 3.6283, 3.6021, 1.7743, 3.7397, 3.7510, 2.9363, 2.7079], device='cuda:5'), covar=tensor([0.0751, 0.0075, 0.0098, 0.1025, 0.0058, 0.0060, 0.0334, 0.0378], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0081, 0.0080, 0.0140, 0.0072, 0.0075, 0.0113, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 02:50:17,552 INFO [train.py:904] (5/8) Epoch 4, batch 3750, loss[loss=0.2265, simple_loss=0.3081, pruned_loss=0.07245, over 16645.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.2962, pruned_loss=0.07865, over 3282280.10 frames. ], batch size: 57, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:50:25,305 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:51:21,424 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.239e+02 3.358e+02 4.060e+02 5.073e+02 1.417e+03, threshold=8.120e+02, percent-clipped=4.0 2023-04-28 02:51:29,947 INFO [train.py:904] (5/8) Epoch 4, batch 3800, loss[loss=0.2425, simple_loss=0.302, pruned_loss=0.09144, over 16758.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.2964, pruned_loss=0.07975, over 3270890.83 frames. ], batch size: 124, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:51:38,152 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 02:52:36,347 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34295.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:52:44,483 INFO [train.py:904] (5/8) Epoch 4, batch 3850, loss[loss=0.2149, simple_loss=0.2895, pruned_loss=0.07018, over 16777.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.297, pruned_loss=0.08079, over 3259881.89 frames. ], batch size: 102, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:52:56,131 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-04-28 02:53:31,114 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:53:49,248 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 02:53:49,432 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.155e+02 3.189e+02 3.757e+02 4.739e+02 9.284e+02, threshold=7.515e+02, percent-clipped=3.0 2023-04-28 02:53:57,018 INFO [train.py:904] (5/8) Epoch 4, batch 3900, loss[loss=0.2441, simple_loss=0.305, pruned_loss=0.09164, over 16742.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.2955, pruned_loss=0.08047, over 3277474.70 frames. ], batch size: 134, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:54:01,410 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 02:54:40,927 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:54:56,454 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9184, 3.7269, 3.8978, 4.0928, 4.1694, 3.7040, 3.9390, 4.1307], device='cuda:5'), covar=tensor([0.0785, 0.0677, 0.1157, 0.0576, 0.0488, 0.1192, 0.1071, 0.0437], device='cuda:5'), in_proj_covar=tensor([0.0351, 0.0425, 0.0555, 0.0445, 0.0331, 0.0318, 0.0345, 0.0358], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 02:55:09,566 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34400.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:55:10,288 INFO [train.py:904] (5/8) Epoch 4, batch 3950, loss[loss=0.3017, simple_loss=0.3485, pruned_loss=0.1275, over 12917.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.296, pruned_loss=0.08151, over 3272106.73 frames. ], batch size: 246, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:56:04,194 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0962, 5.0701, 4.8576, 4.8733, 4.4321, 4.8646, 4.7612, 4.5969], device='cuda:5'), covar=tensor([0.0346, 0.0169, 0.0173, 0.0130, 0.0840, 0.0172, 0.0225, 0.0315], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0165, 0.0209, 0.0175, 0.0243, 0.0197, 0.0150, 0.0214], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 02:56:16,762 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 3.579e+02 4.210e+02 5.438e+02 1.110e+03, threshold=8.420e+02, percent-clipped=9.0 2023-04-28 02:56:24,491 INFO [train.py:904] (5/8) Epoch 4, batch 4000, loss[loss=0.2444, simple_loss=0.3088, pruned_loss=0.09001, over 16881.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.295, pruned_loss=0.08122, over 3279449.55 frames. ], batch size: 109, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:36,440 INFO [train.py:904] (5/8) Epoch 4, batch 4050, loss[loss=0.2453, simple_loss=0.3108, pruned_loss=0.08988, over 12500.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.2946, pruned_loss=0.07928, over 3272726.87 frames. ], batch size: 246, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:36,872 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:58:41,516 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.624e+02 2.984e+02 3.642e+02 7.677e+02, threshold=5.968e+02, percent-clipped=0.0 2023-04-28 02:58:48,893 INFO [train.py:904] (5/8) Epoch 4, batch 4100, loss[loss=0.285, simple_loss=0.355, pruned_loss=0.1075, over 15475.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.2958, pruned_loss=0.07806, over 3265060.06 frames. ], batch size: 191, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:59:54,035 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:59:57,628 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:06,230 INFO [train.py:904] (5/8) Epoch 4, batch 4150, loss[loss=0.2319, simple_loss=0.3135, pruned_loss=0.07515, over 17028.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3042, pruned_loss=0.08151, over 3241162.11 frames. ], batch size: 50, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:00:38,893 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:44,961 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:05,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-28 03:01:10,676 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:15,086 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.947e+02 3.447e+02 4.221e+02 5.425e+02 1.080e+03, threshold=8.441e+02, percent-clipped=15.0 2023-04-28 03:01:22,993 INFO [train.py:904] (5/8) Epoch 4, batch 4200, loss[loss=0.2444, simple_loss=0.3268, pruned_loss=0.08096, over 16701.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3127, pruned_loss=0.08477, over 3200031.37 frames. ], batch size: 134, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:01:28,582 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:13,336 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:02:19,283 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:23,264 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-28 03:02:39,602 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34700.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:40,367 INFO [train.py:904] (5/8) Epoch 4, batch 4250, loss[loss=0.2535, simple_loss=0.3168, pruned_loss=0.09506, over 12288.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3149, pruned_loss=0.08423, over 3181631.08 frames. ], batch size: 248, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:02:53,656 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:02:58,860 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9631, 4.2154, 3.6668, 2.5260, 3.4519, 2.5796, 4.5525, 4.5370], device='cuda:5'), covar=tensor([0.1848, 0.0537, 0.0912, 0.1239, 0.1765, 0.1108, 0.0300, 0.0380], device='cuda:5'), in_proj_covar=tensor([0.0267, 0.0241, 0.0255, 0.0228, 0.0302, 0.0193, 0.0224, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:03:47,978 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 2.926e+02 3.520e+02 4.399e+02 1.246e+03, threshold=7.039e+02, percent-clipped=2.0 2023-04-28 03:03:52,077 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:03:52,450 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 03:03:56,215 INFO [train.py:904] (5/8) Epoch 4, batch 4300, loss[loss=0.25, simple_loss=0.3367, pruned_loss=0.08171, over 16784.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3158, pruned_loss=0.08289, over 3172471.88 frames. ], batch size: 124, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:04:27,296 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:05:10,252 INFO [train.py:904] (5/8) Epoch 4, batch 4350, loss[loss=0.2215, simple_loss=0.3081, pruned_loss=0.06746, over 16707.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3197, pruned_loss=0.08428, over 3175656.45 frames. ], batch size: 83, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:05:10,571 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:05:40,600 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3856, 2.1404, 1.7720, 2.1102, 2.6924, 2.4556, 3.3792, 3.0818], device='cuda:5'), covar=tensor([0.0016, 0.0150, 0.0204, 0.0162, 0.0097, 0.0147, 0.0030, 0.0073], device='cuda:5'), in_proj_covar=tensor([0.0071, 0.0137, 0.0139, 0.0135, 0.0130, 0.0140, 0.0100, 0.0116], device='cuda:5'), out_proj_covar=tensor([9.3393e-05, 1.8166e-04, 1.7800e-04, 1.7509e-04, 1.7467e-04, 1.8681e-04, 1.3216e-04, 1.5654e-04], device='cuda:5') 2023-04-28 03:06:15,523 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.334e+02 4.019e+02 4.887e+02 8.589e+02, threshold=8.038e+02, percent-clipped=2.0 2023-04-28 03:06:20,218 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:22,181 INFO [train.py:904] (5/8) Epoch 4, batch 4400, loss[loss=0.2528, simple_loss=0.3297, pruned_loss=0.08793, over 17095.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3216, pruned_loss=0.08501, over 3180244.08 frames. ], batch size: 47, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:07:32,105 INFO [train.py:904] (5/8) Epoch 4, batch 4450, loss[loss=0.2263, simple_loss=0.3173, pruned_loss=0.06771, over 16517.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3237, pruned_loss=0.08443, over 3210279.66 frames. ], batch size: 75, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:07:46,799 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 03:08:02,260 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0255, 2.7642, 2.5806, 1.7781, 2.7732, 2.8555, 2.5158, 2.3504], device='cuda:5'), covar=tensor([0.0692, 0.0121, 0.0168, 0.0878, 0.0064, 0.0067, 0.0389, 0.0395], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0081, 0.0079, 0.0137, 0.0068, 0.0070, 0.0111, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 03:08:36,347 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.921e+02 3.471e+02 4.143e+02 7.527e+02, threshold=6.942e+02, percent-clipped=0.0 2023-04-28 03:08:40,982 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:08:43,015 INFO [train.py:904] (5/8) Epoch 4, batch 4500, loss[loss=0.2255, simple_loss=0.3088, pruned_loss=0.07107, over 16527.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3231, pruned_loss=0.08407, over 3217060.41 frames. ], batch size: 68, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:09:23,067 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:09:28,669 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:09:29,949 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2079, 3.8127, 3.5072, 1.6383, 2.7129, 2.3430, 3.3831, 3.6424], device='cuda:5'), covar=tensor([0.0222, 0.0426, 0.0408, 0.1706, 0.0723, 0.0853, 0.0611, 0.0630], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0127, 0.0155, 0.0143, 0.0136, 0.0128, 0.0142, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 03:09:54,758 INFO [train.py:904] (5/8) Epoch 4, batch 4550, loss[loss=0.2277, simple_loss=0.305, pruned_loss=0.07521, over 16376.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3234, pruned_loss=0.0845, over 3227119.03 frames. ], batch size: 35, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:10:57,622 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 2.863e+02 3.417e+02 4.239e+02 9.930e+02, threshold=6.834e+02, percent-clipped=3.0 2023-04-28 03:11:04,056 INFO [train.py:904] (5/8) Epoch 4, batch 4600, loss[loss=0.2528, simple_loss=0.3329, pruned_loss=0.08636, over 16876.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3233, pruned_loss=0.08392, over 3229608.67 frames. ], batch size: 109, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:11:25,510 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:12:02,634 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:12:15,712 INFO [train.py:904] (5/8) Epoch 4, batch 4650, loss[loss=0.2139, simple_loss=0.298, pruned_loss=0.06496, over 16749.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3219, pruned_loss=0.08327, over 3230138.66 frames. ], batch size: 89, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:13:20,504 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.712e+02 3.466e+02 4.003e+02 6.879e+02, threshold=6.932e+02, percent-clipped=1.0 2023-04-28 03:13:28,257 INFO [train.py:904] (5/8) Epoch 4, batch 4700, loss[loss=0.2647, simple_loss=0.3334, pruned_loss=0.09793, over 15298.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3185, pruned_loss=0.08179, over 3228271.22 frames. ], batch size: 190, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:13:32,530 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:41,630 INFO [train.py:904] (5/8) Epoch 4, batch 4750, loss[loss=0.2146, simple_loss=0.2924, pruned_loss=0.06842, over 16538.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3141, pruned_loss=0.07972, over 3232850.90 frames. ], batch size: 75, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:14:56,832 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7651, 2.6752, 2.3599, 1.5022, 2.6203, 2.6940, 2.4052, 2.1640], device='cuda:5'), covar=tensor([0.0959, 0.0149, 0.0200, 0.1177, 0.0100, 0.0116, 0.0401, 0.0557], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0084, 0.0082, 0.0143, 0.0070, 0.0074, 0.0114, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 03:15:45,878 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.092e+02 2.820e+02 3.474e+02 4.196e+02 7.562e+02, threshold=6.948e+02, percent-clipped=1.0 2023-04-28 03:15:51,970 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:15:53,755 INFO [train.py:904] (5/8) Epoch 4, batch 4800, loss[loss=0.2565, simple_loss=0.3348, pruned_loss=0.08908, over 16309.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3111, pruned_loss=0.07836, over 3216728.22 frames. ], batch size: 165, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:16:33,419 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:16:38,704 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:16:59,987 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:05,716 INFO [train.py:904] (5/8) Epoch 4, batch 4850, loss[loss=0.2029, simple_loss=0.2967, pruned_loss=0.05458, over 16843.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3117, pruned_loss=0.07748, over 3216157.75 frames. ], batch size: 96, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:17:37,856 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6741, 2.7780, 1.6707, 2.8274, 2.1173, 2.8316, 1.9157, 2.4259], device='cuda:5'), covar=tensor([0.0149, 0.0267, 0.1192, 0.0061, 0.0684, 0.0389, 0.1070, 0.0533], device='cuda:5'), in_proj_covar=tensor([0.0107, 0.0145, 0.0176, 0.0080, 0.0162, 0.0170, 0.0184, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 03:17:41,861 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:48,532 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:18:11,916 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35345.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:18:12,515 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 3.003e+02 3.470e+02 4.055e+02 1.018e+03, threshold=6.940e+02, percent-clipped=2.0 2023-04-28 03:18:19,086 INFO [train.py:904] (5/8) Epoch 4, batch 4900, loss[loss=0.2321, simple_loss=0.3052, pruned_loss=0.07948, over 16689.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3118, pruned_loss=0.07719, over 3191081.35 frames. ], batch size: 57, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:18:29,622 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 03:18:42,112 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:18:53,733 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:19:24,665 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4976, 3.4934, 2.8861, 2.2103, 2.6401, 2.2152, 3.6781, 3.8239], device='cuda:5'), covar=tensor([0.1884, 0.0521, 0.1076, 0.1258, 0.1717, 0.1164, 0.0351, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0267, 0.0238, 0.0254, 0.0225, 0.0292, 0.0189, 0.0224, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:19:32,923 INFO [train.py:904] (5/8) Epoch 4, batch 4950, loss[loss=0.238, simple_loss=0.324, pruned_loss=0.07597, over 16423.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3127, pruned_loss=0.07729, over 3188632.69 frames. ], batch size: 68, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:19:41,913 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:19:53,434 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:20:02,836 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8437, 3.7302, 3.8913, 4.0875, 4.1518, 3.7158, 4.0576, 4.1691], device='cuda:5'), covar=tensor([0.0750, 0.0656, 0.1061, 0.0445, 0.0381, 0.1110, 0.0517, 0.0349], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0394, 0.0527, 0.0407, 0.0303, 0.0293, 0.0319, 0.0325], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:20:22,839 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:37,435 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 3.403e+02 3.998e+02 5.018e+02 8.259e+02, threshold=7.996e+02, percent-clipped=7.0 2023-04-28 03:20:41,767 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:45,623 INFO [train.py:904] (5/8) Epoch 4, batch 5000, loss[loss=0.2052, simple_loss=0.2936, pruned_loss=0.05835, over 16909.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.315, pruned_loss=0.07796, over 3187810.36 frames. ], batch size: 109, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:21:57,694 INFO [train.py:904] (5/8) Epoch 4, batch 5050, loss[loss=0.2274, simple_loss=0.3107, pruned_loss=0.07209, over 16614.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3145, pruned_loss=0.07688, over 3206872.42 frames. ], batch size: 62, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:22:11,881 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6392, 2.6559, 2.1094, 4.1507, 3.7276, 3.8073, 1.4078, 2.8305], device='cuda:5'), covar=tensor([0.1458, 0.0575, 0.1383, 0.0071, 0.0200, 0.0294, 0.1440, 0.0839], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0136, 0.0163, 0.0075, 0.0150, 0.0158, 0.0155, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 03:22:39,996 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:22:54,069 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 03:23:03,499 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.171e+02 3.027e+02 3.520e+02 4.379e+02 8.983e+02, threshold=7.040e+02, percent-clipped=1.0 2023-04-28 03:23:10,569 INFO [train.py:904] (5/8) Epoch 4, batch 5100, loss[loss=0.2096, simple_loss=0.2902, pruned_loss=0.06453, over 16647.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.312, pruned_loss=0.0756, over 3205336.53 frames. ], batch size: 62, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:23:54,277 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 03:24:08,121 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:24:22,835 INFO [train.py:904] (5/8) Epoch 4, batch 5150, loss[loss=0.2325, simple_loss=0.3204, pruned_loss=0.07226, over 16711.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3123, pruned_loss=0.07509, over 3196007.41 frames. ], batch size: 124, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:24:26,049 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 03:25:27,207 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8325, 4.7215, 4.7321, 3.8748, 4.5897, 1.8867, 4.3998, 4.6830], device='cuda:5'), covar=tensor([0.0060, 0.0053, 0.0062, 0.0405, 0.0065, 0.1493, 0.0091, 0.0119], device='cuda:5'), in_proj_covar=tensor([0.0080, 0.0071, 0.0105, 0.0119, 0.0080, 0.0126, 0.0094, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:25:29,064 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.054e+02 2.805e+02 3.383e+02 4.164e+02 1.002e+03, threshold=6.766e+02, percent-clipped=7.0 2023-04-28 03:25:36,104 INFO [train.py:904] (5/8) Epoch 4, batch 5200, loss[loss=0.2168, simple_loss=0.2889, pruned_loss=0.07232, over 16301.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3099, pruned_loss=0.0743, over 3203538.09 frames. ], batch size: 35, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:26:46,337 INFO [train.py:904] (5/8) Epoch 4, batch 5250, loss[loss=0.2182, simple_loss=0.292, pruned_loss=0.07223, over 16622.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3072, pruned_loss=0.07399, over 3217933.31 frames. ], batch size: 57, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:26:47,351 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:27:28,649 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:27:52,068 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.886e+02 3.406e+02 4.102e+02 7.546e+02, threshold=6.813e+02, percent-clipped=1.0 2023-04-28 03:27:55,680 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:00,139 INFO [train.py:904] (5/8) Epoch 4, batch 5300, loss[loss=0.2043, simple_loss=0.2783, pruned_loss=0.06512, over 16947.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3044, pruned_loss=0.07296, over 3215611.10 frames. ], batch size: 109, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:28:28,862 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:47,667 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2363, 4.2710, 4.1667, 2.8269, 3.7802, 4.1700, 4.0174, 1.8740], device='cuda:5'), covar=tensor([0.0298, 0.0012, 0.0019, 0.0202, 0.0037, 0.0031, 0.0030, 0.0317], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0050, 0.0055, 0.0110, 0.0058, 0.0064, 0.0058, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 03:29:02,751 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:11,335 INFO [train.py:904] (5/8) Epoch 4, batch 5350, loss[loss=0.231, simple_loss=0.3171, pruned_loss=0.07244, over 16858.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3022, pruned_loss=0.07177, over 3220179.05 frames. ], batch size: 102, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:29:24,900 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:42,353 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5176, 3.3520, 2.6521, 2.0909, 2.5034, 2.0941, 3.5304, 3.5853], device='cuda:5'), covar=tensor([0.1731, 0.0608, 0.0980, 0.1266, 0.1436, 0.1231, 0.0353, 0.0440], device='cuda:5'), in_proj_covar=tensor([0.0268, 0.0240, 0.0253, 0.0227, 0.0295, 0.0192, 0.0228, 0.0231], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:29:56,975 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:17,965 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 3.081e+02 3.703e+02 4.509e+02 8.729e+02, threshold=7.406e+02, percent-clipped=5.0 2023-04-28 03:30:23,349 INFO [train.py:904] (5/8) Epoch 4, batch 5400, loss[loss=0.3209, simple_loss=0.3757, pruned_loss=0.133, over 11877.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3054, pruned_loss=0.07283, over 3233011.28 frames. ], batch size: 248, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:30:53,728 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:13,889 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:15,596 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 03:31:20,404 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6554, 2.7144, 2.5258, 1.5210, 2.7685, 2.7769, 2.4509, 2.0931], device='cuda:5'), covar=tensor([0.1047, 0.0136, 0.0181, 0.1104, 0.0096, 0.0111, 0.0375, 0.0541], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0083, 0.0077, 0.0137, 0.0067, 0.0072, 0.0110, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 03:31:23,359 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7236, 3.6799, 4.1811, 4.1234, 4.1079, 3.6663, 3.8321, 3.7384], device='cuda:5'), covar=tensor([0.0263, 0.0423, 0.0267, 0.0369, 0.0435, 0.0365, 0.0662, 0.0429], device='cuda:5'), in_proj_covar=tensor([0.0209, 0.0201, 0.0209, 0.0216, 0.0253, 0.0225, 0.0316, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 03:31:38,580 INFO [train.py:904] (5/8) Epoch 4, batch 5450, loss[loss=0.2727, simple_loss=0.3433, pruned_loss=0.101, over 16913.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3098, pruned_loss=0.07559, over 3229816.07 frames. ], batch size: 109, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:32:50,258 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.590e+02 3.796e+02 5.362e+02 7.541e+02 3.116e+03, threshold=1.072e+03, percent-clipped=25.0 2023-04-28 03:32:56,436 INFO [train.py:904] (5/8) Epoch 4, batch 5500, loss[loss=0.3033, simple_loss=0.3635, pruned_loss=0.1216, over 16392.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3205, pruned_loss=0.08363, over 3200917.16 frames. ], batch size: 146, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:33:19,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8421, 2.6390, 2.4334, 4.4863, 1.9958, 4.1009, 2.5833, 2.5590], device='cuda:5'), covar=tensor([0.0443, 0.1260, 0.0717, 0.0208, 0.2333, 0.0358, 0.1191, 0.1722], device='cuda:5'), in_proj_covar=tensor([0.0294, 0.0279, 0.0231, 0.0290, 0.0346, 0.0251, 0.0257, 0.0344], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:34:16,680 INFO [train.py:904] (5/8) Epoch 4, batch 5550, loss[loss=0.2449, simple_loss=0.3262, pruned_loss=0.08179, over 17246.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3298, pruned_loss=0.09099, over 3171800.29 frames. ], batch size: 52, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:34:17,791 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:35:02,272 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:35:29,649 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.087e+02 4.484e+02 5.536e+02 7.020e+02 1.224e+03, threshold=1.107e+03, percent-clipped=3.0 2023-04-28 03:35:33,036 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:35:35,563 INFO [train.py:904] (5/8) Epoch 4, batch 5600, loss[loss=0.2698, simple_loss=0.3439, pruned_loss=0.09783, over 16535.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.3373, pruned_loss=0.0982, over 3120104.16 frames. ], batch size: 75, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:35:48,041 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 03:36:20,422 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:36:47,118 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8506, 2.7074, 2.2558, 3.7412, 3.3791, 3.6187, 1.5656, 2.8213], device='cuda:5'), covar=tensor([0.1185, 0.0399, 0.1050, 0.0082, 0.0201, 0.0287, 0.1187, 0.0626], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0133, 0.0160, 0.0072, 0.0147, 0.0152, 0.0150, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 03:36:57,495 INFO [train.py:904] (5/8) Epoch 4, batch 5650, loss[loss=0.2596, simple_loss=0.3321, pruned_loss=0.09358, over 16573.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3441, pruned_loss=0.1043, over 3076531.74 frames. ], batch size: 62, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:37:41,237 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:09,921 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.856e+02 5.092e+02 6.431e+02 7.893e+02 1.649e+03, threshold=1.286e+03, percent-clipped=5.0 2023-04-28 03:38:17,767 INFO [train.py:904] (5/8) Epoch 4, batch 5700, loss[loss=0.3547, simple_loss=0.3861, pruned_loss=0.1617, over 11333.00 frames. ], tot_loss[loss=0.2803, simple_loss=0.3467, pruned_loss=0.107, over 3051142.85 frames. ], batch size: 248, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:38:42,678 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:05,541 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:13,476 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:36,195 INFO [train.py:904] (5/8) Epoch 4, batch 5750, loss[loss=0.2813, simple_loss=0.3522, pruned_loss=0.1052, over 16847.00 frames. ], tot_loss[loss=0.2825, simple_loss=0.3487, pruned_loss=0.1081, over 3032024.53 frames. ], batch size: 116, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:40:29,641 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:40:42,464 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36242.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:40:49,465 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.564e+02 4.215e+02 5.340e+02 6.535e+02 1.180e+03, threshold=1.068e+03, percent-clipped=0.0 2023-04-28 03:40:55,828 INFO [train.py:904] (5/8) Epoch 4, batch 5800, loss[loss=0.2917, simple_loss=0.36, pruned_loss=0.1117, over 15291.00 frames. ], tot_loss[loss=0.28, simple_loss=0.3481, pruned_loss=0.106, over 3043784.14 frames. ], batch size: 191, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:41:34,526 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7587, 2.3051, 1.5391, 1.9909, 3.0375, 2.8579, 3.6030, 3.2924], device='cuda:5'), covar=tensor([0.0013, 0.0149, 0.0230, 0.0185, 0.0071, 0.0126, 0.0050, 0.0067], device='cuda:5'), in_proj_covar=tensor([0.0067, 0.0138, 0.0140, 0.0137, 0.0129, 0.0141, 0.0100, 0.0118], device='cuda:5'), out_proj_covar=tensor([8.6631e-05, 1.8179e-04, 1.7757e-04, 1.7668e-04, 1.7137e-04, 1.8605e-04, 1.2875e-04, 1.5666e-04], device='cuda:5') 2023-04-28 03:41:53,464 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9858, 4.8453, 4.7540, 4.1049, 4.8384, 1.8614, 4.5459, 4.8568], device='cuda:5'), covar=tensor([0.0052, 0.0055, 0.0074, 0.0313, 0.0051, 0.1492, 0.0079, 0.0099], device='cuda:5'), in_proj_covar=tensor([0.0079, 0.0069, 0.0106, 0.0117, 0.0079, 0.0127, 0.0092, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:42:03,017 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8976, 3.8368, 4.4072, 4.3889, 4.3660, 3.9867, 4.0889, 3.9792], device='cuda:5'), covar=tensor([0.0265, 0.0371, 0.0364, 0.0427, 0.0442, 0.0302, 0.0762, 0.0388], device='cuda:5'), in_proj_covar=tensor([0.0204, 0.0193, 0.0204, 0.0211, 0.0245, 0.0214, 0.0309, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 03:42:12,367 INFO [train.py:904] (5/8) Epoch 4, batch 5850, loss[loss=0.2373, simple_loss=0.3142, pruned_loss=0.08021, over 16627.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.3446, pruned_loss=0.1031, over 3064342.37 frames. ], batch size: 62, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:43:17,910 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-28 03:43:25,143 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5848, 3.4804, 2.8788, 2.3061, 2.5311, 2.0696, 3.5235, 3.6958], device='cuda:5'), covar=tensor([0.1947, 0.0623, 0.1091, 0.1326, 0.1859, 0.1342, 0.0373, 0.0433], device='cuda:5'), in_proj_covar=tensor([0.0276, 0.0245, 0.0263, 0.0232, 0.0306, 0.0197, 0.0231, 0.0233], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:43:29,073 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.296e+02 4.315e+02 5.493e+02 6.918e+02 1.909e+03, threshold=1.099e+03, percent-clipped=3.0 2023-04-28 03:43:34,006 INFO [train.py:904] (5/8) Epoch 4, batch 5900, loss[loss=0.3067, simple_loss=0.3554, pruned_loss=0.129, over 11500.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3442, pruned_loss=0.1025, over 3078083.32 frames. ], batch size: 247, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:44:56,224 INFO [train.py:904] (5/8) Epoch 4, batch 5950, loss[loss=0.2705, simple_loss=0.3402, pruned_loss=0.1005, over 17182.00 frames. ], tot_loss[loss=0.2742, simple_loss=0.3457, pruned_loss=0.1014, over 3078406.67 frames. ], batch size: 44, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:45:37,489 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:46:10,129 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.426e+02 4.337e+02 5.242e+02 6.578e+02 1.706e+03, threshold=1.048e+03, percent-clipped=3.0 2023-04-28 03:46:14,586 INFO [train.py:904] (5/8) Epoch 4, batch 6000, loss[loss=0.2668, simple_loss=0.3344, pruned_loss=0.09959, over 16247.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3445, pruned_loss=0.1005, over 3088382.93 frames. ], batch size: 165, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:46:14,587 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 03:46:23,191 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9149, 4.6634, 4.6677, 3.9998, 4.6183, 2.0213, 4.5670, 4.6521], device='cuda:5'), covar=tensor([0.0034, 0.0044, 0.0059, 0.0242, 0.0042, 0.1430, 0.0054, 0.0115], device='cuda:5'), in_proj_covar=tensor([0.0080, 0.0069, 0.0106, 0.0117, 0.0079, 0.0127, 0.0093, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:46:25,219 INFO [train.py:938] (5/8) Epoch 4, validation: loss=0.1966, simple_loss=0.3069, pruned_loss=0.04309, over 944034.00 frames. 2023-04-28 03:46:25,220 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 03:46:49,568 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:01,432 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:44,077 INFO [train.py:904] (5/8) Epoch 4, batch 6050, loss[loss=0.2337, simple_loss=0.3152, pruned_loss=0.07609, over 16634.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3428, pruned_loss=0.0995, over 3083837.34 frames. ], batch size: 134, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:48:06,440 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:42,935 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:58,313 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.180e+02 4.133e+02 5.686e+02 7.938e+02 1.625e+03, threshold=1.137e+03, percent-clipped=9.0 2023-04-28 03:49:03,039 INFO [train.py:904] (5/8) Epoch 4, batch 6100, loss[loss=0.2717, simple_loss=0.3338, pruned_loss=0.1048, over 16867.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3413, pruned_loss=0.09796, over 3096002.66 frames. ], batch size: 109, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:50:22,077 INFO [train.py:904] (5/8) Epoch 4, batch 6150, loss[loss=0.2611, simple_loss=0.3343, pruned_loss=0.09398, over 16234.00 frames. ], tot_loss[loss=0.2663, simple_loss=0.3392, pruned_loss=0.09667, over 3101912.72 frames. ], batch size: 165, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:50:23,629 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 03:50:29,641 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:35,008 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:41,754 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:51:38,879 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.478e+02 3.823e+02 4.755e+02 6.026e+02 1.290e+03, threshold=9.509e+02, percent-clipped=1.0 2023-04-28 03:51:44,322 INFO [train.py:904] (5/8) Epoch 4, batch 6200, loss[loss=0.3403, simple_loss=0.3709, pruned_loss=0.1549, over 11365.00 frames. ], tot_loss[loss=0.2654, simple_loss=0.3376, pruned_loss=0.09662, over 3088095.05 frames. ], batch size: 246, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:52:05,965 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4059, 1.9229, 1.5206, 1.6714, 2.3612, 2.1017, 2.4891, 2.5504], device='cuda:5'), covar=tensor([0.0028, 0.0163, 0.0236, 0.0214, 0.0091, 0.0158, 0.0063, 0.0092], device='cuda:5'), in_proj_covar=tensor([0.0068, 0.0139, 0.0144, 0.0140, 0.0132, 0.0143, 0.0101, 0.0121], device='cuda:5'), out_proj_covar=tensor([8.7917e-05, 1.8176e-04, 1.8255e-04, 1.8021e-04, 1.7447e-04, 1.8733e-04, 1.2877e-04, 1.6076e-04], device='cuda:5') 2023-04-28 03:52:07,230 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:11,534 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:17,442 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:57,892 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:52:59,112 INFO [train.py:904] (5/8) Epoch 4, batch 6250, loss[loss=0.2833, simple_loss=0.3672, pruned_loss=0.09967, over 16636.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3373, pruned_loss=0.09611, over 3108070.63 frames. ], batch size: 76, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:53:20,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 03:54:09,141 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.662e+02 4.271e+02 5.056e+02 6.233e+02 1.402e+03, threshold=1.011e+03, percent-clipped=5.0 2023-04-28 03:54:14,291 INFO [train.py:904] (5/8) Epoch 4, batch 6300, loss[loss=0.2489, simple_loss=0.3256, pruned_loss=0.08606, over 16467.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3368, pruned_loss=0.09479, over 3118769.74 frames. ], batch size: 146, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:54:31,257 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:55:32,007 INFO [train.py:904] (5/8) Epoch 4, batch 6350, loss[loss=0.2999, simple_loss=0.3662, pruned_loss=0.1168, over 16331.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3382, pruned_loss=0.09666, over 3110088.09 frames. ], batch size: 165, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:55:38,799 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:56:28,099 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:43,688 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.626e+02 4.373e+02 5.288e+02 7.014e+02 2.020e+03, threshold=1.058e+03, percent-clipped=6.0 2023-04-28 03:56:44,636 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 03:56:48,093 INFO [train.py:904] (5/8) Epoch 4, batch 6400, loss[loss=0.2259, simple_loss=0.3122, pruned_loss=0.06982, over 16891.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3373, pruned_loss=0.09619, over 3124166.25 frames. ], batch size: 116, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:57:05,776 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 03:57:10,829 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:57:27,294 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 03:57:40,005 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:04,376 INFO [train.py:904] (5/8) Epoch 4, batch 6450, loss[loss=0.2464, simple_loss=0.3046, pruned_loss=0.09409, over 11588.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3352, pruned_loss=0.09426, over 3120621.89 frames. ], batch size: 247, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:58:20,112 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:50,853 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1155, 5.4419, 5.1390, 5.1766, 4.7898, 4.5542, 4.9784, 5.4910], device='cuda:5'), covar=tensor([0.0524, 0.0559, 0.0812, 0.0402, 0.0560, 0.0589, 0.0519, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0328, 0.0438, 0.0381, 0.0290, 0.0278, 0.0287, 0.0355, 0.0305], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 03:59:18,067 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.485e+02 3.772e+02 4.534e+02 6.144e+02 1.455e+03, threshold=9.067e+02, percent-clipped=1.0 2023-04-28 03:59:23,793 INFO [train.py:904] (5/8) Epoch 4, batch 6500, loss[loss=0.2521, simple_loss=0.3345, pruned_loss=0.08489, over 16832.00 frames. ], tot_loss[loss=0.2594, simple_loss=0.3322, pruned_loss=0.09331, over 3112060.75 frames. ], batch size: 96, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:59:35,217 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:39,316 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:44,060 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:50,179 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:56,674 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:00:44,420 INFO [train.py:904] (5/8) Epoch 4, batch 6550, loss[loss=0.2683, simple_loss=0.3561, pruned_loss=0.09027, over 15463.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3355, pruned_loss=0.09418, over 3135186.84 frames. ], batch size: 191, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:00:50,611 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:00,748 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 04:01:13,075 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:56,567 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.603e+02 4.116e+02 5.104e+02 6.571e+02 1.425e+03, threshold=1.021e+03, percent-clipped=7.0 2023-04-28 04:02:01,492 INFO [train.py:904] (5/8) Epoch 4, batch 6600, loss[loss=0.2967, simple_loss=0.357, pruned_loss=0.1182, over 15466.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3386, pruned_loss=0.09576, over 3129927.58 frames. ], batch size: 190, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:02:06,624 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5013, 3.6718, 1.2665, 3.8412, 2.3013, 3.8173, 1.6473, 2.7026], device='cuda:5'), covar=tensor([0.0127, 0.0231, 0.1988, 0.0047, 0.0745, 0.0333, 0.1803, 0.0521], device='cuda:5'), in_proj_covar=tensor([0.0108, 0.0145, 0.0174, 0.0076, 0.0158, 0.0174, 0.0184, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 04:02:09,609 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:02:23,369 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:54,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8881, 1.5432, 2.0463, 2.5282, 2.7984, 3.0007, 1.7045, 2.9881], device='cuda:5'), covar=tensor([0.0051, 0.0212, 0.0152, 0.0113, 0.0085, 0.0073, 0.0200, 0.0056], device='cuda:5'), in_proj_covar=tensor([0.0102, 0.0131, 0.0120, 0.0112, 0.0117, 0.0082, 0.0130, 0.0076], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 04:03:19,975 INFO [train.py:904] (5/8) Epoch 4, batch 6650, loss[loss=0.3561, simple_loss=0.3949, pruned_loss=0.1587, over 11575.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3393, pruned_loss=0.0971, over 3118701.68 frames. ], batch size: 247, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:03:31,858 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1663, 2.1877, 1.9956, 3.5100, 1.7927, 3.0127, 2.1068, 1.9728], device='cuda:5'), covar=tensor([0.0552, 0.1398, 0.0899, 0.0342, 0.2483, 0.0634, 0.1513, 0.2039], device='cuda:5'), in_proj_covar=tensor([0.0291, 0.0279, 0.0232, 0.0291, 0.0343, 0.0255, 0.0256, 0.0344], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:03:57,703 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2827, 4.3226, 4.3118, 4.4464, 4.3529, 4.9371, 4.5189, 4.2511], device='cuda:5'), covar=tensor([0.1074, 0.1523, 0.1319, 0.1621, 0.2284, 0.0798, 0.1059, 0.2125], device='cuda:5'), in_proj_covar=tensor([0.0254, 0.0353, 0.0339, 0.0312, 0.0403, 0.0366, 0.0283, 0.0412], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 04:04:09,443 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9650, 2.9406, 2.5969, 4.6752, 4.1827, 4.2258, 1.7362, 3.2743], device='cuda:5'), covar=tensor([0.1215, 0.0500, 0.1065, 0.0056, 0.0239, 0.0237, 0.1277, 0.0603], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0136, 0.0163, 0.0074, 0.0156, 0.0157, 0.0155, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 04:04:14,669 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 04:04:33,024 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.678e+02 4.226e+02 5.482e+02 6.973e+02 1.276e+03, threshold=1.096e+03, percent-clipped=4.0 2023-04-28 04:04:37,779 INFO [train.py:904] (5/8) Epoch 4, batch 6700, loss[loss=0.2458, simple_loss=0.3207, pruned_loss=0.08546, over 16723.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3381, pruned_loss=0.09735, over 3110663.52 frames. ], batch size: 124, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:04:53,203 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:05:27,840 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-28 04:05:53,502 INFO [train.py:904] (5/8) Epoch 4, batch 6750, loss[loss=0.3219, simple_loss=0.3729, pruned_loss=0.1355, over 11828.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3379, pruned_loss=0.09816, over 3089845.63 frames. ], batch size: 248, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:06:15,615 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2052, 1.4314, 1.7185, 2.1532, 2.2510, 2.4805, 1.3882, 2.5068], device='cuda:5'), covar=tensor([0.0072, 0.0201, 0.0147, 0.0099, 0.0104, 0.0060, 0.0218, 0.0044], device='cuda:5'), in_proj_covar=tensor([0.0102, 0.0131, 0.0120, 0.0111, 0.0119, 0.0082, 0.0131, 0.0075], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 04:06:58,130 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4622, 4.3262, 4.2412, 3.2555, 4.1436, 1.5302, 4.0162, 4.0837], device='cuda:5'), covar=tensor([0.0059, 0.0054, 0.0082, 0.0407, 0.0069, 0.1769, 0.0093, 0.0157], device='cuda:5'), in_proj_covar=tensor([0.0082, 0.0071, 0.0109, 0.0118, 0.0080, 0.0132, 0.0096, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:07:06,103 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.556e+02 4.104e+02 5.227e+02 6.545e+02 1.583e+03, threshold=1.045e+03, percent-clipped=5.0 2023-04-28 04:07:10,567 INFO [train.py:904] (5/8) Epoch 4, batch 6800, loss[loss=0.2562, simple_loss=0.3327, pruned_loss=0.08985, over 17276.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3372, pruned_loss=0.09725, over 3091063.69 frames. ], batch size: 52, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:07:27,130 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:31,554 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:35,438 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:37,348 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:37,418 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:27,086 INFO [train.py:904] (5/8) Epoch 4, batch 6850, loss[loss=0.2433, simple_loss=0.3362, pruned_loss=0.07517, over 16767.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.339, pruned_loss=0.09774, over 3097708.92 frames. ], batch size: 39, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:08:38,765 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:43,641 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:46,874 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:49,190 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:08,867 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:36,166 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.986e+02 4.056e+02 4.810e+02 6.733e+02 1.178e+03, threshold=9.620e+02, percent-clipped=3.0 2023-04-28 04:09:39,838 INFO [train.py:904] (5/8) Epoch 4, batch 6900, loss[loss=0.3419, simple_loss=0.3823, pruned_loss=0.1507, over 11499.00 frames. ], tot_loss[loss=0.2696, simple_loss=0.3423, pruned_loss=0.09842, over 3088089.69 frames. ], batch size: 248, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:09:48,102 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:09:53,442 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:59,804 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:10:42,535 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 04:10:55,362 INFO [train.py:904] (5/8) Epoch 4, batch 6950, loss[loss=0.2717, simple_loss=0.344, pruned_loss=0.09969, over 16802.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3453, pruned_loss=0.1018, over 3065872.98 frames. ], batch size: 124, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:10:57,973 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 04:11:00,663 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:11:32,996 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:12:10,367 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.259e+02 4.534e+02 5.752e+02 7.141e+02 1.491e+03, threshold=1.150e+03, percent-clipped=11.0 2023-04-28 04:12:11,787 INFO [train.py:904] (5/8) Epoch 4, batch 7000, loss[loss=0.2625, simple_loss=0.3426, pruned_loss=0.09121, over 16373.00 frames. ], tot_loss[loss=0.2736, simple_loss=0.345, pruned_loss=0.1011, over 3065993.10 frames. ], batch size: 35, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:12:28,109 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 04:13:29,632 INFO [train.py:904] (5/8) Epoch 4, batch 7050, loss[loss=0.2688, simple_loss=0.3459, pruned_loss=0.09582, over 17058.00 frames. ], tot_loss[loss=0.2731, simple_loss=0.3453, pruned_loss=0.1005, over 3064123.12 frames. ], batch size: 50, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:13:35,740 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 04:13:43,151 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:14:12,776 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1343, 2.2470, 2.1863, 3.4708, 1.8405, 3.0959, 2.2560, 2.0372], device='cuda:5'), covar=tensor([0.0537, 0.1445, 0.0774, 0.0304, 0.2463, 0.0527, 0.1442, 0.1944], device='cuda:5'), in_proj_covar=tensor([0.0296, 0.0281, 0.0235, 0.0295, 0.0350, 0.0260, 0.0259, 0.0349], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:14:45,396 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 3.107e+02 4.211e+02 5.076e+02 6.594e+02 1.566e+03, threshold=1.015e+03, percent-clipped=2.0 2023-04-28 04:14:46,717 INFO [train.py:904] (5/8) Epoch 4, batch 7100, loss[loss=0.2758, simple_loss=0.3594, pruned_loss=0.09609, over 16686.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.3435, pruned_loss=0.09983, over 3069331.62 frames. ], batch size: 89, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:15:13,102 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:15:44,552 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 04:15:57,163 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5282, 2.2524, 2.2434, 4.1042, 1.9190, 3.4027, 2.3647, 2.3267], device='cuda:5'), covar=tensor([0.0506, 0.1495, 0.0815, 0.0261, 0.2551, 0.0554, 0.1416, 0.1967], device='cuda:5'), in_proj_covar=tensor([0.0297, 0.0284, 0.0236, 0.0297, 0.0353, 0.0264, 0.0262, 0.0353], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:16:02,659 INFO [train.py:904] (5/8) Epoch 4, batch 7150, loss[loss=0.2537, simple_loss=0.3214, pruned_loss=0.09299, over 16652.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3404, pruned_loss=0.09845, over 3081039.27 frames. ], batch size: 57, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:16:23,286 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:24,832 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:37,699 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 04:16:38,275 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37624.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:07,259 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5250, 2.2754, 2.1508, 4.0817, 1.8123, 3.2027, 2.2255, 2.1521], device='cuda:5'), covar=tensor([0.0572, 0.1561, 0.0901, 0.0276, 0.2706, 0.0673, 0.1601, 0.2161], device='cuda:5'), in_proj_covar=tensor([0.0296, 0.0282, 0.0235, 0.0296, 0.0350, 0.0263, 0.0261, 0.0350], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:17:17,126 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.158e+02 3.757e+02 4.855e+02 6.210e+02 1.451e+03, threshold=9.710e+02, percent-clipped=1.0 2023-04-28 04:17:18,898 INFO [train.py:904] (5/8) Epoch 4, batch 7200, loss[loss=0.2447, simple_loss=0.3287, pruned_loss=0.08034, over 16445.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3377, pruned_loss=0.09607, over 3080937.75 frames. ], batch size: 146, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:17:33,289 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:36,207 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:18:40,024 INFO [train.py:904] (5/8) Epoch 4, batch 7250, loss[loss=0.2272, simple_loss=0.3121, pruned_loss=0.07115, over 16878.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3343, pruned_loss=0.0941, over 3069562.26 frames. ], batch size: 102, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:18:52,050 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:10,390 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:55,473 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 3.869e+02 4.810e+02 5.847e+02 1.256e+03, threshold=9.620e+02, percent-clipped=3.0 2023-04-28 04:19:57,424 INFO [train.py:904] (5/8) Epoch 4, batch 7300, loss[loss=0.284, simple_loss=0.3515, pruned_loss=0.1082, over 15158.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3328, pruned_loss=0.09316, over 3092385.66 frames. ], batch size: 190, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:20:33,100 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 04:21:14,983 INFO [train.py:904] (5/8) Epoch 4, batch 7350, loss[loss=0.2283, simple_loss=0.3082, pruned_loss=0.07419, over 16579.00 frames. ], tot_loss[loss=0.2589, simple_loss=0.3321, pruned_loss=0.09282, over 3093585.60 frames. ], batch size: 75, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:21:16,679 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3727, 4.1501, 4.0822, 1.7433, 4.3715, 4.4151, 2.9165, 3.0965], device='cuda:5'), covar=tensor([0.0874, 0.0082, 0.0160, 0.1184, 0.0035, 0.0031, 0.0353, 0.0413], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0082, 0.0082, 0.0140, 0.0069, 0.0072, 0.0112, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 04:21:18,840 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 04:21:27,310 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9787, 2.4498, 2.2557, 3.2053, 2.8088, 3.2232, 1.8357, 2.6270], device='cuda:5'), covar=tensor([0.1056, 0.0452, 0.0916, 0.0097, 0.0242, 0.0276, 0.1058, 0.0627], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0136, 0.0164, 0.0075, 0.0157, 0.0158, 0.0156, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 04:21:51,003 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 04:21:55,021 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9102, 3.6430, 3.2793, 1.6833, 2.7021, 2.1794, 3.2003, 3.5112], device='cuda:5'), covar=tensor([0.0270, 0.0497, 0.0486, 0.1742, 0.0818, 0.0968, 0.0702, 0.0606], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0123, 0.0156, 0.0143, 0.0137, 0.0128, 0.0144, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 04:22:29,893 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.483e+02 3.907e+02 4.910e+02 6.467e+02 1.073e+03, threshold=9.820e+02, percent-clipped=2.0 2023-04-28 04:22:31,814 INFO [train.py:904] (5/8) Epoch 4, batch 7400, loss[loss=0.2379, simple_loss=0.3199, pruned_loss=0.07792, over 17208.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3335, pruned_loss=0.0936, over 3102711.48 frames. ], batch size: 45, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:22:40,571 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9986, 2.3185, 1.6681, 1.9883, 2.7893, 2.6491, 3.1892, 3.0890], device='cuda:5'), covar=tensor([0.0025, 0.0137, 0.0207, 0.0175, 0.0074, 0.0121, 0.0045, 0.0068], device='cuda:5'), in_proj_covar=tensor([0.0065, 0.0136, 0.0142, 0.0137, 0.0129, 0.0142, 0.0101, 0.0119], device='cuda:5'), out_proj_covar=tensor([8.3212e-05, 1.7548e-04, 1.8035e-04, 1.7286e-04, 1.6689e-04, 1.8554e-04, 1.2663e-04, 1.5557e-04], device='cuda:5') 2023-04-28 04:23:22,394 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8708, 3.5854, 3.1347, 1.8948, 2.7081, 2.0791, 3.1863, 3.4247], device='cuda:5'), covar=tensor([0.0234, 0.0451, 0.0472, 0.1503, 0.0706, 0.0927, 0.0623, 0.0666], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0121, 0.0154, 0.0141, 0.0135, 0.0127, 0.0142, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 04:23:49,014 INFO [train.py:904] (5/8) Epoch 4, batch 7450, loss[loss=0.2501, simple_loss=0.335, pruned_loss=0.08265, over 16709.00 frames. ], tot_loss[loss=0.2623, simple_loss=0.3349, pruned_loss=0.09486, over 3096789.68 frames. ], batch size: 134, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:24:26,705 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37924.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:06,400 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.829e+02 4.545e+02 5.686e+02 6.937e+02 1.455e+03, threshold=1.137e+03, percent-clipped=4.0 2023-04-28 04:25:07,738 INFO [train.py:904] (5/8) Epoch 4, batch 7500, loss[loss=0.2478, simple_loss=0.3299, pruned_loss=0.08283, over 16497.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3365, pruned_loss=0.09537, over 3070999.57 frames. ], batch size: 35, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:25:39,745 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-04-28 04:25:40,571 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:49,131 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0546, 2.9763, 2.6046, 2.0346, 2.5230, 2.0936, 2.6874, 2.8058], device='cuda:5'), covar=tensor([0.0295, 0.0451, 0.0509, 0.1378, 0.0691, 0.0882, 0.0523, 0.0545], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0122, 0.0156, 0.0142, 0.0137, 0.0128, 0.0143, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 04:25:50,309 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:27,347 INFO [train.py:904] (5/8) Epoch 4, batch 7550, loss[loss=0.2639, simple_loss=0.3353, pruned_loss=0.09626, over 15297.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3352, pruned_loss=0.09519, over 3075429.93 frames. ], batch size: 190, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:26:54,996 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:24,033 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:37,942 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-28 04:27:38,148 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.661e+02 3.786e+02 4.899e+02 6.266e+02 1.464e+03, threshold=9.797e+02, percent-clipped=2.0 2023-04-28 04:27:40,073 INFO [train.py:904] (5/8) Epoch 4, batch 7600, loss[loss=0.2663, simple_loss=0.3387, pruned_loss=0.09696, over 16337.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3356, pruned_loss=0.09623, over 3078892.32 frames. ], batch size: 146, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:28:05,542 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:28:32,691 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 04:28:55,072 INFO [train.py:904] (5/8) Epoch 4, batch 7650, loss[loss=0.2342, simple_loss=0.307, pruned_loss=0.0807, over 17024.00 frames. ], tot_loss[loss=0.2663, simple_loss=0.3373, pruned_loss=0.09762, over 3079007.84 frames. ], batch size: 55, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:28:56,113 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8207, 4.7312, 4.5796, 3.8501, 4.5309, 1.6967, 4.3982, 4.5464], device='cuda:5'), covar=tensor([0.0053, 0.0047, 0.0082, 0.0296, 0.0053, 0.1576, 0.0065, 0.0116], device='cuda:5'), in_proj_covar=tensor([0.0082, 0.0071, 0.0109, 0.0118, 0.0080, 0.0132, 0.0095, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:30:08,820 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.664e+02 4.200e+02 5.231e+02 7.341e+02 1.286e+03, threshold=1.046e+03, percent-clipped=4.0 2023-04-28 04:30:09,982 INFO [train.py:904] (5/8) Epoch 4, batch 7700, loss[loss=0.2632, simple_loss=0.3405, pruned_loss=0.09292, over 16740.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3383, pruned_loss=0.09873, over 3074548.39 frames. ], batch size: 89, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:31:26,749 INFO [train.py:904] (5/8) Epoch 4, batch 7750, loss[loss=0.3264, simple_loss=0.364, pruned_loss=0.1444, over 11561.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.338, pruned_loss=0.09813, over 3083591.48 frames. ], batch size: 248, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:31:42,439 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5063, 2.2363, 1.7716, 1.8964, 2.6155, 2.4449, 2.9792, 2.8100], device='cuda:5'), covar=tensor([0.0048, 0.0171, 0.0198, 0.0211, 0.0090, 0.0140, 0.0059, 0.0095], device='cuda:5'), in_proj_covar=tensor([0.0068, 0.0139, 0.0143, 0.0139, 0.0130, 0.0143, 0.0103, 0.0120], device='cuda:5'), out_proj_covar=tensor([8.7605e-05, 1.8002e-04, 1.8098e-04, 1.7557e-04, 1.6862e-04, 1.8540e-04, 1.2856e-04, 1.5641e-04], device='cuda:5') 2023-04-28 04:32:12,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8724, 3.8419, 3.8231, 3.1793, 3.8184, 1.7116, 3.6464, 3.7630], device='cuda:5'), covar=tensor([0.0089, 0.0068, 0.0104, 0.0287, 0.0064, 0.1626, 0.0086, 0.0126], device='cuda:5'), in_proj_covar=tensor([0.0081, 0.0070, 0.0107, 0.0117, 0.0078, 0.0130, 0.0095, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:32:15,963 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0956, 1.5297, 2.2307, 2.7653, 2.6043, 3.0248, 1.6107, 3.1998], device='cuda:5'), covar=tensor([0.0047, 0.0218, 0.0142, 0.0106, 0.0090, 0.0074, 0.0213, 0.0052], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0130, 0.0117, 0.0111, 0.0115, 0.0083, 0.0130, 0.0075], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 04:32:40,401 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 4.651e+02 5.629e+02 7.358e+02 1.215e+03, threshold=1.126e+03, percent-clipped=5.0 2023-04-28 04:32:41,614 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8443, 1.4939, 2.0700, 2.6181, 2.5458, 2.8616, 1.6451, 2.8863], device='cuda:5'), covar=tensor([0.0056, 0.0226, 0.0151, 0.0107, 0.0089, 0.0087, 0.0197, 0.0062], device='cuda:5'), in_proj_covar=tensor([0.0102, 0.0131, 0.0118, 0.0113, 0.0116, 0.0084, 0.0131, 0.0075], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 04:32:42,177 INFO [train.py:904] (5/8) Epoch 4, batch 7800, loss[loss=0.2599, simple_loss=0.3404, pruned_loss=0.08972, over 16490.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3385, pruned_loss=0.09868, over 3082400.64 frames. ], batch size: 75, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:33:12,046 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1255, 3.0194, 2.6635, 2.0146, 2.5853, 2.1288, 2.6806, 2.9600], device='cuda:5'), covar=tensor([0.0277, 0.0373, 0.0432, 0.1366, 0.0580, 0.0816, 0.0507, 0.0473], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0122, 0.0153, 0.0141, 0.0133, 0.0127, 0.0141, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 04:33:58,994 INFO [train.py:904] (5/8) Epoch 4, batch 7850, loss[loss=0.2889, simple_loss=0.3565, pruned_loss=0.1107, over 15292.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.3392, pruned_loss=0.09865, over 3067295.26 frames. ], batch size: 191, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:34:30,566 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 04:34:50,570 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:07,639 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 04:35:12,447 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.779e+02 3.909e+02 4.709e+02 5.998e+02 1.164e+03, threshold=9.418e+02, percent-clipped=1.0 2023-04-28 04:35:14,728 INFO [train.py:904] (5/8) Epoch 4, batch 7900, loss[loss=0.2877, simple_loss=0.3592, pruned_loss=0.1081, over 15361.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3377, pruned_loss=0.0975, over 3072333.61 frames. ], batch size: 190, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:35:18,130 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8448, 4.1323, 3.8549, 3.9328, 3.6111, 3.6840, 3.8472, 4.0277], device='cuda:5'), covar=tensor([0.0589, 0.0628, 0.0865, 0.0488, 0.0555, 0.1205, 0.0558, 0.0788], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0430, 0.0379, 0.0281, 0.0279, 0.0294, 0.0356, 0.0309], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:35:28,532 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:36:34,674 INFO [train.py:904] (5/8) Epoch 4, batch 7950, loss[loss=0.2387, simple_loss=0.3147, pruned_loss=0.08137, over 16878.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.3388, pruned_loss=0.09892, over 3041708.22 frames. ], batch size: 42, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:37:05,065 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:37:49,212 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.619e+02 4.198e+02 4.896e+02 6.041e+02 1.381e+03, threshold=9.792e+02, percent-clipped=5.0 2023-04-28 04:37:50,962 INFO [train.py:904] (5/8) Epoch 4, batch 8000, loss[loss=0.2671, simple_loss=0.3483, pruned_loss=0.09297, over 17137.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3393, pruned_loss=0.09987, over 3026886.40 frames. ], batch size: 47, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:38:22,630 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 04:38:25,252 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5481, 4.2974, 4.3044, 2.9655, 3.8655, 4.2666, 4.0075, 2.1770], device='cuda:5'), covar=tensor([0.0324, 0.0020, 0.0031, 0.0227, 0.0039, 0.0054, 0.0036, 0.0311], device='cuda:5'), in_proj_covar=tensor([0.0113, 0.0049, 0.0056, 0.0110, 0.0057, 0.0065, 0.0059, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 04:39:04,878 INFO [train.py:904] (5/8) Epoch 4, batch 8050, loss[loss=0.2658, simple_loss=0.3454, pruned_loss=0.09308, over 16787.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3389, pruned_loss=0.09905, over 3029008.57 frames. ], batch size: 83, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:40:21,967 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.546e+02 4.083e+02 4.841e+02 6.395e+02 1.721e+03, threshold=9.682e+02, percent-clipped=6.0 2023-04-28 04:40:23,277 INFO [train.py:904] (5/8) Epoch 4, batch 8100, loss[loss=0.2896, simple_loss=0.3477, pruned_loss=0.1157, over 15414.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.338, pruned_loss=0.09791, over 3050233.56 frames. ], batch size: 190, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:41:41,540 INFO [train.py:904] (5/8) Epoch 4, batch 8150, loss[loss=0.237, simple_loss=0.3131, pruned_loss=0.08041, over 16153.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.3342, pruned_loss=0.09554, over 3070775.78 frames. ], batch size: 165, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:42:10,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4897, 4.7562, 4.8316, 4.8090, 4.7708, 5.3170, 4.8695, 4.6946], device='cuda:5'), covar=tensor([0.0961, 0.1396, 0.1256, 0.1692, 0.2420, 0.0887, 0.1142, 0.2172], device='cuda:5'), in_proj_covar=tensor([0.0258, 0.0359, 0.0347, 0.0315, 0.0411, 0.0384, 0.0292, 0.0421], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 04:42:34,025 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:42:57,101 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.420e+02 4.459e+02 5.761e+02 7.393e+02 1.688e+03, threshold=1.152e+03, percent-clipped=6.0 2023-04-28 04:42:59,075 INFO [train.py:904] (5/8) Epoch 4, batch 8200, loss[loss=0.2314, simple_loss=0.3117, pruned_loss=0.07555, over 16412.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.332, pruned_loss=0.0947, over 3082892.07 frames. ], batch size: 75, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:43:32,170 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 04:43:48,925 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5738, 4.8331, 4.5062, 4.5890, 4.2024, 4.1129, 4.3629, 4.8257], device='cuda:5'), covar=tensor([0.0623, 0.0690, 0.0841, 0.0446, 0.0598, 0.0916, 0.0614, 0.0720], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0440, 0.0382, 0.0283, 0.0280, 0.0296, 0.0358, 0.0310], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:43:53,411 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:44:15,013 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0399, 3.3423, 3.5286, 3.5361, 3.5171, 3.2835, 3.3496, 3.3926], device='cuda:5'), covar=tensor([0.0340, 0.0444, 0.0366, 0.0407, 0.0455, 0.0345, 0.0667, 0.0356], device='cuda:5'), in_proj_covar=tensor([0.0214, 0.0207, 0.0214, 0.0216, 0.0260, 0.0224, 0.0326, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 04:44:23,335 INFO [train.py:904] (5/8) Epoch 4, batch 8250, loss[loss=0.2243, simple_loss=0.3135, pruned_loss=0.0675, over 16782.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3311, pruned_loss=0.09278, over 3057635.57 frames. ], batch size: 102, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:44:49,023 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:44:49,400 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 04:45:02,059 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3124, 1.3734, 1.7364, 2.2233, 2.2305, 2.2934, 1.3476, 2.3865], device='cuda:5'), covar=tensor([0.0062, 0.0221, 0.0132, 0.0102, 0.0088, 0.0092, 0.0215, 0.0060], device='cuda:5'), in_proj_covar=tensor([0.0100, 0.0130, 0.0117, 0.0110, 0.0113, 0.0081, 0.0129, 0.0073], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 04:45:43,967 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 3.522e+02 4.130e+02 5.467e+02 1.123e+03, threshold=8.260e+02, percent-clipped=0.0 2023-04-28 04:45:45,838 INFO [train.py:904] (5/8) Epoch 4, batch 8300, loss[loss=0.2093, simple_loss=0.3038, pruned_loss=0.0574, over 16846.00 frames. ], tot_loss[loss=0.2524, simple_loss=0.3271, pruned_loss=0.08884, over 3052817.41 frames. ], batch size: 102, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:47:07,360 INFO [train.py:904] (5/8) Epoch 4, batch 8350, loss[loss=0.263, simple_loss=0.3243, pruned_loss=0.1009, over 12083.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3248, pruned_loss=0.08555, over 3043062.71 frames. ], batch size: 246, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:47:24,767 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5031, 3.7608, 2.9219, 2.3193, 2.7171, 2.1670, 3.8953, 3.7697], device='cuda:5'), covar=tensor([0.2116, 0.0501, 0.1055, 0.1389, 0.1778, 0.1301, 0.0292, 0.0465], device='cuda:5'), in_proj_covar=tensor([0.0264, 0.0233, 0.0251, 0.0225, 0.0281, 0.0188, 0.0219, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:47:44,476 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:48:29,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.235e+02 3.176e+02 3.884e+02 4.976e+02 1.172e+03, threshold=7.768e+02, percent-clipped=5.0 2023-04-28 04:48:30,629 INFO [train.py:904] (5/8) Epoch 4, batch 8400, loss[loss=0.2177, simple_loss=0.3039, pruned_loss=0.06575, over 16807.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3215, pruned_loss=0.0823, over 3047594.80 frames. ], batch size: 116, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:49:18,561 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0721, 4.4439, 3.5265, 2.4949, 3.3521, 2.3194, 4.6697, 4.5453], device='cuda:5'), covar=tensor([0.1962, 0.0467, 0.1002, 0.1327, 0.2054, 0.1352, 0.0281, 0.0343], device='cuda:5'), in_proj_covar=tensor([0.0263, 0.0231, 0.0248, 0.0224, 0.0275, 0.0188, 0.0219, 0.0220], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 04:49:26,933 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:49:53,151 INFO [train.py:904] (5/8) Epoch 4, batch 8450, loss[loss=0.2263, simple_loss=0.3087, pruned_loss=0.07188, over 16191.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3192, pruned_loss=0.08008, over 3052551.18 frames. ], batch size: 165, lr: 1.56e-02, grad_scale: 4.0 2023-04-28 04:50:15,312 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4921, 1.2679, 1.6435, 2.1842, 2.2534, 2.3477, 1.4942, 2.4169], device='cuda:5'), covar=tensor([0.0051, 0.0233, 0.0159, 0.0119, 0.0091, 0.0088, 0.0221, 0.0059], device='cuda:5'), in_proj_covar=tensor([0.0100, 0.0130, 0.0117, 0.0112, 0.0116, 0.0081, 0.0130, 0.0074], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 04:50:51,571 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-28 04:51:13,822 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.235e+02 3.132e+02 4.045e+02 5.414e+02 1.265e+03, threshold=8.090e+02, percent-clipped=8.0 2023-04-28 04:51:13,838 INFO [train.py:904] (5/8) Epoch 4, batch 8500, loss[loss=0.2086, simple_loss=0.287, pruned_loss=0.06509, over 15281.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.314, pruned_loss=0.07629, over 3058839.25 frames. ], batch size: 190, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:52:37,330 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 04:52:39,878 INFO [train.py:904] (5/8) Epoch 4, batch 8550, loss[loss=0.2263, simple_loss=0.3121, pruned_loss=0.07026, over 16503.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.311, pruned_loss=0.07477, over 3045577.17 frames. ], batch size: 75, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:53:10,006 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:53:26,596 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-04-28 04:54:21,294 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 3.391e+02 3.909e+02 5.367e+02 1.114e+03, threshold=7.819e+02, percent-clipped=4.0 2023-04-28 04:54:21,322 INFO [train.py:904] (5/8) Epoch 4, batch 8600, loss[loss=0.202, simple_loss=0.2811, pruned_loss=0.06142, over 12461.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3112, pruned_loss=0.07363, over 3049424.57 frames. ], batch size: 246, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:54:51,181 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:58,633 INFO [train.py:904] (5/8) Epoch 4, batch 8650, loss[loss=0.1978, simple_loss=0.2971, pruned_loss=0.04927, over 16840.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3086, pruned_loss=0.07158, over 3043866.60 frames. ], batch size: 102, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:57:44,933 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 3.067e+02 3.859e+02 4.894e+02 7.251e+02, threshold=7.718e+02, percent-clipped=0.0 2023-04-28 04:57:44,949 INFO [train.py:904] (5/8) Epoch 4, batch 8700, loss[loss=0.2003, simple_loss=0.2928, pruned_loss=0.05395, over 16871.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3049, pruned_loss=0.06977, over 3052137.70 frames. ], batch size: 102, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:58:36,857 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:50,440 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:59:20,259 INFO [train.py:904] (5/8) Epoch 4, batch 8750, loss[loss=0.2683, simple_loss=0.347, pruned_loss=0.09483, over 15325.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3054, pruned_loss=0.06938, over 3061855.14 frames. ], batch size: 191, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:00:45,700 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4771, 4.3983, 4.3826, 3.7746, 4.2553, 1.6379, 4.0694, 4.2949], device='cuda:5'), covar=tensor([0.0048, 0.0048, 0.0065, 0.0220, 0.0057, 0.1525, 0.0073, 0.0103], device='cuda:5'), in_proj_covar=tensor([0.0080, 0.0069, 0.0106, 0.0107, 0.0079, 0.0130, 0.0094, 0.0101], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 05:01:05,259 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:01:14,461 INFO [train.py:904] (5/8) Epoch 4, batch 8800, loss[loss=0.2165, simple_loss=0.307, pruned_loss=0.06299, over 16218.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3032, pruned_loss=0.06797, over 3049706.54 frames. ], batch size: 165, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:01:15,990 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.302e+02 3.583e+02 4.376e+02 5.208e+02 1.205e+03, threshold=8.753e+02, percent-clipped=6.0 2023-04-28 05:02:23,765 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:02:58,301 INFO [train.py:904] (5/8) Epoch 4, batch 8850, loss[loss=0.2328, simple_loss=0.3045, pruned_loss=0.08053, over 12765.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3055, pruned_loss=0.0671, over 3057218.61 frames. ], batch size: 249, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:03:34,691 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 05:04:08,642 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 05:04:32,505 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:04:44,188 INFO [train.py:904] (5/8) Epoch 4, batch 8900, loss[loss=0.2171, simple_loss=0.3018, pruned_loss=0.0662, over 16740.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3046, pruned_loss=0.06554, over 3055743.11 frames. ], batch size: 83, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:04:49,520 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.906e+02 3.441e+02 4.038e+02 4.851e+02 9.886e+02, threshold=8.076e+02, percent-clipped=2.0 2023-04-28 05:06:47,809 INFO [train.py:904] (5/8) Epoch 4, batch 8950, loss[loss=0.1954, simple_loss=0.2841, pruned_loss=0.05337, over 16240.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3048, pruned_loss=0.06627, over 3066771.56 frames. ], batch size: 165, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:06:51,567 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4371, 4.4185, 4.2612, 4.1071, 3.8187, 4.3346, 4.1935, 4.0487], device='cuda:5'), covar=tensor([0.0389, 0.0265, 0.0187, 0.0152, 0.0704, 0.0246, 0.0282, 0.0384], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0149, 0.0184, 0.0150, 0.0201, 0.0177, 0.0136, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 05:08:35,742 INFO [train.py:904] (5/8) Epoch 4, batch 9000, loss[loss=0.2024, simple_loss=0.2798, pruned_loss=0.06254, over 11995.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.3014, pruned_loss=0.06458, over 3069728.78 frames. ], batch size: 248, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:08:35,743 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 05:08:45,782 INFO [train.py:938] (5/8) Epoch 4, validation: loss=0.1802, simple_loss=0.283, pruned_loss=0.0387, over 944034.00 frames. 2023-04-28 05:08:45,783 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 05:08:49,544 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 05:08:49,860 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.055e+02 3.044e+02 3.757e+02 4.852e+02 9.817e+02, threshold=7.514e+02, percent-clipped=1.0 2023-04-28 05:09:11,746 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:09:18,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0364, 3.8329, 3.2719, 1.7765, 2.7604, 2.3259, 3.1568, 3.5486], device='cuda:5'), covar=tensor([0.0238, 0.0355, 0.0487, 0.1487, 0.0717, 0.0867, 0.0721, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0113, 0.0153, 0.0140, 0.0133, 0.0127, 0.0138, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 05:09:43,524 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:10:29,891 INFO [train.py:904] (5/8) Epoch 4, batch 9050, loss[loss=0.2224, simple_loss=0.3005, pruned_loss=0.0721, over 15381.00 frames. ], tot_loss[loss=0.217, simple_loss=0.303, pruned_loss=0.06556, over 3073106.88 frames. ], batch size: 191, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:11:17,782 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:21,930 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:25,513 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8263, 3.0262, 2.5553, 4.1741, 3.8791, 3.9430, 1.3467, 2.9385], device='cuda:5'), covar=tensor([0.1343, 0.0497, 0.1019, 0.0090, 0.0177, 0.0318, 0.1472, 0.0671], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0136, 0.0165, 0.0075, 0.0143, 0.0161, 0.0158, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 05:11:54,779 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:12:14,626 INFO [train.py:904] (5/8) Epoch 4, batch 9100, loss[loss=0.1893, simple_loss=0.2769, pruned_loss=0.05086, over 17197.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3023, pruned_loss=0.06562, over 3081970.81 frames. ], batch size: 46, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:12:18,750 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.564e+02 3.376e+02 4.042e+02 5.182e+02 1.418e+03, threshold=8.084e+02, percent-clipped=5.0 2023-04-28 05:12:45,214 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9029, 2.6946, 2.6892, 1.8230, 2.4967, 2.6243, 2.6498, 1.7740], device='cuda:5'), covar=tensor([0.0262, 0.0023, 0.0037, 0.0190, 0.0048, 0.0042, 0.0035, 0.0282], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0049, 0.0056, 0.0108, 0.0055, 0.0062, 0.0058, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 05:14:15,388 INFO [train.py:904] (5/8) Epoch 4, batch 9150, loss[loss=0.2034, simple_loss=0.2839, pruned_loss=0.06149, over 12037.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.3025, pruned_loss=0.06533, over 3063319.63 frames. ], batch size: 248, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:14:53,105 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5607, 4.5615, 3.8739, 2.1344, 3.0588, 2.7507, 3.6873, 4.1665], device='cuda:5'), covar=tensor([0.0269, 0.0336, 0.0445, 0.1454, 0.0680, 0.0786, 0.0701, 0.0703], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0114, 0.0152, 0.0140, 0.0133, 0.0126, 0.0139, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 05:15:01,392 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0882, 2.2169, 2.1189, 2.0603, 2.7591, 2.4131, 3.1173, 3.0323], device='cuda:5'), covar=tensor([0.0018, 0.0190, 0.0203, 0.0236, 0.0101, 0.0177, 0.0053, 0.0072], device='cuda:5'), in_proj_covar=tensor([0.0064, 0.0142, 0.0142, 0.0141, 0.0134, 0.0144, 0.0099, 0.0116], device='cuda:5'), out_proj_covar=tensor([8.0100e-05, 1.8250e-04, 1.7791e-04, 1.7611e-04, 1.7236e-04, 1.8517e-04, 1.2081e-04, 1.4859e-04], device='cuda:5') 2023-04-28 05:15:41,141 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:15:43,181 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:00,936 INFO [train.py:904] (5/8) Epoch 4, batch 9200, loss[loss=0.2095, simple_loss=0.296, pruned_loss=0.06145, over 16677.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2985, pruned_loss=0.0642, over 3061964.76 frames. ], batch size: 134, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:16:04,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.381e+02 3.414e+02 4.371e+02 6.281e+02 1.474e+03, threshold=8.741e+02, percent-clipped=12.0 2023-04-28 05:17:35,056 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 05:17:35,641 INFO [train.py:904] (5/8) Epoch 4, batch 9250, loss[loss=0.215, simple_loss=0.3057, pruned_loss=0.06215, over 16715.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2977, pruned_loss=0.06415, over 3043647.49 frames. ], batch size: 134, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:17:50,494 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2302, 1.2352, 1.8038, 2.0484, 2.1540, 2.1896, 1.4738, 2.1760], device='cuda:5'), covar=tensor([0.0069, 0.0213, 0.0108, 0.0107, 0.0098, 0.0076, 0.0185, 0.0049], device='cuda:5'), in_proj_covar=tensor([0.0100, 0.0129, 0.0115, 0.0110, 0.0114, 0.0079, 0.0129, 0.0071], device='cuda:5'), out_proj_covar=tensor([1.4334e-04, 1.8311e-04, 1.6899e-04, 1.5949e-04, 1.6292e-04, 1.1022e-04, 1.8347e-04, 9.9316e-05], device='cuda:5') 2023-04-28 05:18:37,594 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9298, 2.6683, 2.6935, 1.8924, 2.6221, 2.6849, 2.6431, 1.6440], device='cuda:5'), covar=tensor([0.0248, 0.0026, 0.0036, 0.0189, 0.0045, 0.0045, 0.0037, 0.0331], device='cuda:5'), in_proj_covar=tensor([0.0108, 0.0049, 0.0054, 0.0106, 0.0054, 0.0061, 0.0057, 0.0101], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 05:19:14,604 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7581, 3.9913, 3.9274, 1.7273, 4.2090, 4.3153, 3.2164, 3.3182], device='cuda:5'), covar=tensor([0.0547, 0.0081, 0.0197, 0.1175, 0.0039, 0.0037, 0.0254, 0.0277], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0084, 0.0078, 0.0140, 0.0068, 0.0073, 0.0111, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 05:19:26,138 INFO [train.py:904] (5/8) Epoch 4, batch 9300, loss[loss=0.1961, simple_loss=0.2838, pruned_loss=0.0542, over 15248.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2955, pruned_loss=0.06289, over 3042736.03 frames. ], batch size: 190, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:19:30,022 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.165e+02 3.034e+02 3.554e+02 4.321e+02 7.893e+02, threshold=7.107e+02, percent-clipped=0.0 2023-04-28 05:20:06,176 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:11,809 INFO [train.py:904] (5/8) Epoch 4, batch 9350, loss[loss=0.2117, simple_loss=0.2821, pruned_loss=0.07062, over 11952.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2952, pruned_loss=0.06297, over 3031570.48 frames. ], batch size: 250, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:21:41,288 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:48,591 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:52,644 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7562, 1.3687, 2.0524, 2.5907, 2.4248, 2.7545, 1.5957, 2.6358], device='cuda:5'), covar=tensor([0.0050, 0.0237, 0.0129, 0.0104, 0.0095, 0.0066, 0.0202, 0.0047], device='cuda:5'), in_proj_covar=tensor([0.0103, 0.0134, 0.0119, 0.0115, 0.0118, 0.0082, 0.0133, 0.0072], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 05:22:05,606 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:30,160 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:51,030 INFO [train.py:904] (5/8) Epoch 4, batch 9400, loss[loss=0.229, simple_loss=0.3114, pruned_loss=0.07329, over 15323.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2942, pruned_loss=0.06268, over 3010593.98 frames. ], batch size: 191, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:22:57,594 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.054e+02 3.057e+02 3.854e+02 4.791e+02 1.161e+03, threshold=7.709e+02, percent-clipped=3.0 2023-04-28 05:23:42,053 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:09,958 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:32,558 INFO [train.py:904] (5/8) Epoch 4, batch 9450, loss[loss=0.1985, simple_loss=0.2858, pruned_loss=0.05558, over 16398.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2957, pruned_loss=0.0631, over 2991315.00 frames. ], batch size: 146, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:24:47,357 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3036, 4.0605, 4.2857, 4.5435, 4.6067, 4.1382, 4.6139, 4.6088], device='cuda:5'), covar=tensor([0.0933, 0.0633, 0.1007, 0.0427, 0.0454, 0.0688, 0.0416, 0.0361], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0385, 0.0484, 0.0395, 0.0298, 0.0285, 0.0315, 0.0323], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 05:25:54,233 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:26:13,854 INFO [train.py:904] (5/8) Epoch 4, batch 9500, loss[loss=0.2133, simple_loss=0.3025, pruned_loss=0.06203, over 16146.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2957, pruned_loss=0.06243, over 3025691.02 frames. ], batch size: 165, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:26:21,224 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.115e+02 3.608e+02 4.489e+02 5.705e+02 9.491e+02, threshold=8.979e+02, percent-clipped=6.0 2023-04-28 05:27:31,420 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:27:48,799 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:28:04,408 INFO [train.py:904] (5/8) Epoch 4, batch 9550, loss[loss=0.2381, simple_loss=0.3244, pruned_loss=0.07585, over 16271.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2963, pruned_loss=0.06286, over 3037532.81 frames. ], batch size: 165, lr: 1.53e-02, grad_scale: 2.0 2023-04-28 05:29:46,542 INFO [train.py:904] (5/8) Epoch 4, batch 9600, loss[loss=0.2027, simple_loss=0.2903, pruned_loss=0.05752, over 16627.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2974, pruned_loss=0.06361, over 3041539.61 frames. ], batch size: 76, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:29:50,836 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 05:29:52,056 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.278e+02 3.502e+02 4.489e+02 5.494e+02 1.109e+03, threshold=8.977e+02, percent-clipped=3.0 2023-04-28 05:31:23,982 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1017, 4.0629, 3.7348, 1.8497, 3.0116, 2.4672, 3.3360, 3.7242], device='cuda:5'), covar=tensor([0.0316, 0.0437, 0.0418, 0.1629, 0.0681, 0.0906, 0.0796, 0.0696], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0114, 0.0155, 0.0141, 0.0134, 0.0127, 0.0139, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 05:31:33,083 INFO [train.py:904] (5/8) Epoch 4, batch 9650, loss[loss=0.2229, simple_loss=0.297, pruned_loss=0.07437, over 12525.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2993, pruned_loss=0.06428, over 3028550.07 frames. ], batch size: 248, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:32:17,598 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:17,723 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0489, 2.7490, 2.6162, 1.7922, 2.8962, 2.8923, 2.5942, 2.3901], device='cuda:5'), covar=tensor([0.0664, 0.0137, 0.0155, 0.0944, 0.0085, 0.0095, 0.0300, 0.0400], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0080, 0.0073, 0.0136, 0.0066, 0.0070, 0.0106, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 05:32:25,205 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:47,563 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1187, 3.1142, 2.9053, 1.9514, 2.6228, 2.0393, 2.7176, 2.7517], device='cuda:5'), covar=tensor([0.0358, 0.0423, 0.0409, 0.1467, 0.0600, 0.0941, 0.0667, 0.0733], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0113, 0.0154, 0.0142, 0.0134, 0.0127, 0.0139, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 05:33:12,508 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1861, 4.0308, 4.6198, 4.6211, 4.5898, 4.1697, 4.2508, 4.0401], device='cuda:5'), covar=tensor([0.0199, 0.0336, 0.0366, 0.0381, 0.0321, 0.0251, 0.0645, 0.0350], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0190, 0.0199, 0.0202, 0.0232, 0.0208, 0.0292, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-04-28 05:33:21,187 INFO [train.py:904] (5/8) Epoch 4, batch 9700, loss[loss=0.1983, simple_loss=0.2762, pruned_loss=0.06015, over 12344.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2982, pruned_loss=0.06413, over 3035894.85 frames. ], batch size: 248, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:33:26,547 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 3.154e+02 3.870e+02 5.322e+02 1.510e+03, threshold=7.740e+02, percent-clipped=2.0 2023-04-28 05:33:53,099 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:01,341 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:54,788 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:56,554 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-28 05:35:03,578 INFO [train.py:904] (5/8) Epoch 4, batch 9750, loss[loss=0.2245, simple_loss=0.3121, pruned_loss=0.06844, over 16226.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2968, pruned_loss=0.06376, over 3044971.07 frames. ], batch size: 165, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:35:38,815 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:35:54,443 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1201, 4.1495, 3.8823, 3.9031, 3.5969, 4.0653, 3.8112, 3.7797], device='cuda:5'), covar=tensor([0.0386, 0.0277, 0.0192, 0.0142, 0.0693, 0.0221, 0.0506, 0.0369], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0146, 0.0184, 0.0152, 0.0202, 0.0173, 0.0133, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 05:35:57,164 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3835, 4.0485, 4.1537, 2.7978, 3.6689, 4.0871, 3.8380, 2.8162], device='cuda:5'), covar=tensor([0.0296, 0.0011, 0.0021, 0.0196, 0.0032, 0.0027, 0.0023, 0.0202], device='cuda:5'), in_proj_covar=tensor([0.0110, 0.0050, 0.0056, 0.0108, 0.0054, 0.0062, 0.0057, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 05:36:45,114 INFO [train.py:904] (5/8) Epoch 4, batch 9800, loss[loss=0.2495, simple_loss=0.3447, pruned_loss=0.07714, over 16270.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2981, pruned_loss=0.06305, over 3059363.05 frames. ], batch size: 165, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:36:51,070 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.170e+02 3.844e+02 4.650e+02 7.827e+02, threshold=7.689e+02, percent-clipped=1.0 2023-04-28 05:36:56,604 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:23,425 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5197, 3.4636, 3.3649, 3.0529, 3.4251, 1.9126, 3.2625, 3.0049], device='cuda:5'), covar=tensor([0.0066, 0.0056, 0.0073, 0.0160, 0.0060, 0.1374, 0.0077, 0.0128], device='cuda:5'), in_proj_covar=tensor([0.0077, 0.0065, 0.0101, 0.0101, 0.0076, 0.0130, 0.0091, 0.0097], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 05:37:25,574 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:37,847 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:17,567 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40295.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:38:20,792 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2023-04-28 05:38:24,613 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9820, 2.7898, 2.6665, 1.6767, 2.8986, 2.9198, 2.5205, 2.3709], device='cuda:5'), covar=tensor([0.0721, 0.0128, 0.0164, 0.1089, 0.0081, 0.0097, 0.0378, 0.0447], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0082, 0.0077, 0.0140, 0.0068, 0.0072, 0.0110, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 05:38:29,682 INFO [train.py:904] (5/8) Epoch 4, batch 9850, loss[loss=0.2086, simple_loss=0.2852, pruned_loss=0.06601, over 12229.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2993, pruned_loss=0.06331, over 3050817.26 frames. ], batch size: 247, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:38:57,031 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4822, 1.9477, 1.4157, 1.6260, 2.3609, 2.0371, 2.4559, 2.4840], device='cuda:5'), covar=tensor([0.0017, 0.0164, 0.0234, 0.0228, 0.0090, 0.0156, 0.0057, 0.0079], device='cuda:5'), in_proj_covar=tensor([0.0063, 0.0141, 0.0139, 0.0141, 0.0133, 0.0140, 0.0097, 0.0114], device='cuda:5'), out_proj_covar=tensor([7.7143e-05, 1.7956e-04, 1.7305e-04, 1.7560e-04, 1.7047e-04, 1.7865e-04, 1.1827e-04, 1.4543e-04], device='cuda:5') 2023-04-28 05:39:18,064 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9716, 3.1956, 3.0365, 1.4777, 3.3312, 3.3857, 2.7250, 2.4693], device='cuda:5'), covar=tensor([0.0880, 0.0107, 0.0177, 0.1280, 0.0071, 0.0064, 0.0370, 0.0489], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0083, 0.0077, 0.0140, 0.0069, 0.0072, 0.0111, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 05:39:31,330 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:39,218 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:05,941 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:40:21,697 INFO [train.py:904] (5/8) Epoch 4, batch 9900, loss[loss=0.2033, simple_loss=0.3061, pruned_loss=0.05026, over 17238.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2996, pruned_loss=0.0628, over 3054016.17 frames. ], batch size: 52, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:40:27,914 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 3.210e+02 3.891e+02 5.069e+02 8.589e+02, threshold=7.781e+02, percent-clipped=1.0 2023-04-28 05:41:56,795 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:42:18,087 INFO [train.py:904] (5/8) Epoch 4, batch 9950, loss[loss=0.2037, simple_loss=0.293, pruned_loss=0.05721, over 16442.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.3017, pruned_loss=0.06309, over 3076226.44 frames. ], batch size: 68, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:43:13,883 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:20,294 INFO [train.py:904] (5/8) Epoch 4, batch 10000, loss[loss=0.1838, simple_loss=0.2786, pruned_loss=0.04448, over 16581.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2995, pruned_loss=0.06212, over 3102177.89 frames. ], batch size: 62, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:44:26,643 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.009e+02 3.138e+02 3.849e+02 4.764e+02 1.083e+03, threshold=7.697e+02, percent-clipped=5.0 2023-04-28 05:45:01,305 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:01,361 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:47,315 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:46:02,932 INFO [train.py:904] (5/8) Epoch 4, batch 10050, loss[loss=0.2144, simple_loss=0.2933, pruned_loss=0.06781, over 12077.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.299, pruned_loss=0.06184, over 3078116.84 frames. ], batch size: 247, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:46:38,832 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:38,805 INFO [train.py:904] (5/8) Epoch 4, batch 10100, loss[loss=0.2047, simple_loss=0.2839, pruned_loss=0.06278, over 15442.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2998, pruned_loss=0.06243, over 3100006.80 frames. ], batch size: 191, lr: 1.52e-02, grad_scale: 8.0 2023-04-28 05:47:39,388 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:42,227 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:42,815 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.191e+02 3.527e+02 4.401e+02 5.596e+02 1.013e+03, threshold=8.802e+02, percent-clipped=4.0 2023-04-28 05:48:27,676 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40574.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:25,190 INFO [train.py:904] (5/8) Epoch 5, batch 0, loss[loss=0.2492, simple_loss=0.3268, pruned_loss=0.08574, over 16692.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.3268, pruned_loss=0.08574, over 16692.00 frames. ], batch size: 57, lr: 1.42e-02, grad_scale: 8.0 2023-04-28 05:49:25,191 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 05:49:32,553 INFO [train.py:938] (5/8) Epoch 5, validation: loss=0.1789, simple_loss=0.2817, pruned_loss=0.03802, over 944034.00 frames. 2023-04-28 05:49:32,553 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 05:49:49,593 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6439, 3.7202, 2.8402, 2.4843, 2.8160, 2.4290, 3.6685, 3.6611], device='cuda:5'), covar=tensor([0.1602, 0.0422, 0.0960, 0.1208, 0.1433, 0.1149, 0.0360, 0.0522], device='cuda:5'), in_proj_covar=tensor([0.0259, 0.0238, 0.0253, 0.0227, 0.0234, 0.0190, 0.0220, 0.0219], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 05:50:11,887 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:34,964 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-04-28 05:50:42,809 INFO [train.py:904] (5/8) Epoch 5, batch 50, loss[loss=0.2161, simple_loss=0.3035, pruned_loss=0.06438, over 17237.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3215, pruned_loss=0.09471, over 747447.86 frames. ], batch size: 52, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:50:43,254 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:49,839 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.847e+02 4.852e+02 6.011e+02 1.299e+03, threshold=9.705e+02, percent-clipped=3.0 2023-04-28 05:51:29,209 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:51:50,921 INFO [train.py:904] (5/8) Epoch 5, batch 100, loss[loss=0.2753, simple_loss=0.3269, pruned_loss=0.1118, over 16828.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3133, pruned_loss=0.08756, over 1322895.02 frames. ], batch size: 109, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:52:02,099 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.91 vs. limit=5.0 2023-04-28 05:52:07,435 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:52:30,883 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0183, 3.7131, 3.9678, 4.2517, 4.2868, 3.8664, 4.1969, 4.2604], device='cuda:5'), covar=tensor([0.0805, 0.0806, 0.1219, 0.0451, 0.0423, 0.0871, 0.0627, 0.0359], device='cuda:5'), in_proj_covar=tensor([0.0358, 0.0439, 0.0556, 0.0436, 0.0328, 0.0318, 0.0353, 0.0363], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 05:52:59,309 INFO [train.py:904] (5/8) Epoch 5, batch 150, loss[loss=0.1969, simple_loss=0.2749, pruned_loss=0.05944, over 17194.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3079, pruned_loss=0.08365, over 1772521.38 frames. ], batch size: 44, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:53:08,168 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.263e+02 3.680e+02 4.214e+02 5.175e+02 1.144e+03, threshold=8.427e+02, percent-clipped=1.0 2023-04-28 05:53:12,808 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3388, 3.6034, 3.5667, 1.7389, 3.7182, 3.7860, 3.0760, 2.9218], device='cuda:5'), covar=tensor([0.0710, 0.0090, 0.0101, 0.1113, 0.0057, 0.0066, 0.0260, 0.0376], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0083, 0.0077, 0.0141, 0.0069, 0.0073, 0.0109, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 05:53:44,541 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4218, 4.4476, 4.5300, 4.5911, 4.4514, 5.0539, 4.7482, 4.4273], device='cuda:5'), covar=tensor([0.1334, 0.1639, 0.1369, 0.1800, 0.2907, 0.1145, 0.1178, 0.2355], device='cuda:5'), in_proj_covar=tensor([0.0257, 0.0377, 0.0362, 0.0327, 0.0426, 0.0385, 0.0297, 0.0432], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 05:54:09,230 INFO [train.py:904] (5/8) Epoch 5, batch 200, loss[loss=0.211, simple_loss=0.2884, pruned_loss=0.06681, over 15893.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3084, pruned_loss=0.08343, over 2109887.42 frames. ], batch size: 35, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:54:40,879 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6888, 2.7827, 2.1681, 2.3040, 3.1208, 2.8978, 3.7241, 3.3588], device='cuda:5'), covar=tensor([0.0018, 0.0143, 0.0202, 0.0207, 0.0092, 0.0149, 0.0069, 0.0078], device='cuda:5'), in_proj_covar=tensor([0.0070, 0.0145, 0.0147, 0.0145, 0.0138, 0.0146, 0.0108, 0.0122], device='cuda:5'), out_proj_covar=tensor([8.5322e-05, 1.8420e-04, 1.8202e-04, 1.7898e-04, 1.7649e-04, 1.8619e-04, 1.3156e-04, 1.5391e-04], device='cuda:5') 2023-04-28 05:55:02,265 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0283, 4.3508, 2.0001, 4.6104, 2.8313, 4.5434, 2.1741, 3.1473], device='cuda:5'), covar=tensor([0.0106, 0.0197, 0.1390, 0.0032, 0.0752, 0.0279, 0.1390, 0.0581], device='cuda:5'), in_proj_covar=tensor([0.0112, 0.0146, 0.0174, 0.0081, 0.0158, 0.0175, 0.0185, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 05:55:13,943 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:17,652 INFO [train.py:904] (5/8) Epoch 5, batch 250, loss[loss=0.2269, simple_loss=0.3101, pruned_loss=0.07182, over 17028.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3065, pruned_loss=0.08162, over 2372927.67 frames. ], batch size: 50, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:55:18,008 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:25,563 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.087e+02 3.707e+02 4.511e+02 5.174e+02 8.449e+02, threshold=9.022e+02, percent-clipped=1.0 2023-04-28 05:55:38,043 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:50,370 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:24,971 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:26,398 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:27,714 INFO [train.py:904] (5/8) Epoch 5, batch 300, loss[loss=0.1759, simple_loss=0.2593, pruned_loss=0.04626, over 16981.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3028, pruned_loss=0.07924, over 2587379.85 frames. ], batch size: 41, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:56:58,126 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:05,100 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:07,845 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:18,897 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4890, 3.1249, 3.9283, 2.5633, 3.6400, 3.8608, 3.6911, 2.0522], device='cuda:5'), covar=tensor([0.0301, 0.0136, 0.0030, 0.0246, 0.0037, 0.0050, 0.0037, 0.0314], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0058, 0.0060, 0.0115, 0.0058, 0.0066, 0.0062, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 05:57:38,141 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 05:57:39,686 INFO [train.py:904] (5/8) Epoch 5, batch 350, loss[loss=0.2635, simple_loss=0.3169, pruned_loss=0.105, over 12091.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3005, pruned_loss=0.07788, over 2735820.95 frames. ], batch size: 245, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:57:48,104 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.380e+02 3.200e+02 3.923e+02 5.275e+02 9.514e+02, threshold=7.845e+02, percent-clipped=1.0 2023-04-28 05:57:49,912 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7802, 3.8466, 2.7884, 2.3792, 2.9294, 2.2574, 3.8887, 4.0547], device='cuda:5'), covar=tensor([0.2001, 0.0616, 0.1283, 0.1379, 0.2164, 0.1347, 0.0404, 0.0590], device='cuda:5'), in_proj_covar=tensor([0.0276, 0.0251, 0.0267, 0.0241, 0.0284, 0.0202, 0.0233, 0.0243], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 05:57:54,160 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:58:14,897 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:28,112 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:44,159 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:48,868 INFO [train.py:904] (5/8) Epoch 5, batch 400, loss[loss=0.2514, simple_loss=0.3098, pruned_loss=0.09653, over 16876.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.2999, pruned_loss=0.07833, over 2861942.15 frames. ], batch size: 90, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 05:58:59,030 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:20,633 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 05:59:25,128 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7383, 4.2093, 4.3420, 1.7826, 4.6473, 4.7095, 3.2788, 3.7447], device='cuda:5'), covar=tensor([0.0712, 0.0107, 0.0228, 0.1199, 0.0052, 0.0051, 0.0305, 0.0315], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0086, 0.0081, 0.0144, 0.0071, 0.0077, 0.0113, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 05:59:36,591 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:38,765 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0959, 4.0652, 3.9126, 3.8524, 3.5548, 3.9690, 3.7518, 3.7761], device='cuda:5'), covar=tensor([0.0416, 0.0252, 0.0240, 0.0203, 0.0697, 0.0275, 0.0658, 0.0431], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0171, 0.0213, 0.0179, 0.0239, 0.0204, 0.0153, 0.0229], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 05:59:51,158 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41045.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:00:00,537 INFO [train.py:904] (5/8) Epoch 5, batch 450, loss[loss=0.1941, simple_loss=0.276, pruned_loss=0.05606, over 17237.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.2973, pruned_loss=0.07682, over 2963165.41 frames. ], batch size: 44, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:00:09,415 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 3.241e+02 3.884e+02 4.938e+02 1.053e+03, threshold=7.768e+02, percent-clipped=3.0 2023-04-28 06:00:11,793 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:00:43,942 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.12 vs. limit=2.0 2023-04-28 06:01:10,613 INFO [train.py:904] (5/8) Epoch 5, batch 500, loss[loss=0.2215, simple_loss=0.3127, pruned_loss=0.06517, over 17023.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2951, pruned_loss=0.07512, over 3045937.34 frames. ], batch size: 50, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:01:12,329 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:16,689 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:02:15,292 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:02:19,019 INFO [train.py:904] (5/8) Epoch 5, batch 550, loss[loss=0.2537, simple_loss=0.3111, pruned_loss=0.09817, over 16741.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.294, pruned_loss=0.07518, over 3088682.53 frames. ], batch size: 124, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:02:27,246 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.046e+02 3.287e+02 3.927e+02 4.639e+02 9.245e+02, threshold=7.855e+02, percent-clipped=2.0 2023-04-28 06:02:36,024 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:18,589 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4479, 4.3924, 4.4419, 3.6993, 4.3117, 1.7213, 4.1521, 4.2543], device='cuda:5'), covar=tensor([0.0093, 0.0073, 0.0092, 0.0318, 0.0077, 0.1765, 0.0098, 0.0127], device='cuda:5'), in_proj_covar=tensor([0.0091, 0.0075, 0.0122, 0.0123, 0.0089, 0.0140, 0.0105, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:03:22,053 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:27,700 INFO [train.py:904] (5/8) Epoch 5, batch 600, loss[loss=0.224, simple_loss=0.2875, pruned_loss=0.08026, over 16694.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2931, pruned_loss=0.07459, over 3147992.44 frames. ], batch size: 134, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:03:29,219 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9129, 3.9261, 4.4454, 4.4561, 4.4197, 3.9855, 4.0759, 3.9314], device='cuda:5'), covar=tensor([0.0305, 0.0486, 0.0366, 0.0391, 0.0418, 0.0322, 0.0821, 0.0489], device='cuda:5'), in_proj_covar=tensor([0.0233, 0.0226, 0.0234, 0.0236, 0.0271, 0.0244, 0.0350, 0.0209], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 06:03:54,902 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:04:33,949 INFO [train.py:904] (5/8) Epoch 5, batch 650, loss[loss=0.2228, simple_loss=0.2906, pruned_loss=0.07749, over 16457.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2913, pruned_loss=0.07306, over 3193129.11 frames. ], batch size: 75, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:04:41,765 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:04:42,450 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.330e+02 3.128e+02 3.726e+02 4.799e+02 9.270e+02, threshold=7.452e+02, percent-clipped=1.0 2023-04-28 06:04:45,777 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7392, 5.0804, 4.2997, 5.1456, 4.7180, 4.2888, 4.7453, 5.1301], device='cuda:5'), covar=tensor([0.1490, 0.1460, 0.2819, 0.0645, 0.1130, 0.1469, 0.1023, 0.1576], device='cuda:5'), in_proj_covar=tensor([0.0355, 0.0487, 0.0413, 0.0310, 0.0306, 0.0314, 0.0386, 0.0348], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:05:39,911 INFO [train.py:904] (5/8) Epoch 5, batch 700, loss[loss=0.2077, simple_loss=0.2921, pruned_loss=0.06166, over 16673.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2911, pruned_loss=0.07273, over 3226739.80 frames. ], batch size: 57, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:05:49,004 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:05,344 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3025, 3.1922, 3.2906, 3.4781, 3.5112, 3.1940, 3.3111, 3.5466], device='cuda:5'), covar=tensor([0.0699, 0.0638, 0.1019, 0.0463, 0.0465, 0.1856, 0.1025, 0.0486], device='cuda:5'), in_proj_covar=tensor([0.0386, 0.0475, 0.0605, 0.0476, 0.0357, 0.0347, 0.0382, 0.0397], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:06:10,401 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 06:06:49,386 INFO [train.py:904] (5/8) Epoch 5, batch 750, loss[loss=0.2022, simple_loss=0.2828, pruned_loss=0.06082, over 16401.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2911, pruned_loss=0.07301, over 3239941.35 frames. ], batch size: 75, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:06:52,788 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:55,089 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:57,956 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 3.021e+02 3.672e+02 4.220e+02 6.723e+02, threshold=7.344e+02, percent-clipped=0.0 2023-04-28 06:07:58,831 INFO [train.py:904] (5/8) Epoch 5, batch 800, loss[loss=0.2435, simple_loss=0.3038, pruned_loss=0.09162, over 15539.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.291, pruned_loss=0.07265, over 3263472.13 frames. ], batch size: 190, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:07:59,212 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:08:36,296 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9053, 4.5702, 4.8059, 5.0994, 5.1981, 4.6000, 5.1890, 5.1660], device='cuda:5'), covar=tensor([0.0715, 0.0755, 0.1319, 0.0442, 0.0402, 0.0554, 0.0376, 0.0390], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0474, 0.0610, 0.0478, 0.0365, 0.0348, 0.0378, 0.0405], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:08:48,229 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7854, 4.2852, 4.6624, 3.0976, 4.1875, 4.6329, 4.2763, 2.4754], device='cuda:5'), covar=tensor([0.0263, 0.0018, 0.0016, 0.0204, 0.0028, 0.0027, 0.0026, 0.0247], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0060, 0.0060, 0.0113, 0.0059, 0.0067, 0.0061, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 06:09:08,626 INFO [train.py:904] (5/8) Epoch 5, batch 850, loss[loss=0.2024, simple_loss=0.2843, pruned_loss=0.06026, over 17140.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2899, pruned_loss=0.07171, over 3283261.04 frames. ], batch size: 47, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:09:16,405 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 3.257e+02 3.799e+02 4.682e+02 9.659e+02, threshold=7.597e+02, percent-clipped=4.0 2023-04-28 06:09:17,873 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:10:16,023 INFO [train.py:904] (5/8) Epoch 5, batch 900, loss[loss=0.2657, simple_loss=0.3221, pruned_loss=0.1047, over 12280.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2882, pruned_loss=0.07034, over 3281262.96 frames. ], batch size: 246, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:10:44,985 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:11:27,570 INFO [train.py:904] (5/8) Epoch 5, batch 950, loss[loss=0.1952, simple_loss=0.28, pruned_loss=0.05518, over 17241.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2876, pruned_loss=0.06981, over 3290534.63 frames. ], batch size: 45, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:11:34,567 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:11:35,287 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.063e+02 2.970e+02 3.801e+02 5.319e+02 1.620e+03, threshold=7.602e+02, percent-clipped=9.0 2023-04-28 06:11:53,682 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:37,223 INFO [train.py:904] (5/8) Epoch 5, batch 1000, loss[loss=0.2258, simple_loss=0.2871, pruned_loss=0.08223, over 16465.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2871, pruned_loss=0.06978, over 3292808.90 frames. ], batch size: 68, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:12:41,515 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:37,840 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3142, 3.9607, 4.1055, 1.8061, 4.2778, 4.2660, 3.3821, 3.2738], device='cuda:5'), covar=tensor([0.0900, 0.0116, 0.0192, 0.1221, 0.0052, 0.0079, 0.0281, 0.0421], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0086, 0.0084, 0.0142, 0.0071, 0.0080, 0.0114, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 06:13:45,707 INFO [train.py:904] (5/8) Epoch 5, batch 1050, loss[loss=0.1867, simple_loss=0.2673, pruned_loss=0.05305, over 17206.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2868, pruned_loss=0.06908, over 3296679.21 frames. ], batch size: 44, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:13:49,035 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:50,403 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:54,770 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.880e+02 3.783e+02 4.446e+02 1.011e+03, threshold=7.566e+02, percent-clipped=5.0 2023-04-28 06:14:56,096 INFO [train.py:904] (5/8) Epoch 5, batch 1100, loss[loss=0.1792, simple_loss=0.2618, pruned_loss=0.04826, over 16847.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2856, pruned_loss=0.06835, over 3303083.77 frames. ], batch size: 42, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:14:56,430 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:14:56,607 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:14:58,830 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4753, 4.8829, 4.6176, 4.6205, 4.3051, 4.3090, 4.3904, 4.8767], device='cuda:5'), covar=tensor([0.0702, 0.0699, 0.0808, 0.0456, 0.0613, 0.0908, 0.0653, 0.0776], device='cuda:5'), in_proj_covar=tensor([0.0365, 0.0500, 0.0421, 0.0320, 0.0313, 0.0317, 0.0395, 0.0360], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:15:04,203 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9460, 4.6693, 4.8702, 5.2164, 5.3716, 4.7107, 5.2790, 5.2127], device='cuda:5'), covar=tensor([0.0872, 0.0797, 0.1368, 0.0467, 0.0345, 0.0502, 0.0449, 0.0450], device='cuda:5'), in_proj_covar=tensor([0.0394, 0.0484, 0.0616, 0.0492, 0.0369, 0.0357, 0.0389, 0.0409], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:15:15,822 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:15:53,248 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2731, 5.2027, 5.0459, 3.9910, 4.9583, 1.8940, 4.7719, 5.0073], device='cuda:5'), covar=tensor([0.0061, 0.0055, 0.0080, 0.0435, 0.0071, 0.1836, 0.0099, 0.0150], device='cuda:5'), in_proj_covar=tensor([0.0097, 0.0081, 0.0128, 0.0134, 0.0096, 0.0144, 0.0111, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 06:16:02,694 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:05,307 INFO [train.py:904] (5/8) Epoch 5, batch 1150, loss[loss=0.2131, simple_loss=0.2744, pruned_loss=0.07588, over 16462.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.285, pruned_loss=0.06764, over 3302112.51 frames. ], batch size: 75, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:16:12,835 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.850e+02 3.603e+02 4.605e+02 7.939e+02, threshold=7.207e+02, percent-clipped=2.0 2023-04-28 06:16:15,815 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:17:14,383 INFO [train.py:904] (5/8) Epoch 5, batch 1200, loss[loss=0.2394, simple_loss=0.2987, pruned_loss=0.0901, over 16896.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2843, pruned_loss=0.06724, over 3298711.23 frames. ], batch size: 109, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:17:21,141 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:17:21,397 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7523, 3.3446, 2.8192, 5.1433, 4.6636, 4.3416, 1.5059, 3.5772], device='cuda:5'), covar=tensor([0.1355, 0.0544, 0.1070, 0.0062, 0.0258, 0.0296, 0.1398, 0.0535], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0140, 0.0164, 0.0082, 0.0172, 0.0169, 0.0156, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 06:18:13,156 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8815, 4.2602, 4.4847, 3.3263, 4.1159, 4.5184, 4.2334, 2.5724], device='cuda:5'), covar=tensor([0.0272, 0.0028, 0.0023, 0.0195, 0.0035, 0.0032, 0.0025, 0.0280], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0062, 0.0062, 0.0118, 0.0062, 0.0070, 0.0065, 0.0108], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 06:18:23,587 INFO [train.py:904] (5/8) Epoch 5, batch 1250, loss[loss=0.2106, simple_loss=0.2975, pruned_loss=0.06183, over 16750.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2849, pruned_loss=0.06772, over 3306682.14 frames. ], batch size: 57, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:18:31,525 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 3.380e+02 4.131e+02 4.913e+02 1.055e+03, threshold=8.263e+02, percent-clipped=6.0 2023-04-28 06:19:30,799 INFO [train.py:904] (5/8) Epoch 5, batch 1300, loss[loss=0.2116, simple_loss=0.2777, pruned_loss=0.07272, over 16692.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2854, pruned_loss=0.06822, over 3299209.25 frames. ], batch size: 134, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:19:38,924 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:19:45,272 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8853, 5.3526, 5.4570, 5.4013, 5.3385, 5.9522, 5.5739, 5.2768], device='cuda:5'), covar=tensor([0.0671, 0.1391, 0.1252, 0.1582, 0.2358, 0.0815, 0.0889, 0.1827], device='cuda:5'), in_proj_covar=tensor([0.0267, 0.0395, 0.0377, 0.0334, 0.0450, 0.0406, 0.0312, 0.0453], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 06:20:38,081 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:20:42,275 INFO [train.py:904] (5/8) Epoch 5, batch 1350, loss[loss=0.2381, simple_loss=0.2985, pruned_loss=0.08887, over 16725.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2867, pruned_loss=0.06838, over 3306832.56 frames. ], batch size: 124, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:20:51,207 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 3.407e+02 4.000e+02 4.882e+02 1.065e+03, threshold=8.000e+02, percent-clipped=1.0 2023-04-28 06:21:05,582 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:31,976 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 06:21:55,643 INFO [train.py:904] (5/8) Epoch 5, batch 1400, loss[loss=0.2311, simple_loss=0.2884, pruned_loss=0.08694, over 16334.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2861, pruned_loss=0.06806, over 3313317.24 frames. ], batch size: 165, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:22:08,286 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:22:09,944 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:22:15,441 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7685, 2.8002, 2.4461, 4.0135, 3.5834, 3.7566, 1.5304, 2.9055], device='cuda:5'), covar=tensor([0.1255, 0.0488, 0.0991, 0.0079, 0.0221, 0.0331, 0.1265, 0.0618], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0139, 0.0164, 0.0082, 0.0170, 0.0168, 0.0157, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 06:22:19,632 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-28 06:23:01,016 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7395, 4.3811, 3.9716, 2.1698, 3.1601, 2.5999, 3.9064, 3.9896], device='cuda:5'), covar=tensor([0.0219, 0.0475, 0.0406, 0.1369, 0.0641, 0.0823, 0.0549, 0.0745], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0131, 0.0151, 0.0139, 0.0131, 0.0124, 0.0138, 0.0136], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 06:23:05,349 INFO [train.py:904] (5/8) Epoch 5, batch 1450, loss[loss=0.2271, simple_loss=0.2796, pruned_loss=0.08729, over 16866.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2844, pruned_loss=0.06736, over 3320866.66 frames. ], batch size: 96, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:23:15,560 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.035e+02 3.059e+02 3.831e+02 4.678e+02 9.639e+02, threshold=7.661e+02, percent-clipped=1.0 2023-04-28 06:24:02,334 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6310, 3.6871, 2.8206, 2.3016, 2.7799, 2.0876, 3.5155, 3.7757], device='cuda:5'), covar=tensor([0.2010, 0.0514, 0.1180, 0.1525, 0.2028, 0.1540, 0.0403, 0.0618], device='cuda:5'), in_proj_covar=tensor([0.0274, 0.0252, 0.0268, 0.0242, 0.0298, 0.0202, 0.0238, 0.0260], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:24:14,180 INFO [train.py:904] (5/8) Epoch 5, batch 1500, loss[loss=0.2282, simple_loss=0.2865, pruned_loss=0.08494, over 16839.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2847, pruned_loss=0.068, over 3321592.20 frames. ], batch size: 102, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:24:47,022 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9565, 4.3150, 2.0323, 4.5713, 2.7856, 4.6021, 2.1530, 3.3044], device='cuda:5'), covar=tensor([0.0104, 0.0202, 0.1443, 0.0031, 0.0700, 0.0207, 0.1221, 0.0455], device='cuda:5'), in_proj_covar=tensor([0.0115, 0.0152, 0.0172, 0.0085, 0.0156, 0.0182, 0.0182, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 06:25:21,165 INFO [train.py:904] (5/8) Epoch 5, batch 1550, loss[loss=0.2431, simple_loss=0.2989, pruned_loss=0.09367, over 16819.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2868, pruned_loss=0.07031, over 3318185.99 frames. ], batch size: 102, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:25:32,956 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.189e+02 3.587e+02 4.011e+02 4.553e+02 8.694e+02, threshold=8.021e+02, percent-clipped=2.0 2023-04-28 06:25:40,348 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2580, 5.2303, 5.0405, 4.2717, 4.9239, 1.8915, 4.8082, 5.0090], device='cuda:5'), covar=tensor([0.0057, 0.0043, 0.0078, 0.0327, 0.0067, 0.1571, 0.0073, 0.0114], device='cuda:5'), in_proj_covar=tensor([0.0097, 0.0080, 0.0127, 0.0134, 0.0097, 0.0142, 0.0111, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 06:26:32,628 INFO [train.py:904] (5/8) Epoch 5, batch 1600, loss[loss=0.22, simple_loss=0.2836, pruned_loss=0.07825, over 16688.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2883, pruned_loss=0.07053, over 3313524.70 frames. ], batch size: 124, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:26:35,243 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2072, 3.4620, 3.5398, 1.7684, 3.7228, 3.6275, 3.0015, 2.8289], device='cuda:5'), covar=tensor([0.0730, 0.0104, 0.0126, 0.1057, 0.0053, 0.0082, 0.0292, 0.0383], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0086, 0.0082, 0.0141, 0.0070, 0.0078, 0.0112, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 06:26:41,246 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 06:26:59,406 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 06:27:35,558 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 06:27:39,765 INFO [train.py:904] (5/8) Epoch 5, batch 1650, loss[loss=0.2197, simple_loss=0.2966, pruned_loss=0.07136, over 16439.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2899, pruned_loss=0.07119, over 3312909.33 frames. ], batch size: 68, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:27:49,823 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.983e+02 3.262e+02 3.932e+02 4.837e+02 9.371e+02, threshold=7.863e+02, percent-clipped=1.0 2023-04-28 06:27:56,381 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:31,675 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7675, 1.7806, 1.4704, 1.5137, 1.8896, 1.7352, 1.8182, 1.9495], device='cuda:5'), covar=tensor([0.0040, 0.0103, 0.0157, 0.0152, 0.0074, 0.0115, 0.0071, 0.0076], device='cuda:5'), in_proj_covar=tensor([0.0084, 0.0152, 0.0152, 0.0149, 0.0147, 0.0154, 0.0127, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:28:49,289 INFO [train.py:904] (5/8) Epoch 5, batch 1700, loss[loss=0.2591, simple_loss=0.3167, pruned_loss=0.1008, over 16882.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2923, pruned_loss=0.07255, over 3316095.88 frames. ], batch size: 109, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:28:54,204 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:29:01,283 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42310.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:29:29,631 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2003, 2.5185, 2.4309, 4.7261, 1.9930, 3.8020, 2.5310, 2.6378], device='cuda:5'), covar=tensor([0.0488, 0.1861, 0.0964, 0.0254, 0.2972, 0.0688, 0.1664, 0.2466], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0311, 0.0252, 0.0311, 0.0361, 0.0299, 0.0278, 0.0383], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:29:32,956 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1484, 1.5502, 2.3096, 2.8878, 2.8811, 2.8638, 1.6178, 2.9901], device='cuda:5'), covar=tensor([0.0062, 0.0217, 0.0155, 0.0109, 0.0079, 0.0108, 0.0215, 0.0055], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0139, 0.0126, 0.0126, 0.0123, 0.0089, 0.0138, 0.0079], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 06:29:53,141 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 06:29:56,509 INFO [train.py:904] (5/8) Epoch 5, batch 1750, loss[loss=0.2052, simple_loss=0.2895, pruned_loss=0.06044, over 17129.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2936, pruned_loss=0.07254, over 3320735.67 frames. ], batch size: 47, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:30:05,726 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.019e+02 3.418e+02 4.322e+02 5.636e+02 1.741e+03, threshold=8.644e+02, percent-clipped=7.0 2023-04-28 06:30:05,999 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:30:35,478 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:30:50,562 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0672, 2.3568, 2.3090, 4.7510, 1.9568, 3.6929, 2.3645, 2.4294], device='cuda:5'), covar=tensor([0.0472, 0.1950, 0.1023, 0.0210, 0.2886, 0.0676, 0.1760, 0.2640], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0312, 0.0252, 0.0311, 0.0360, 0.0298, 0.0278, 0.0383], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:31:06,110 INFO [train.py:904] (5/8) Epoch 5, batch 1800, loss[loss=0.2051, simple_loss=0.2855, pruned_loss=0.06238, over 17234.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2939, pruned_loss=0.07174, over 3320786.46 frames. ], batch size: 45, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:32:01,420 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:16,841 INFO [train.py:904] (5/8) Epoch 5, batch 1850, loss[loss=0.2034, simple_loss=0.2798, pruned_loss=0.06353, over 16833.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2934, pruned_loss=0.07108, over 3330785.25 frames. ], batch size: 42, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:32:26,221 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 3.144e+02 3.803e+02 4.355e+02 7.438e+02, threshold=7.606e+02, percent-clipped=0.0 2023-04-28 06:33:25,575 INFO [train.py:904] (5/8) Epoch 5, batch 1900, loss[loss=0.2051, simple_loss=0.2929, pruned_loss=0.05864, over 17299.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2919, pruned_loss=0.06951, over 3332631.00 frames. ], batch size: 52, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:33:32,593 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:35,501 INFO [train.py:904] (5/8) Epoch 5, batch 1950, loss[loss=0.1932, simple_loss=0.2915, pruned_loss=0.04741, over 17054.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2927, pruned_loss=0.06977, over 3320925.20 frames. ], batch size: 50, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:34:45,741 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 06:34:47,176 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 3.199e+02 3.788e+02 4.472e+02 9.515e+02, threshold=7.576e+02, percent-clipped=2.0 2023-04-28 06:34:52,112 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:59,500 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:35:10,015 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1553, 4.9743, 5.0618, 4.3013, 4.9693, 1.9244, 4.8286, 5.0770], device='cuda:5'), covar=tensor([0.0065, 0.0059, 0.0087, 0.0355, 0.0062, 0.1566, 0.0079, 0.0113], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0082, 0.0129, 0.0136, 0.0098, 0.0141, 0.0112, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 06:35:48,045 INFO [train.py:904] (5/8) Epoch 5, batch 2000, loss[loss=0.2113, simple_loss=0.2984, pruned_loss=0.06216, over 17134.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2919, pruned_loss=0.06888, over 3321531.81 frames. ], batch size: 47, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:35:51,836 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:36:00,889 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:50,010 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3103, 3.8505, 3.3397, 1.9525, 2.8002, 2.2747, 3.6375, 3.6556], device='cuda:5'), covar=tensor([0.0178, 0.0436, 0.0506, 0.1492, 0.0697, 0.0951, 0.0481, 0.0698], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0133, 0.0152, 0.0141, 0.0133, 0.0125, 0.0140, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 06:36:55,050 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:57,538 INFO [train.py:904] (5/8) Epoch 5, batch 2050, loss[loss=0.1876, simple_loss=0.2658, pruned_loss=0.05469, over 17018.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2926, pruned_loss=0.06934, over 3314086.21 frames. ], batch size: 41, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:36:59,092 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:37:06,896 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 3.076e+02 3.576e+02 4.219e+02 7.856e+02, threshold=7.152e+02, percent-clipped=1.0 2023-04-28 06:38:05,327 INFO [train.py:904] (5/8) Epoch 5, batch 2100, loss[loss=0.2486, simple_loss=0.3114, pruned_loss=0.09287, over 16833.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2938, pruned_loss=0.0706, over 3312806.55 frames. ], batch size: 102, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:38:18,861 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:38:47,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4375, 3.3699, 2.4547, 2.1482, 2.5839, 2.0395, 3.2721, 3.4312], device='cuda:5'), covar=tensor([0.2053, 0.0675, 0.1339, 0.1598, 0.2088, 0.1578, 0.0532, 0.0751], device='cuda:5'), in_proj_covar=tensor([0.0275, 0.0251, 0.0268, 0.0242, 0.0302, 0.0200, 0.0238, 0.0262], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:38:52,944 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:38:55,448 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1236, 5.0839, 4.9024, 4.3308, 4.8531, 2.2809, 4.7006, 4.9865], device='cuda:5'), covar=tensor([0.0060, 0.0055, 0.0083, 0.0320, 0.0065, 0.1320, 0.0078, 0.0115], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0082, 0.0127, 0.0134, 0.0096, 0.0139, 0.0112, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 06:39:15,227 INFO [train.py:904] (5/8) Epoch 5, batch 2150, loss[loss=0.2157, simple_loss=0.3016, pruned_loss=0.06492, over 17261.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2943, pruned_loss=0.07086, over 3305681.88 frames. ], batch size: 52, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:39:24,100 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.183e+02 3.251e+02 3.882e+02 4.566e+02 8.930e+02, threshold=7.764e+02, percent-clipped=4.0 2023-04-28 06:39:31,709 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 06:40:23,580 INFO [train.py:904] (5/8) Epoch 5, batch 2200, loss[loss=0.2279, simple_loss=0.3139, pruned_loss=0.07099, over 17078.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2958, pruned_loss=0.07198, over 3301836.27 frames. ], batch size: 53, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:40:59,104 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5493, 4.4204, 4.3740, 4.3009, 4.0324, 4.4192, 4.2772, 4.1467], device='cuda:5'), covar=tensor([0.0370, 0.0308, 0.0186, 0.0197, 0.0722, 0.0270, 0.0399, 0.0501], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0193, 0.0231, 0.0202, 0.0266, 0.0221, 0.0170, 0.0254], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 06:41:06,897 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6825, 2.2865, 1.9143, 1.9679, 2.7754, 2.6205, 3.0607, 2.8155], device='cuda:5'), covar=tensor([0.0060, 0.0179, 0.0216, 0.0225, 0.0104, 0.0152, 0.0096, 0.0116], device='cuda:5'), in_proj_covar=tensor([0.0085, 0.0152, 0.0149, 0.0147, 0.0148, 0.0153, 0.0127, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:41:34,697 INFO [train.py:904] (5/8) Epoch 5, batch 2250, loss[loss=0.3523, simple_loss=0.3949, pruned_loss=0.1548, over 11924.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.2986, pruned_loss=0.07435, over 3290913.52 frames. ], batch size: 247, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:41:43,708 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 3.207e+02 3.862e+02 4.939e+02 1.226e+03, threshold=7.724e+02, percent-clipped=4.0 2023-04-28 06:41:46,050 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:49,602 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:14,470 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:44,596 INFO [train.py:904] (5/8) Epoch 5, batch 2300, loss[loss=0.1963, simple_loss=0.2797, pruned_loss=0.05644, over 16866.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.2997, pruned_loss=0.07457, over 3298254.66 frames. ], batch size: 96, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:42:58,848 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8793, 2.3845, 2.3267, 4.4511, 1.9497, 3.5144, 2.3773, 2.4613], device='cuda:5'), covar=tensor([0.0461, 0.1806, 0.0966, 0.0261, 0.2689, 0.0781, 0.1783, 0.2203], device='cuda:5'), in_proj_covar=tensor([0.0312, 0.0313, 0.0252, 0.0309, 0.0359, 0.0300, 0.0279, 0.0385], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:43:11,619 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:20,359 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4150, 4.4218, 3.7650, 2.0136, 3.0027, 2.3683, 3.7570, 4.1279], device='cuda:5'), covar=tensor([0.0288, 0.0397, 0.0477, 0.1461, 0.0661, 0.0963, 0.0628, 0.0744], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0133, 0.0152, 0.0140, 0.0132, 0.0126, 0.0139, 0.0137], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 06:43:38,222 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:52,771 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 06:43:53,252 INFO [train.py:904] (5/8) Epoch 5, batch 2350, loss[loss=0.2352, simple_loss=0.3002, pruned_loss=0.0851, over 16844.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2993, pruned_loss=0.07485, over 3311991.44 frames. ], batch size: 90, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:44:03,299 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.280e+02 3.235e+02 3.889e+02 4.889e+02 8.777e+02, threshold=7.778e+02, percent-clipped=2.0 2023-04-28 06:44:51,988 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8437, 4.8432, 5.3602, 5.3770, 5.3321, 4.9471, 4.9492, 4.5827], device='cuda:5'), covar=tensor([0.0200, 0.0248, 0.0350, 0.0330, 0.0326, 0.0223, 0.0647, 0.0341], device='cuda:5'), in_proj_covar=tensor([0.0239, 0.0232, 0.0236, 0.0237, 0.0285, 0.0249, 0.0357, 0.0213], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 06:44:55,258 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 06:45:02,228 INFO [train.py:904] (5/8) Epoch 5, batch 2400, loss[loss=0.2269, simple_loss=0.3146, pruned_loss=0.06965, over 17288.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2995, pruned_loss=0.07405, over 3314430.21 frames. ], batch size: 52, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:45:08,768 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:45:52,165 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:46:13,640 INFO [train.py:904] (5/8) Epoch 5, batch 2450, loss[loss=0.2171, simple_loss=0.2988, pruned_loss=0.06765, over 16249.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.2993, pruned_loss=0.074, over 3307641.74 frames. ], batch size: 165, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:46:16,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8260, 3.1220, 2.5477, 4.4185, 4.0096, 3.9914, 1.5362, 2.9930], device='cuda:5'), covar=tensor([0.1176, 0.0417, 0.0894, 0.0054, 0.0205, 0.0311, 0.1207, 0.0592], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0140, 0.0163, 0.0084, 0.0177, 0.0170, 0.0157, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 06:46:26,012 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 3.112e+02 3.702e+02 4.495e+02 8.852e+02, threshold=7.404e+02, percent-clipped=2.0 2023-04-28 06:46:50,886 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3678, 1.4372, 1.9716, 2.2549, 2.3992, 2.4248, 1.5689, 2.5171], device='cuda:5'), covar=tensor([0.0061, 0.0225, 0.0138, 0.0118, 0.0097, 0.0105, 0.0179, 0.0040], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0143, 0.0127, 0.0131, 0.0127, 0.0092, 0.0138, 0.0080], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 06:46:58,056 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:47:23,969 INFO [train.py:904] (5/8) Epoch 5, batch 2500, loss[loss=0.1923, simple_loss=0.2753, pruned_loss=0.05464, over 16467.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2987, pruned_loss=0.07328, over 3312647.71 frames. ], batch size: 68, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:48:33,495 INFO [train.py:904] (5/8) Epoch 5, batch 2550, loss[loss=0.2024, simple_loss=0.2882, pruned_loss=0.05829, over 17039.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2992, pruned_loss=0.07375, over 3309424.08 frames. ], batch size: 50, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:48:45,561 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.843e+02 3.229e+02 3.908e+02 4.674e+02 8.443e+02, threshold=7.816e+02, percent-clipped=2.0 2023-04-28 06:48:48,799 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:49:43,648 INFO [train.py:904] (5/8) Epoch 5, batch 2600, loss[loss=0.2301, simple_loss=0.2999, pruned_loss=0.08012, over 16750.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.2992, pruned_loss=0.07329, over 3306973.37 frames. ], batch size: 83, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:49:55,550 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:01,844 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:22,623 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 06:50:31,268 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43235.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:53,893 INFO [train.py:904] (5/8) Epoch 5, batch 2650, loss[loss=0.2272, simple_loss=0.2989, pruned_loss=0.07775, over 16426.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.2989, pruned_loss=0.07247, over 3318950.63 frames. ], batch size: 146, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:51:05,580 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 3.400e+02 3.860e+02 4.730e+02 9.799e+02, threshold=7.720e+02, percent-clipped=2.0 2023-04-28 06:51:05,998 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6825, 4.7043, 5.2052, 5.2001, 5.2158, 4.6374, 4.7377, 4.4336], device='cuda:5'), covar=tensor([0.0238, 0.0297, 0.0304, 0.0318, 0.0346, 0.0326, 0.0774, 0.0398], device='cuda:5'), in_proj_covar=tensor([0.0249, 0.0239, 0.0241, 0.0246, 0.0295, 0.0258, 0.0368, 0.0221], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 06:51:23,733 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0674, 4.3687, 4.5059, 1.8974, 4.7707, 4.8389, 3.4138, 3.9377], device='cuda:5'), covar=tensor([0.0563, 0.0095, 0.0099, 0.1093, 0.0030, 0.0037, 0.0268, 0.0250], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0086, 0.0087, 0.0143, 0.0073, 0.0080, 0.0116, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 06:52:02,503 INFO [train.py:904] (5/8) Epoch 5, batch 2700, loss[loss=0.1954, simple_loss=0.2702, pruned_loss=0.06032, over 16846.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2983, pruned_loss=0.07171, over 3324344.23 frames. ], batch size: 39, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:52:09,077 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:53:12,584 INFO [train.py:904] (5/8) Epoch 5, batch 2750, loss[loss=0.1997, simple_loss=0.2933, pruned_loss=0.05303, over 17082.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2979, pruned_loss=0.0713, over 3329096.22 frames. ], batch size: 53, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:53:15,878 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:53:25,429 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.783e+02 3.325e+02 4.453e+02 8.515e+02, threshold=6.649e+02, percent-clipped=2.0 2023-04-28 06:53:31,261 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 06:54:17,714 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 06:54:22,964 INFO [train.py:904] (5/8) Epoch 5, batch 2800, loss[loss=0.2288, simple_loss=0.2914, pruned_loss=0.08312, over 15623.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2971, pruned_loss=0.07057, over 3333079.13 frames. ], batch size: 191, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:54:45,289 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6030, 3.3651, 2.8688, 1.9920, 2.5083, 2.1319, 3.1110, 3.1900], device='cuda:5'), covar=tensor([0.0262, 0.0546, 0.0563, 0.1394, 0.0785, 0.0917, 0.0582, 0.0686], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0133, 0.0154, 0.0140, 0.0132, 0.0124, 0.0140, 0.0139], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 06:54:52,906 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7530, 2.2950, 2.4578, 4.4540, 1.9763, 3.3116, 2.3562, 2.4522], device='cuda:5'), covar=tensor([0.0511, 0.1757, 0.0839, 0.0255, 0.2634, 0.0859, 0.1647, 0.2018], device='cuda:5'), in_proj_covar=tensor([0.0315, 0.0316, 0.0255, 0.0309, 0.0362, 0.0307, 0.0281, 0.0386], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 06:55:33,340 INFO [train.py:904] (5/8) Epoch 5, batch 2850, loss[loss=0.204, simple_loss=0.2896, pruned_loss=0.05916, over 17129.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2975, pruned_loss=0.07107, over 3321252.75 frames. ], batch size: 48, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:45,527 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.090e+02 3.971e+02 4.824e+02 1.597e+03, threshold=7.942e+02, percent-clipped=16.0 2023-04-28 06:56:40,174 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 2023-04-28 06:56:41,608 INFO [train.py:904] (5/8) Epoch 5, batch 2900, loss[loss=0.2139, simple_loss=0.287, pruned_loss=0.07035, over 16419.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.296, pruned_loss=0.07105, over 3321312.55 frames. ], batch size: 36, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:57:00,358 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:01,497 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:28,564 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:49,474 INFO [train.py:904] (5/8) Epoch 5, batch 2950, loss[loss=0.2131, simple_loss=0.2988, pruned_loss=0.06369, over 16679.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2958, pruned_loss=0.07205, over 3316827.81 frames. ], batch size: 57, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:02,031 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 3.523e+02 4.392e+02 5.474e+02 1.022e+03, threshold=8.784e+02, percent-clipped=2.0 2023-04-28 06:58:05,994 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:24,112 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:34,167 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:45,378 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9124, 4.2389, 4.4894, 2.0317, 4.7702, 4.8198, 3.3326, 3.8577], device='cuda:5'), covar=tensor([0.0640, 0.0105, 0.0165, 0.1049, 0.0045, 0.0051, 0.0294, 0.0282], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0083, 0.0085, 0.0138, 0.0071, 0.0078, 0.0113, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 06:58:59,871 INFO [train.py:904] (5/8) Epoch 5, batch 3000, loss[loss=0.213, simple_loss=0.2832, pruned_loss=0.07137, over 16327.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2951, pruned_loss=0.07157, over 3320403.33 frames. ], batch size: 165, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:59,871 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 06:59:08,835 INFO [train.py:938] (5/8) Epoch 5, validation: loss=0.1564, simple_loss=0.2632, pruned_loss=0.02477, over 944034.00 frames. 2023-04-28 06:59:08,836 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 06:59:34,115 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1860, 3.9899, 4.0489, 1.6458, 4.3541, 4.4974, 3.2789, 3.2270], device='cuda:5'), covar=tensor([0.1078, 0.0106, 0.0187, 0.1305, 0.0068, 0.0050, 0.0325, 0.0468], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0083, 0.0084, 0.0138, 0.0070, 0.0078, 0.0113, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 07:00:18,537 INFO [train.py:904] (5/8) Epoch 5, batch 3050, loss[loss=0.2017, simple_loss=0.2738, pruned_loss=0.06482, over 16023.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2949, pruned_loss=0.07143, over 3321543.80 frames. ], batch size: 35, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:00:31,428 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.472e+02 3.368e+02 3.840e+02 5.233e+02 1.219e+03, threshold=7.679e+02, percent-clipped=3.0 2023-04-28 07:01:25,947 INFO [train.py:904] (5/8) Epoch 5, batch 3100, loss[loss=0.2081, simple_loss=0.2774, pruned_loss=0.06936, over 16792.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2949, pruned_loss=0.07198, over 3311787.38 frames. ], batch size: 102, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:02:33,561 INFO [train.py:904] (5/8) Epoch 5, batch 3150, loss[loss=0.2328, simple_loss=0.2906, pruned_loss=0.08753, over 16476.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2937, pruned_loss=0.07082, over 3308348.70 frames. ], batch size: 146, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:02:46,880 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.292e+02 2.988e+02 3.776e+02 4.574e+02 1.068e+03, threshold=7.553e+02, percent-clipped=4.0 2023-04-28 07:02:47,379 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1863, 5.0892, 4.9914, 4.3641, 4.9373, 1.9289, 4.6983, 5.0418], device='cuda:5'), covar=tensor([0.0059, 0.0057, 0.0076, 0.0344, 0.0061, 0.1600, 0.0101, 0.0136], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0087, 0.0130, 0.0137, 0.0100, 0.0140, 0.0114, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:03:22,314 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:03:37,503 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 07:03:42,799 INFO [train.py:904] (5/8) Epoch 5, batch 3200, loss[loss=0.2369, simple_loss=0.3079, pruned_loss=0.08295, over 16755.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2928, pruned_loss=0.07003, over 3314158.96 frames. ], batch size: 83, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:03:55,523 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 07:04:17,753 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6010, 2.5714, 1.8443, 2.3411, 3.0630, 2.8524, 3.8068, 3.4704], device='cuda:5'), covar=tensor([0.0033, 0.0165, 0.0239, 0.0180, 0.0118, 0.0161, 0.0075, 0.0077], device='cuda:5'), in_proj_covar=tensor([0.0085, 0.0151, 0.0151, 0.0148, 0.0151, 0.0156, 0.0131, 0.0137], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:04:48,331 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:04:52,134 INFO [train.py:904] (5/8) Epoch 5, batch 3250, loss[loss=0.197, simple_loss=0.286, pruned_loss=0.05397, over 17181.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2927, pruned_loss=0.07, over 3321003.60 frames. ], batch size: 46, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:05:06,670 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.952e+02 3.819e+02 4.745e+02 7.662e+02, threshold=7.638e+02, percent-clipped=2.0 2023-04-28 07:05:19,459 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:06:03,328 INFO [train.py:904] (5/8) Epoch 5, batch 3300, loss[loss=0.189, simple_loss=0.2665, pruned_loss=0.05574, over 16829.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2939, pruned_loss=0.07044, over 3321530.42 frames. ], batch size: 39, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:06:14,718 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:06:43,151 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-04-28 07:06:45,744 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 07:07:12,759 INFO [train.py:904] (5/8) Epoch 5, batch 3350, loss[loss=0.2512, simple_loss=0.3157, pruned_loss=0.09337, over 16306.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2953, pruned_loss=0.0712, over 3311878.50 frames. ], batch size: 165, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:07:24,261 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43959.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:07:26,941 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 2.864e+02 3.521e+02 4.539e+02 9.461e+02, threshold=7.042e+02, percent-clipped=3.0 2023-04-28 07:07:28,560 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4306, 4.3131, 4.3316, 3.8602, 4.3024, 1.7831, 4.1163, 4.1209], device='cuda:5'), covar=tensor([0.0066, 0.0065, 0.0080, 0.0234, 0.0061, 0.1389, 0.0078, 0.0122], device='cuda:5'), in_proj_covar=tensor([0.0097, 0.0086, 0.0129, 0.0136, 0.0100, 0.0138, 0.0114, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:07:39,255 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:08:22,245 INFO [train.py:904] (5/8) Epoch 5, batch 3400, loss[loss=0.2053, simple_loss=0.2899, pruned_loss=0.06039, over 17123.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2946, pruned_loss=0.07046, over 3321122.78 frames. ], batch size: 49, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:08:47,633 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:09:05,596 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3878, 4.3251, 4.2684, 4.1733, 3.8675, 4.3326, 4.1660, 4.0784], device='cuda:5'), covar=tensor([0.0416, 0.0305, 0.0213, 0.0185, 0.0871, 0.0281, 0.0430, 0.0475], device='cuda:5'), in_proj_covar=tensor([0.0200, 0.0203, 0.0236, 0.0208, 0.0276, 0.0234, 0.0174, 0.0261], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:09:31,576 INFO [train.py:904] (5/8) Epoch 5, batch 3450, loss[loss=0.2292, simple_loss=0.3197, pruned_loss=0.06932, over 16663.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2922, pruned_loss=0.06934, over 3318299.08 frames. ], batch size: 57, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:09:44,810 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 3.034e+02 3.683e+02 4.468e+02 1.074e+03, threshold=7.367e+02, percent-clipped=2.0 2023-04-28 07:10:01,335 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0982, 4.4494, 3.3556, 2.6025, 3.3506, 2.4076, 4.6566, 4.6019], device='cuda:5'), covar=tensor([0.2085, 0.0592, 0.1214, 0.1583, 0.2365, 0.1533, 0.0302, 0.0517], device='cuda:5'), in_proj_covar=tensor([0.0278, 0.0253, 0.0268, 0.0243, 0.0306, 0.0204, 0.0238, 0.0268], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:10:02,268 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9912, 5.3553, 5.4777, 5.2676, 5.2575, 5.8170, 5.5652, 5.3223], device='cuda:5'), covar=tensor([0.0790, 0.1340, 0.1226, 0.1501, 0.2498, 0.0828, 0.0845, 0.1781], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0409, 0.0393, 0.0349, 0.0467, 0.0424, 0.0328, 0.0472], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:10:04,877 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2716, 1.8396, 2.6682, 3.0415, 3.0473, 3.4485, 2.2515, 3.2348], device='cuda:5'), covar=tensor([0.0067, 0.0226, 0.0118, 0.0125, 0.0101, 0.0078, 0.0160, 0.0068], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0144, 0.0128, 0.0132, 0.0128, 0.0095, 0.0139, 0.0083], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 07:10:39,130 INFO [train.py:904] (5/8) Epoch 5, batch 3500, loss[loss=0.241, simple_loss=0.306, pruned_loss=0.088, over 16683.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2912, pruned_loss=0.06931, over 3318968.88 frames. ], batch size: 124, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:10:44,307 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7757, 4.9212, 5.0455, 4.9105, 4.8676, 5.4997, 5.2146, 4.8919], device='cuda:5'), covar=tensor([0.1047, 0.1562, 0.1211, 0.1791, 0.2653, 0.0944, 0.1017, 0.2133], device='cuda:5'), in_proj_covar=tensor([0.0292, 0.0409, 0.0394, 0.0349, 0.0468, 0.0424, 0.0329, 0.0474], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:11:06,715 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 07:11:12,083 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3815, 3.1891, 3.5712, 2.6256, 3.3459, 3.6885, 3.5127, 2.1307], device='cuda:5'), covar=tensor([0.0253, 0.0122, 0.0032, 0.0188, 0.0046, 0.0035, 0.0040, 0.0252], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0060, 0.0061, 0.0114, 0.0062, 0.0069, 0.0067, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:11:37,837 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:11:49,623 INFO [train.py:904] (5/8) Epoch 5, batch 3550, loss[loss=0.2204, simple_loss=0.3041, pruned_loss=0.0684, over 17111.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2905, pruned_loss=0.06898, over 3322061.53 frames. ], batch size: 49, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:12:03,034 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.484e+02 3.064e+02 3.621e+02 4.319e+02 6.916e+02, threshold=7.242e+02, percent-clipped=0.0 2023-04-28 07:12:17,242 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:12:23,275 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8748, 2.5314, 1.7727, 2.2385, 2.9195, 2.6839, 3.2509, 3.1287], device='cuda:5'), covar=tensor([0.0059, 0.0145, 0.0211, 0.0193, 0.0092, 0.0155, 0.0080, 0.0087], device='cuda:5'), in_proj_covar=tensor([0.0084, 0.0150, 0.0149, 0.0147, 0.0149, 0.0154, 0.0133, 0.0136], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:12:59,212 INFO [train.py:904] (5/8) Epoch 5, batch 3600, loss[loss=0.208, simple_loss=0.2734, pruned_loss=0.0713, over 16795.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2893, pruned_loss=0.06841, over 3310749.89 frames. ], batch size: 116, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:13:23,287 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:14:09,675 INFO [train.py:904] (5/8) Epoch 5, batch 3650, loss[loss=0.2023, simple_loss=0.2691, pruned_loss=0.06776, over 16935.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2877, pruned_loss=0.06842, over 3317764.75 frames. ], batch size: 116, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:14:25,434 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 3.015e+02 3.671e+02 4.591e+02 9.976e+02, threshold=7.342e+02, percent-clipped=3.0 2023-04-28 07:14:30,541 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:15:22,483 INFO [train.py:904] (5/8) Epoch 5, batch 3700, loss[loss=0.1932, simple_loss=0.2626, pruned_loss=0.06193, over 16614.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2858, pruned_loss=0.06938, over 3306291.43 frames. ], batch size: 89, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:15:40,352 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2309, 3.4755, 4.0568, 2.7341, 3.6590, 4.0718, 3.8808, 2.2986], device='cuda:5'), covar=tensor([0.0339, 0.0102, 0.0032, 0.0228, 0.0039, 0.0035, 0.0038, 0.0289], device='cuda:5'), in_proj_covar=tensor([0.0115, 0.0059, 0.0060, 0.0113, 0.0061, 0.0068, 0.0065, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:15:45,306 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:16:36,542 INFO [train.py:904] (5/8) Epoch 5, batch 3750, loss[loss=0.2079, simple_loss=0.2743, pruned_loss=0.07068, over 16776.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2867, pruned_loss=0.07116, over 3272736.33 frames. ], batch size: 102, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:16:52,718 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.955e+02 3.644e+02 4.475e+02 7.424e+02, threshold=7.289e+02, percent-clipped=1.0 2023-04-28 07:16:59,798 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6813, 3.6770, 2.7485, 2.2776, 2.8126, 2.1909, 3.5174, 3.7484], device='cuda:5'), covar=tensor([0.1955, 0.0609, 0.1111, 0.1507, 0.1952, 0.1487, 0.0463, 0.0619], device='cuda:5'), in_proj_covar=tensor([0.0276, 0.0252, 0.0268, 0.0243, 0.0304, 0.0206, 0.0238, 0.0266], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:17:49,983 INFO [train.py:904] (5/8) Epoch 5, batch 3800, loss[loss=0.2246, simple_loss=0.3014, pruned_loss=0.07394, over 16494.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2885, pruned_loss=0.0733, over 3266935.77 frames. ], batch size: 35, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:17:51,571 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-28 07:18:30,759 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-28 07:18:50,558 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:19:00,304 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0115, 4.0840, 4.4168, 4.4318, 4.4089, 4.0576, 4.1178, 4.0468], device='cuda:5'), covar=tensor([0.0263, 0.0374, 0.0299, 0.0327, 0.0340, 0.0271, 0.0655, 0.0470], device='cuda:5'), in_proj_covar=tensor([0.0245, 0.0236, 0.0241, 0.0244, 0.0290, 0.0254, 0.0360, 0.0216], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 07:19:01,037 INFO [train.py:904] (5/8) Epoch 5, batch 3850, loss[loss=0.2041, simple_loss=0.2753, pruned_loss=0.06643, over 16637.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2885, pruned_loss=0.07399, over 3269516.91 frames. ], batch size: 134, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:19:16,586 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.978e+02 3.515e+02 4.161e+02 7.942e+02, threshold=7.030e+02, percent-clipped=2.0 2023-04-28 07:19:57,734 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44491.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:20:11,839 INFO [train.py:904] (5/8) Epoch 5, batch 3900, loss[loss=0.2228, simple_loss=0.2867, pruned_loss=0.07941, over 16760.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2885, pruned_loss=0.07458, over 3267047.25 frames. ], batch size: 134, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:20:49,705 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8697, 5.2112, 5.3586, 5.3133, 5.2959, 5.7846, 5.4220, 5.1772], device='cuda:5'), covar=tensor([0.0884, 0.1248, 0.1246, 0.1379, 0.1824, 0.0762, 0.1064, 0.1980], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0397, 0.0389, 0.0339, 0.0448, 0.0411, 0.0321, 0.0459], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:20:49,817 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:20:54,762 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6558, 5.0082, 4.6614, 4.7018, 4.3667, 4.3126, 4.4542, 5.0102], device='cuda:5'), covar=tensor([0.0670, 0.0674, 0.0921, 0.0499, 0.0644, 0.0942, 0.0676, 0.0716], device='cuda:5'), in_proj_covar=tensor([0.0366, 0.0493, 0.0418, 0.0322, 0.0313, 0.0320, 0.0399, 0.0353], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:21:22,448 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44550.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:21:23,258 INFO [train.py:904] (5/8) Epoch 5, batch 3950, loss[loss=0.2234, simple_loss=0.2866, pruned_loss=0.08011, over 16502.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.287, pruned_loss=0.07459, over 3279207.28 frames. ], batch size: 75, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:21:24,994 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9268, 1.6984, 2.2107, 2.8497, 2.7982, 2.8355, 1.3956, 3.0275], device='cuda:5'), covar=tensor([0.0053, 0.0192, 0.0144, 0.0083, 0.0072, 0.0067, 0.0240, 0.0036], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0142, 0.0128, 0.0128, 0.0127, 0.0093, 0.0138, 0.0082], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 07:21:37,716 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.214e+02 3.071e+02 3.714e+02 4.451e+02 1.184e+03, threshold=7.429e+02, percent-clipped=3.0 2023-04-28 07:21:44,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8467, 4.1611, 3.8921, 3.9554, 3.6224, 3.6173, 3.7177, 4.0994], device='cuda:5'), covar=tensor([0.0709, 0.0747, 0.0922, 0.0540, 0.0628, 0.1600, 0.0763, 0.0878], device='cuda:5'), in_proj_covar=tensor([0.0369, 0.0497, 0.0421, 0.0325, 0.0315, 0.0324, 0.0402, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:21:44,259 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:16,299 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:34,894 INFO [train.py:904] (5/8) Epoch 5, batch 4000, loss[loss=0.2144, simple_loss=0.2769, pruned_loss=0.07592, over 16850.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2873, pruned_loss=0.07513, over 3264101.60 frames. ], batch size: 116, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:22:49,514 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44611.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:51,786 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:54,899 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:23:00,997 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6744, 3.8245, 2.8260, 2.3973, 2.7641, 2.2035, 3.6393, 3.8322], device='cuda:5'), covar=tensor([0.2064, 0.0495, 0.1203, 0.1610, 0.2024, 0.1514, 0.0463, 0.0615], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0252, 0.0268, 0.0248, 0.0312, 0.0207, 0.0239, 0.0267], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:23:45,848 INFO [train.py:904] (5/8) Epoch 5, batch 4050, loss[loss=0.2278, simple_loss=0.2966, pruned_loss=0.07948, over 16617.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2868, pruned_loss=0.07327, over 3268653.81 frames. ], batch size: 134, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:24:02,948 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.760e+02 3.326e+02 4.025e+02 7.816e+02, threshold=6.651e+02, percent-clipped=1.0 2023-04-28 07:24:04,437 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:24:59,132 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2279, 4.2976, 4.5437, 4.5093, 4.5994, 4.2103, 4.0831, 3.9814], device='cuda:5'), covar=tensor([0.0298, 0.0389, 0.0419, 0.0627, 0.0462, 0.0368, 0.1017, 0.0474], device='cuda:5'), in_proj_covar=tensor([0.0250, 0.0243, 0.0245, 0.0251, 0.0297, 0.0260, 0.0374, 0.0220], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 07:25:01,125 INFO [train.py:904] (5/8) Epoch 5, batch 4100, loss[loss=0.2116, simple_loss=0.2978, pruned_loss=0.06275, over 16746.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2875, pruned_loss=0.07185, over 3263829.88 frames. ], batch size: 83, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:26:16,728 INFO [train.py:904] (5/8) Epoch 5, batch 4150, loss[loss=0.3565, simple_loss=0.397, pruned_loss=0.158, over 11793.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.2966, pruned_loss=0.07586, over 3240475.67 frames. ], batch size: 248, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:26:34,908 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.818e+02 3.479e+02 4.379e+02 9.731e+02, threshold=6.958e+02, percent-clipped=3.0 2023-04-28 07:27:16,737 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 07:27:35,382 INFO [train.py:904] (5/8) Epoch 5, batch 4200, loss[loss=0.2699, simple_loss=0.3485, pruned_loss=0.09567, over 15288.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3043, pruned_loss=0.0788, over 3192942.97 frames. ], batch size: 190, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:28:51,249 INFO [train.py:904] (5/8) Epoch 5, batch 4250, loss[loss=0.2081, simple_loss=0.2991, pruned_loss=0.05855, over 16589.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3076, pruned_loss=0.07907, over 3175446.59 frames. ], batch size: 68, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:29:07,187 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 3.072e+02 3.670e+02 4.674e+02 1.163e+03, threshold=7.340e+02, percent-clipped=6.0 2023-04-28 07:29:18,444 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9621, 1.6601, 1.4760, 1.5566, 1.8173, 1.6856, 1.9199, 1.9007], device='cuda:5'), covar=tensor([0.0043, 0.0150, 0.0189, 0.0166, 0.0105, 0.0147, 0.0077, 0.0105], device='cuda:5'), in_proj_covar=tensor([0.0080, 0.0150, 0.0151, 0.0147, 0.0146, 0.0153, 0.0124, 0.0136], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:29:38,861 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44883.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:29:44,217 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 07:30:03,043 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3202, 2.0404, 2.3055, 3.2100, 3.1350, 3.6539, 1.8936, 3.4959], device='cuda:5'), covar=tensor([0.0067, 0.0207, 0.0159, 0.0076, 0.0077, 0.0052, 0.0216, 0.0034], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0142, 0.0127, 0.0126, 0.0126, 0.0091, 0.0138, 0.0080], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 07:30:06,015 INFO [train.py:904] (5/8) Epoch 5, batch 4300, loss[loss=0.2246, simple_loss=0.305, pruned_loss=0.07209, over 17009.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3078, pruned_loss=0.07711, over 3179688.55 frames. ], batch size: 55, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:30:13,101 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:30:52,115 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3463, 2.7153, 2.7416, 4.9031, 2.2295, 3.7129, 2.6510, 2.7946], device='cuda:5'), covar=tensor([0.0412, 0.1699, 0.0871, 0.0184, 0.2503, 0.0741, 0.1544, 0.1997], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0322, 0.0259, 0.0309, 0.0364, 0.0308, 0.0287, 0.0391], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:31:19,636 INFO [train.py:904] (5/8) Epoch 5, batch 4350, loss[loss=0.2461, simple_loss=0.3231, pruned_loss=0.08453, over 16651.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3116, pruned_loss=0.07859, over 3175528.89 frames. ], batch size: 134, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:31:36,573 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 3.143e+02 3.858e+02 4.477e+02 9.466e+02, threshold=7.715e+02, percent-clipped=1.0 2023-04-28 07:32:25,611 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3498, 2.8970, 2.5396, 2.3439, 2.2723, 2.0042, 2.8555, 2.9460], device='cuda:5'), covar=tensor([0.1602, 0.0573, 0.1008, 0.1144, 0.1791, 0.1308, 0.0342, 0.0505], device='cuda:5'), in_proj_covar=tensor([0.0277, 0.0251, 0.0271, 0.0244, 0.0312, 0.0204, 0.0239, 0.0260], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:32:35,460 INFO [train.py:904] (5/8) Epoch 5, batch 4400, loss[loss=0.267, simple_loss=0.3283, pruned_loss=0.1028, over 11626.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3137, pruned_loss=0.08009, over 3151270.20 frames. ], batch size: 249, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:05,754 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 07:33:48,061 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 07:33:49,609 INFO [train.py:904] (5/8) Epoch 5, batch 4450, loss[loss=0.2423, simple_loss=0.3331, pruned_loss=0.0757, over 16844.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.317, pruned_loss=0.0802, over 3172987.78 frames. ], batch size: 116, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:56,517 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:34:06,120 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.841e+02 3.317e+02 4.350e+02 8.803e+02, threshold=6.635e+02, percent-clipped=2.0 2023-04-28 07:34:19,857 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 07:34:23,598 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 07:34:33,611 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:35:01,846 INFO [train.py:904] (5/8) Epoch 5, batch 4500, loss[loss=0.2165, simple_loss=0.3069, pruned_loss=0.06304, over 16913.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.317, pruned_loss=0.08016, over 3176316.43 frames. ], batch size: 96, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:35:26,551 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:36:02,265 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:36:13,666 INFO [train.py:904] (5/8) Epoch 5, batch 4550, loss[loss=0.2605, simple_loss=0.3252, pruned_loss=0.09788, over 16625.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.317, pruned_loss=0.08028, over 3181138.22 frames. ], batch size: 57, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:36:14,967 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2768, 3.1515, 3.2391, 3.4594, 3.4708, 3.1592, 3.4082, 3.5140], device='cuda:5'), covar=tensor([0.0756, 0.0721, 0.1107, 0.0490, 0.0492, 0.1948, 0.0783, 0.0470], device='cuda:5'), in_proj_covar=tensor([0.0356, 0.0446, 0.0564, 0.0455, 0.0338, 0.0343, 0.0352, 0.0376], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:36:19,219 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9339, 3.9750, 3.7545, 3.6611, 3.4038, 3.8264, 3.5772, 3.5197], device='cuda:5'), covar=tensor([0.0362, 0.0198, 0.0201, 0.0175, 0.0739, 0.0206, 0.0626, 0.0434], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0172, 0.0205, 0.0180, 0.0238, 0.0200, 0.0150, 0.0220], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:36:30,362 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.972e+02 2.621e+02 3.061e+02 3.788e+02 6.051e+02, threshold=6.121e+02, percent-clipped=0.0 2023-04-28 07:37:00,729 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:37:10,523 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8405, 5.0770, 4.8125, 4.7951, 4.5311, 4.3364, 4.5576, 5.1849], device='cuda:5'), covar=tensor([0.0608, 0.0564, 0.0864, 0.0480, 0.0562, 0.0685, 0.0566, 0.0555], device='cuda:5'), in_proj_covar=tensor([0.0345, 0.0458, 0.0399, 0.0301, 0.0294, 0.0308, 0.0378, 0.0336], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:37:27,634 INFO [train.py:904] (5/8) Epoch 5, batch 4600, loss[loss=0.2091, simple_loss=0.2969, pruned_loss=0.0607, over 16828.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3173, pruned_loss=0.07971, over 3182585.49 frames. ], batch size: 102, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:37:30,776 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 07:37:35,990 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:12,282 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:40,743 INFO [train.py:904] (5/8) Epoch 5, batch 4650, loss[loss=0.2592, simple_loss=0.3456, pruned_loss=0.08642, over 16271.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.316, pruned_loss=0.07921, over 3188715.60 frames. ], batch size: 165, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:38:45,218 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:57,177 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.661e+02 3.124e+02 3.951e+02 6.864e+02, threshold=6.249e+02, percent-clipped=3.0 2023-04-28 07:39:27,508 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6366, 3.3150, 2.9523, 1.7085, 2.5078, 2.0381, 3.0502, 3.1736], device='cuda:5'), covar=tensor([0.0264, 0.0497, 0.0517, 0.1696, 0.0838, 0.1024, 0.0616, 0.0670], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0129, 0.0151, 0.0140, 0.0133, 0.0126, 0.0139, 0.0136], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 07:39:55,277 INFO [train.py:904] (5/8) Epoch 5, batch 4700, loss[loss=0.2139, simple_loss=0.2985, pruned_loss=0.06465, over 16787.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3124, pruned_loss=0.07731, over 3192746.29 frames. ], batch size: 124, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:40:40,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2839, 3.2351, 2.6590, 2.1046, 2.4024, 2.0691, 3.4670, 3.4388], device='cuda:5'), covar=tensor([0.2208, 0.0737, 0.1198, 0.1510, 0.1658, 0.1435, 0.0406, 0.0518], device='cuda:5'), in_proj_covar=tensor([0.0283, 0.0254, 0.0271, 0.0247, 0.0309, 0.0206, 0.0239, 0.0255], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:40:54,172 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 07:41:03,171 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:41:06,794 INFO [train.py:904] (5/8) Epoch 5, batch 4750, loss[loss=0.2006, simple_loss=0.2822, pruned_loss=0.05951, over 16379.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3083, pruned_loss=0.07515, over 3197679.69 frames. ], batch size: 75, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:41:22,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.710e+02 3.330e+02 4.145e+02 7.902e+02, threshold=6.661e+02, percent-clipped=5.0 2023-04-28 07:42:08,158 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3499, 2.8495, 2.5274, 2.2481, 2.2417, 2.0678, 2.9149, 3.0225], device='cuda:5'), covar=tensor([0.1673, 0.0793, 0.1093, 0.1376, 0.1658, 0.1350, 0.0424, 0.0548], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0252, 0.0268, 0.0244, 0.0305, 0.0203, 0.0238, 0.0253], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:42:20,537 INFO [train.py:904] (5/8) Epoch 5, batch 4800, loss[loss=0.2156, simple_loss=0.3021, pruned_loss=0.06455, over 17132.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3048, pruned_loss=0.07299, over 3208043.13 frames. ], batch size: 47, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:42:33,265 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 07:42:38,691 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:43:14,519 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:43:34,382 INFO [train.py:904] (5/8) Epoch 5, batch 4850, loss[loss=0.1964, simple_loss=0.2831, pruned_loss=0.0549, over 16488.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3061, pruned_loss=0.07301, over 3200540.59 frames. ], batch size: 68, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:43:37,225 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 07:43:50,677 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.510e+02 3.182e+02 3.913e+02 7.711e+02, threshold=6.364e+02, percent-clipped=1.0 2023-04-28 07:44:25,004 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-04-28 07:44:47,402 INFO [train.py:904] (5/8) Epoch 5, batch 4900, loss[loss=0.2114, simple_loss=0.2912, pruned_loss=0.06575, over 16598.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3062, pruned_loss=0.07255, over 3186935.52 frames. ], batch size: 62, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:46:01,017 INFO [train.py:904] (5/8) Epoch 5, batch 4950, loss[loss=0.2532, simple_loss=0.3265, pruned_loss=0.08996, over 12214.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3061, pruned_loss=0.07232, over 3187463.77 frames. ], batch size: 246, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:46:10,792 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4130, 2.1462, 2.1511, 3.9630, 1.7386, 2.8154, 2.1404, 2.2216], device='cuda:5'), covar=tensor([0.0678, 0.2324, 0.1190, 0.0434, 0.3455, 0.1121, 0.2125, 0.2557], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0317, 0.0256, 0.0306, 0.0364, 0.0301, 0.0281, 0.0380], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:46:15,483 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.470e+02 3.018e+02 3.573e+02 4.428e+02 9.346e+02, threshold=7.147e+02, percent-clipped=9.0 2023-04-28 07:46:26,090 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-28 07:47:13,812 INFO [train.py:904] (5/8) Epoch 5, batch 5000, loss[loss=0.2194, simple_loss=0.3015, pruned_loss=0.06862, over 16495.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3071, pruned_loss=0.07206, over 3201932.40 frames. ], batch size: 68, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:47:45,427 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8608, 4.6414, 4.8689, 5.1197, 5.2915, 4.6625, 5.2280, 5.2309], device='cuda:5'), covar=tensor([0.0817, 0.0740, 0.1047, 0.0383, 0.0275, 0.0478, 0.0279, 0.0300], device='cuda:5'), in_proj_covar=tensor([0.0366, 0.0445, 0.0563, 0.0455, 0.0338, 0.0341, 0.0355, 0.0369], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:48:15,777 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0077, 2.5174, 2.5009, 4.7688, 2.0253, 3.5244, 2.5920, 2.7280], device='cuda:5'), covar=tensor([0.0509, 0.1752, 0.0923, 0.0199, 0.2618, 0.0791, 0.1628, 0.1875], device='cuda:5'), in_proj_covar=tensor([0.0313, 0.0317, 0.0258, 0.0306, 0.0364, 0.0303, 0.0283, 0.0382], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 07:48:27,066 INFO [train.py:904] (5/8) Epoch 5, batch 5050, loss[loss=0.2161, simple_loss=0.2958, pruned_loss=0.06818, over 16833.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3072, pruned_loss=0.07187, over 3200669.73 frames. ], batch size: 42, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:48:41,278 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0578, 4.9644, 4.8385, 4.7164, 4.3332, 4.8643, 4.8831, 4.5752], device='cuda:5'), covar=tensor([0.0373, 0.0237, 0.0174, 0.0158, 0.0884, 0.0252, 0.0145, 0.0349], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0182, 0.0214, 0.0185, 0.0244, 0.0212, 0.0153, 0.0229], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:48:43,130 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.781e+02 3.302e+02 3.985e+02 6.386e+02, threshold=6.604e+02, percent-clipped=0.0 2023-04-28 07:49:39,982 INFO [train.py:904] (5/8) Epoch 5, batch 5100, loss[loss=0.193, simple_loss=0.2735, pruned_loss=0.05629, over 16647.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3056, pruned_loss=0.07134, over 3191298.17 frames. ], batch size: 62, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:49:42,331 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:49:44,513 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 07:49:56,751 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:50:33,734 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:50:53,045 INFO [train.py:904] (5/8) Epoch 5, batch 5150, loss[loss=0.2061, simple_loss=0.3049, pruned_loss=0.05365, over 16845.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3056, pruned_loss=0.07077, over 3166155.20 frames. ], batch size: 102, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:51:07,068 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:08,847 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.626e+02 3.241e+02 3.847e+02 8.905e+02, threshold=6.482e+02, percent-clipped=5.0 2023-04-28 07:51:10,604 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:26,065 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45774.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:43,503 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45785.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:48,415 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2662, 4.0094, 3.4853, 1.9416, 3.0988, 2.4144, 3.6372, 3.8613], device='cuda:5'), covar=tensor([0.0199, 0.0389, 0.0516, 0.1467, 0.0599, 0.0788, 0.0458, 0.0482], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0126, 0.0153, 0.0140, 0.0133, 0.0126, 0.0138, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 07:52:05,645 INFO [train.py:904] (5/8) Epoch 5, batch 5200, loss[loss=0.2007, simple_loss=0.287, pruned_loss=0.05723, over 16834.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3047, pruned_loss=0.07079, over 3173185.11 frames. ], batch size: 83, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:52:50,637 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8094, 2.2078, 1.6670, 1.9368, 2.6584, 2.4431, 2.9266, 2.9016], device='cuda:5'), covar=tensor([0.0033, 0.0183, 0.0255, 0.0229, 0.0098, 0.0158, 0.0057, 0.0084], device='cuda:5'), in_proj_covar=tensor([0.0077, 0.0150, 0.0153, 0.0150, 0.0147, 0.0153, 0.0120, 0.0135], device='cuda:5'), out_proj_covar=tensor([9.7769e-05, 1.8527e-04, 1.8511e-04, 1.8067e-04, 1.8369e-04, 1.8880e-04, 1.4428e-04, 1.6684e-04], device='cuda:5') 2023-04-28 07:52:54,249 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:53:16,344 INFO [train.py:904] (5/8) Epoch 5, batch 5250, loss[loss=0.1858, simple_loss=0.2712, pruned_loss=0.05018, over 16576.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3027, pruned_loss=0.0708, over 3165437.73 frames. ], batch size: 75, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:53:19,329 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 07:53:27,282 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:53:31,063 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.734e+02 3.218e+02 3.993e+02 9.159e+02, threshold=6.436e+02, percent-clipped=4.0 2023-04-28 07:53:57,133 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 07:54:14,927 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-28 07:54:26,352 INFO [train.py:904] (5/8) Epoch 5, batch 5300, loss[loss=0.2047, simple_loss=0.2757, pruned_loss=0.06687, over 17003.00 frames. ], tot_loss[loss=0.219, simple_loss=0.299, pruned_loss=0.0695, over 3168368.97 frames. ], batch size: 55, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:54:53,145 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:55:15,648 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4677, 3.7803, 1.5042, 4.0788, 2.5567, 3.8432, 1.6860, 2.7538], device='cuda:5'), covar=tensor([0.0172, 0.0195, 0.2155, 0.0034, 0.0722, 0.0303, 0.1803, 0.0623], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0147, 0.0176, 0.0080, 0.0159, 0.0180, 0.0185, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 07:55:17,950 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1022, 4.1224, 3.9253, 3.8684, 3.5245, 4.0320, 3.7958, 3.7651], device='cuda:5'), covar=tensor([0.0354, 0.0285, 0.0206, 0.0179, 0.0773, 0.0279, 0.0488, 0.0408], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0191, 0.0219, 0.0189, 0.0247, 0.0216, 0.0155, 0.0235], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:55:36,415 INFO [train.py:904] (5/8) Epoch 5, batch 5350, loss[loss=0.197, simple_loss=0.2843, pruned_loss=0.05482, over 16532.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2969, pruned_loss=0.06775, over 3186495.29 frames. ], batch size: 68, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:55:49,901 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1369, 3.7317, 3.6846, 2.3706, 3.2121, 3.5894, 3.5169, 1.9111], device='cuda:5'), covar=tensor([0.0340, 0.0015, 0.0024, 0.0248, 0.0052, 0.0042, 0.0030, 0.0321], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0054, 0.0060, 0.0115, 0.0060, 0.0068, 0.0064, 0.0108], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 07:55:53,056 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.936e+02 3.263e+02 4.063e+02 1.076e+03, threshold=6.526e+02, percent-clipped=3.0 2023-04-28 07:56:52,503 INFO [train.py:904] (5/8) Epoch 5, batch 5400, loss[loss=0.2457, simple_loss=0.3218, pruned_loss=0.08482, over 17049.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3007, pruned_loss=0.06951, over 3186463.13 frames. ], batch size: 50, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:56:58,418 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 07:57:51,646 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6345, 3.9888, 4.2130, 1.8374, 4.3715, 4.4890, 3.2201, 3.1490], device='cuda:5'), covar=tensor([0.0737, 0.0118, 0.0086, 0.1116, 0.0037, 0.0028, 0.0277, 0.0394], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0088, 0.0079, 0.0142, 0.0070, 0.0074, 0.0115, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 07:58:09,293 INFO [train.py:904] (5/8) Epoch 5, batch 5450, loss[loss=0.2965, simple_loss=0.3635, pruned_loss=0.1147, over 16420.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3039, pruned_loss=0.07122, over 3187974.35 frames. ], batch size: 146, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:58:11,186 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:58:19,242 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:58:24,994 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 3.104e+02 3.655e+02 4.572e+02 1.003e+03, threshold=7.310e+02, percent-clipped=8.0 2023-04-28 07:59:22,173 INFO [train.py:904] (5/8) Epoch 5, batch 5500, loss[loss=0.3024, simple_loss=0.3567, pruned_loss=0.124, over 11816.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3131, pruned_loss=0.07786, over 3159442.83 frames. ], batch size: 250, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:07,611 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:00:13,884 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7455, 2.8977, 2.0046, 2.4612, 3.2701, 2.9422, 3.6551, 3.4519], device='cuda:5'), covar=tensor([0.0020, 0.0133, 0.0241, 0.0196, 0.0083, 0.0144, 0.0079, 0.0085], device='cuda:5'), in_proj_covar=tensor([0.0077, 0.0150, 0.0153, 0.0149, 0.0146, 0.0154, 0.0123, 0.0134], device='cuda:5'), out_proj_covar=tensor([9.7073e-05, 1.8431e-04, 1.8480e-04, 1.7978e-04, 1.8105e-04, 1.8953e-04, 1.4721e-04, 1.6541e-04], device='cuda:5') 2023-04-28 08:00:39,135 INFO [train.py:904] (5/8) Epoch 5, batch 5550, loss[loss=0.308, simple_loss=0.3688, pruned_loss=0.1236, over 16258.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3218, pruned_loss=0.08496, over 3140326.86 frames. ], batch size: 165, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:56,875 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.500e+02 4.429e+02 5.470e+02 6.755e+02 1.488e+03, threshold=1.094e+03, percent-clipped=17.0 2023-04-28 08:01:58,729 INFO [train.py:904] (5/8) Epoch 5, batch 5600, loss[loss=0.3163, simple_loss=0.3552, pruned_loss=0.1387, over 11267.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3289, pruned_loss=0.0921, over 3077113.52 frames. ], batch size: 246, lr: 1.33e-02, grad_scale: 16.0 2023-04-28 08:02:10,177 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2927, 5.5664, 5.2124, 5.3400, 4.9160, 4.6021, 5.1122, 5.6502], device='cuda:5'), covar=tensor([0.0639, 0.0573, 0.0853, 0.0414, 0.0561, 0.0631, 0.0517, 0.0612], device='cuda:5'), in_proj_covar=tensor([0.0352, 0.0468, 0.0408, 0.0305, 0.0291, 0.0310, 0.0380, 0.0336], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:02:23,969 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:03:23,171 INFO [train.py:904] (5/8) Epoch 5, batch 5650, loss[loss=0.3587, simple_loss=0.3897, pruned_loss=0.1639, over 11525.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3343, pruned_loss=0.09688, over 3055609.00 frames. ], batch size: 248, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:03:42,415 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.653e+02 4.676e+02 6.240e+02 7.908e+02 1.367e+03, threshold=1.248e+03, percent-clipped=1.0 2023-04-28 08:04:03,050 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4701, 1.9223, 1.3118, 1.5796, 2.3494, 2.0907, 2.4890, 2.4573], device='cuda:5'), covar=tensor([0.0042, 0.0202, 0.0313, 0.0249, 0.0108, 0.0196, 0.0071, 0.0111], device='cuda:5'), in_proj_covar=tensor([0.0075, 0.0150, 0.0151, 0.0149, 0.0145, 0.0152, 0.0121, 0.0132], device='cuda:5'), out_proj_covar=tensor([9.4668e-05, 1.8401e-04, 1.8235e-04, 1.7945e-04, 1.8021e-04, 1.8678e-04, 1.4433e-04, 1.6307e-04], device='cuda:5') 2023-04-28 08:04:04,268 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:04:43,834 INFO [train.py:904] (5/8) Epoch 5, batch 5700, loss[loss=0.2741, simple_loss=0.3478, pruned_loss=0.1002, over 16378.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3354, pruned_loss=0.09758, over 3066204.54 frames. ], batch size: 165, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:04:56,884 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 08:05:12,806 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3205, 1.9561, 1.9366, 3.7466, 1.8381, 2.8822, 2.2111, 2.1168], device='cuda:5'), covar=tensor([0.0560, 0.2024, 0.1152, 0.0308, 0.2830, 0.0919, 0.1735, 0.2270], device='cuda:5'), in_proj_covar=tensor([0.0313, 0.0315, 0.0253, 0.0302, 0.0365, 0.0300, 0.0281, 0.0374], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:05:41,882 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:05:58,996 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 08:06:04,716 INFO [train.py:904] (5/8) Epoch 5, batch 5750, loss[loss=0.2467, simple_loss=0.3226, pruned_loss=0.08545, over 17046.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.3387, pruned_loss=0.09908, over 3063030.43 frames. ], batch size: 55, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:06:16,686 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:06:23,503 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.622e+02 4.192e+02 4.948e+02 6.377e+02 1.174e+03, threshold=9.897e+02, percent-clipped=0.0 2023-04-28 08:06:30,151 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7942, 1.4238, 1.5250, 1.7123, 1.8307, 1.7905, 1.5070, 1.6817], device='cuda:5'), covar=tensor([0.0079, 0.0158, 0.0082, 0.0112, 0.0091, 0.0062, 0.0163, 0.0042], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0140, 0.0126, 0.0123, 0.0125, 0.0088, 0.0139, 0.0078], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 08:07:26,292 INFO [train.py:904] (5/8) Epoch 5, batch 5800, loss[loss=0.2553, simple_loss=0.335, pruned_loss=0.08776, over 16881.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3382, pruned_loss=0.09794, over 3047140.51 frames. ], batch size: 116, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:07:27,624 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-04-28 08:07:29,103 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6866, 4.9211, 5.0114, 5.0843, 4.9736, 5.4888, 5.0791, 4.9023], device='cuda:5'), covar=tensor([0.0795, 0.1438, 0.1361, 0.1228, 0.1980, 0.0828, 0.0999, 0.1872], device='cuda:5'), in_proj_covar=tensor([0.0266, 0.0365, 0.0363, 0.0318, 0.0424, 0.0387, 0.0302, 0.0439], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 08:07:35,755 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:14,592 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:31,963 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 08:08:46,735 INFO [train.py:904] (5/8) Epoch 5, batch 5850, loss[loss=0.2412, simple_loss=0.3215, pruned_loss=0.08045, over 16880.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3353, pruned_loss=0.0952, over 3065882.07 frames. ], batch size: 116, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:09:06,687 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.458e+02 3.799e+02 4.852e+02 6.197e+02 1.255e+03, threshold=9.704e+02, percent-clipped=4.0 2023-04-28 08:09:29,290 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:10:09,079 INFO [train.py:904] (5/8) Epoch 5, batch 5900, loss[loss=0.2676, simple_loss=0.3429, pruned_loss=0.09613, over 16882.00 frames. ], tot_loss[loss=0.2618, simple_loss=0.3345, pruned_loss=0.09452, over 3074334.89 frames. ], batch size: 109, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:10:17,906 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 08:10:36,172 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:10:36,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0761, 3.0574, 2.7037, 1.9676, 2.5879, 2.0880, 2.7219, 2.7962], device='cuda:5'), covar=tensor([0.0273, 0.0370, 0.0457, 0.1329, 0.0610, 0.0825, 0.0481, 0.0523], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0123, 0.0151, 0.0137, 0.0131, 0.0123, 0.0138, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 08:10:43,099 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5430, 4.0634, 3.2360, 2.2984, 2.9799, 2.4990, 4.2249, 4.1150], device='cuda:5'), covar=tensor([0.2549, 0.0571, 0.1294, 0.1767, 0.1696, 0.1362, 0.0421, 0.0506], device='cuda:5'), in_proj_covar=tensor([0.0277, 0.0244, 0.0264, 0.0240, 0.0290, 0.0199, 0.0238, 0.0246], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:11:31,954 INFO [train.py:904] (5/8) Epoch 5, batch 5950, loss[loss=0.2553, simple_loss=0.3327, pruned_loss=0.08898, over 16669.00 frames. ], tot_loss[loss=0.2602, simple_loss=0.3349, pruned_loss=0.09281, over 3069866.12 frames. ], batch size: 134, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:11:52,757 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:11:53,494 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 3.478e+02 4.456e+02 5.303e+02 9.511e+02, threshold=8.911e+02, percent-clipped=0.0 2023-04-28 08:11:59,201 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6212, 2.8735, 2.3921, 4.2641, 3.6387, 4.0079, 1.5783, 2.9048], device='cuda:5'), covar=tensor([0.1427, 0.0534, 0.1168, 0.0071, 0.0303, 0.0305, 0.1375, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0141, 0.0168, 0.0083, 0.0175, 0.0175, 0.0163, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 08:12:24,692 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6414, 4.3847, 4.6659, 4.9210, 5.0159, 4.4117, 4.9965, 4.9552], device='cuda:5'), covar=tensor([0.0943, 0.0786, 0.1145, 0.0430, 0.0366, 0.0694, 0.0425, 0.0386], device='cuda:5'), in_proj_covar=tensor([0.0371, 0.0455, 0.0572, 0.0465, 0.0351, 0.0347, 0.0373, 0.0387], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:12:43,365 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:12:51,981 INFO [train.py:904] (5/8) Epoch 5, batch 6000, loss[loss=0.2777, simple_loss=0.3365, pruned_loss=0.1094, over 11711.00 frames. ], tot_loss[loss=0.2589, simple_loss=0.3335, pruned_loss=0.09216, over 3077304.71 frames. ], batch size: 248, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:12:51,981 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 08:13:04,055 INFO [train.py:938] (5/8) Epoch 5, validation: loss=0.1879, simple_loss=0.2992, pruned_loss=0.03826, over 944034.00 frames. 2023-04-28 08:13:04,056 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 08:13:05,777 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:13:51,752 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:14:07,771 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8000, 3.0349, 2.7463, 4.6718, 3.6817, 4.3556, 2.0102, 3.2795], device='cuda:5'), covar=tensor([0.1304, 0.0527, 0.1016, 0.0052, 0.0245, 0.0263, 0.1119, 0.0606], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0141, 0.0168, 0.0083, 0.0177, 0.0176, 0.0161, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 08:14:27,403 INFO [train.py:904] (5/8) Epoch 5, batch 6050, loss[loss=0.2526, simple_loss=0.3425, pruned_loss=0.0813, over 16918.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.3324, pruned_loss=0.09175, over 3059240.00 frames. ], batch size: 116, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:14:37,530 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:14:48,331 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:14:49,008 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 4.108e+02 4.980e+02 6.205e+02 1.363e+03, threshold=9.960e+02, percent-clipped=7.0 2023-04-28 08:15:37,501 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6650, 2.1133, 1.5160, 1.8040, 2.6423, 2.3330, 2.8250, 2.9221], device='cuda:5'), covar=tensor([0.0040, 0.0179, 0.0286, 0.0231, 0.0095, 0.0175, 0.0073, 0.0085], device='cuda:5'), in_proj_covar=tensor([0.0076, 0.0152, 0.0153, 0.0151, 0.0146, 0.0154, 0.0126, 0.0134], device='cuda:5'), out_proj_covar=tensor([9.5490e-05, 1.8644e-04, 1.8373e-04, 1.8169e-04, 1.8095e-04, 1.8974e-04, 1.5049e-04, 1.6566e-04], device='cuda:5') 2023-04-28 08:15:45,459 INFO [train.py:904] (5/8) Epoch 5, batch 6100, loss[loss=0.2123, simple_loss=0.2951, pruned_loss=0.06477, over 16500.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3307, pruned_loss=0.0894, over 3093597.91 frames. ], batch size: 68, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:17:01,204 INFO [train.py:904] (5/8) Epoch 5, batch 6150, loss[loss=0.2515, simple_loss=0.3157, pruned_loss=0.09364, over 16619.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3293, pruned_loss=0.08909, over 3103163.88 frames. ], batch size: 62, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:17:18,243 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8207, 3.6638, 2.9377, 1.7177, 2.5571, 2.1607, 3.1635, 3.4259], device='cuda:5'), covar=tensor([0.0271, 0.0436, 0.0635, 0.1724, 0.0854, 0.0926, 0.0711, 0.0673], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0124, 0.0152, 0.0139, 0.0132, 0.0124, 0.0138, 0.0134], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 08:17:22,721 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.183e+02 3.664e+02 4.778e+02 7.050e+02 1.596e+03, threshold=9.557e+02, percent-clipped=5.0 2023-04-28 08:17:28,736 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2347, 5.4879, 5.1595, 5.2345, 4.8078, 4.6434, 5.0263, 5.5561], device='cuda:5'), covar=tensor([0.0604, 0.0607, 0.0867, 0.0443, 0.0536, 0.0653, 0.0543, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0367, 0.0478, 0.0416, 0.0314, 0.0301, 0.0318, 0.0391, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:18:21,100 INFO [train.py:904] (5/8) Epoch 5, batch 6200, loss[loss=0.2709, simple_loss=0.3247, pruned_loss=0.1085, over 11438.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3273, pruned_loss=0.08848, over 3115530.13 frames. ], batch size: 247, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:18:36,459 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:19:37,526 INFO [train.py:904] (5/8) Epoch 5, batch 6250, loss[loss=0.2113, simple_loss=0.2971, pruned_loss=0.06275, over 16520.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3262, pruned_loss=0.08768, over 3105725.27 frames. ], batch size: 146, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:19:56,665 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5232, 2.2766, 1.6807, 1.8509, 2.6445, 2.3697, 2.8215, 2.8122], device='cuda:5'), covar=tensor([0.0046, 0.0167, 0.0241, 0.0227, 0.0095, 0.0162, 0.0069, 0.0107], device='cuda:5'), in_proj_covar=tensor([0.0076, 0.0151, 0.0153, 0.0150, 0.0146, 0.0153, 0.0125, 0.0134], device='cuda:5'), out_proj_covar=tensor([9.5329e-05, 1.8435e-04, 1.8307e-04, 1.7970e-04, 1.8048e-04, 1.8854e-04, 1.4920e-04, 1.6555e-04], device='cuda:5') 2023-04-28 08:19:57,341 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.405e+02 3.628e+02 4.524e+02 5.799e+02 1.377e+03, threshold=9.049e+02, percent-clipped=5.0 2023-04-28 08:20:10,125 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:20:21,386 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9819, 3.1801, 1.5495, 3.3082, 2.2652, 3.2569, 1.7600, 2.4830], device='cuda:5'), covar=tensor([0.0156, 0.0268, 0.1621, 0.0071, 0.0766, 0.0439, 0.1456, 0.0636], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0147, 0.0178, 0.0081, 0.0162, 0.0183, 0.0186, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 08:20:55,078 INFO [train.py:904] (5/8) Epoch 5, batch 6300, loss[loss=0.283, simple_loss=0.3415, pruned_loss=0.1122, over 11296.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3265, pruned_loss=0.088, over 3083971.60 frames. ], batch size: 247, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:21:43,983 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:22:12,106 INFO [train.py:904] (5/8) Epoch 5, batch 6350, loss[loss=0.2501, simple_loss=0.3238, pruned_loss=0.08824, over 16472.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3292, pruned_loss=0.09082, over 3078642.03 frames. ], batch size: 68, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:22:13,442 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:22:19,516 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 08:22:22,796 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:22:31,678 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.349e+02 3.778e+02 4.701e+02 5.946e+02 1.466e+03, threshold=9.402e+02, percent-clipped=8.0 2023-04-28 08:22:34,366 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9746, 2.4336, 2.3573, 3.2459, 2.7825, 3.2893, 1.8168, 2.6647], device='cuda:5'), covar=tensor([0.1154, 0.0448, 0.0903, 0.0114, 0.0241, 0.0357, 0.1173, 0.0698], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0145, 0.0170, 0.0086, 0.0180, 0.0180, 0.0165, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 08:22:39,917 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7939, 5.2290, 5.3786, 5.2069, 5.2586, 5.7882, 5.4112, 5.1141], device='cuda:5'), covar=tensor([0.0835, 0.1416, 0.1272, 0.1360, 0.2165, 0.0819, 0.1162, 0.2163], device='cuda:5'), in_proj_covar=tensor([0.0273, 0.0375, 0.0379, 0.0331, 0.0435, 0.0399, 0.0308, 0.0456], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 08:22:57,127 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:23:18,929 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-04-28 08:23:28,020 INFO [train.py:904] (5/8) Epoch 5, batch 6400, loss[loss=0.3002, simple_loss=0.3543, pruned_loss=0.123, over 11769.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.328, pruned_loss=0.09085, over 3081899.20 frames. ], batch size: 247, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:23:44,163 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7664, 4.8284, 4.5977, 3.9091, 4.6104, 1.6641, 4.3442, 4.4603], device='cuda:5'), covar=tensor([0.0057, 0.0038, 0.0072, 0.0277, 0.0052, 0.1672, 0.0077, 0.0120], device='cuda:5'), in_proj_covar=tensor([0.0092, 0.0079, 0.0120, 0.0127, 0.0091, 0.0143, 0.0107, 0.0118], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:24:42,542 INFO [train.py:904] (5/8) Epoch 5, batch 6450, loss[loss=0.2446, simple_loss=0.3027, pruned_loss=0.09326, over 12052.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3269, pruned_loss=0.08923, over 3072703.15 frames. ], batch size: 248, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:25:02,353 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.185e+02 3.689e+02 4.668e+02 5.826e+02 1.210e+03, threshold=9.337e+02, percent-clipped=4.0 2023-04-28 08:25:59,703 INFO [train.py:904] (5/8) Epoch 5, batch 6500, loss[loss=0.294, simple_loss=0.3397, pruned_loss=0.1242, over 11157.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3245, pruned_loss=0.08847, over 3072138.58 frames. ], batch size: 246, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:26:47,187 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-28 08:26:49,039 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 08:27:18,012 INFO [train.py:904] (5/8) Epoch 5, batch 6550, loss[loss=0.2333, simple_loss=0.3329, pruned_loss=0.0669, over 16846.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3281, pruned_loss=0.08942, over 3088518.58 frames. ], batch size: 96, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:27:37,111 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.214e+02 3.842e+02 4.732e+02 5.920e+02 1.050e+03, threshold=9.464e+02, percent-clipped=2.0 2023-04-28 08:27:41,306 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:27:57,881 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-28 08:28:00,326 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-28 08:28:33,667 INFO [train.py:904] (5/8) Epoch 5, batch 6600, loss[loss=0.2628, simple_loss=0.3291, pruned_loss=0.09828, over 17003.00 frames. ], tot_loss[loss=0.255, simple_loss=0.3302, pruned_loss=0.08985, over 3105050.42 frames. ], batch size: 55, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:35,157 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:29:51,337 INFO [train.py:904] (5/8) Epoch 5, batch 6650, loss[loss=0.2491, simple_loss=0.3216, pruned_loss=0.08829, over 16746.00 frames. ], tot_loss[loss=0.2568, simple_loss=0.3312, pruned_loss=0.09119, over 3095522.97 frames. ], batch size: 124, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:52,450 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:30:02,774 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:30:05,814 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 08:30:11,440 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.376e+02 3.750e+02 4.777e+02 6.524e+02 9.688e+02, threshold=9.554e+02, percent-clipped=1.0 2023-04-28 08:30:55,207 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7575, 3.4846, 3.0435, 1.7589, 2.7450, 2.1719, 3.1099, 3.2957], device='cuda:5'), covar=tensor([0.0267, 0.0499, 0.0536, 0.1715, 0.0764, 0.0928, 0.0678, 0.0649], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0125, 0.0152, 0.0140, 0.0131, 0.0123, 0.0139, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 08:31:04,891 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:06,916 INFO [train.py:904] (5/8) Epoch 5, batch 6700, loss[loss=0.2409, simple_loss=0.3157, pruned_loss=0.08299, over 16522.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3297, pruned_loss=0.09119, over 3086690.65 frames. ], batch size: 75, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:31:09,274 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:16,567 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:32:26,004 INFO [train.py:904] (5/8) Epoch 5, batch 6750, loss[loss=0.2341, simple_loss=0.3161, pruned_loss=0.07609, over 16717.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3283, pruned_loss=0.09076, over 3108070.73 frames. ], batch size: 89, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:32:34,080 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2964, 3.2865, 3.2943, 3.4833, 3.4916, 3.2280, 3.4375, 3.5047], device='cuda:5'), covar=tensor([0.0773, 0.0596, 0.1020, 0.0523, 0.0572, 0.2040, 0.0851, 0.0566], device='cuda:5'), in_proj_covar=tensor([0.0373, 0.0464, 0.0586, 0.0472, 0.0355, 0.0353, 0.0376, 0.0397], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:32:45,843 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.669e+02 3.980e+02 4.802e+02 5.958e+02 9.394e+02, threshold=9.603e+02, percent-clipped=0.0 2023-04-28 08:33:09,271 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:33:28,281 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0089, 3.8712, 4.0242, 4.2352, 4.3276, 3.9279, 4.2970, 4.2729], device='cuda:5'), covar=tensor([0.0920, 0.0762, 0.1189, 0.0485, 0.0419, 0.1026, 0.0462, 0.0514], device='cuda:5'), in_proj_covar=tensor([0.0372, 0.0461, 0.0582, 0.0470, 0.0350, 0.0349, 0.0373, 0.0395], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:33:40,662 INFO [train.py:904] (5/8) Epoch 5, batch 6800, loss[loss=0.2435, simple_loss=0.3346, pruned_loss=0.07617, over 16925.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3282, pruned_loss=0.09046, over 3103307.91 frames. ], batch size: 96, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:33:53,282 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0179, 1.3809, 1.7312, 2.0472, 2.1125, 2.3297, 1.4570, 2.0931], device='cuda:5'), covar=tensor([0.0082, 0.0222, 0.0116, 0.0111, 0.0117, 0.0057, 0.0208, 0.0041], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0143, 0.0127, 0.0123, 0.0126, 0.0091, 0.0140, 0.0079], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 08:34:41,084 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:34:57,648 INFO [train.py:904] (5/8) Epoch 5, batch 6850, loss[loss=0.3016, simple_loss=0.3549, pruned_loss=0.1242, over 11782.00 frames. ], tot_loss[loss=0.256, simple_loss=0.3297, pruned_loss=0.09117, over 3076778.24 frames. ], batch size: 247, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:35:03,996 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-28 08:35:17,576 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.600e+02 3.467e+02 4.392e+02 5.504e+02 1.073e+03, threshold=8.784e+02, percent-clipped=2.0 2023-04-28 08:35:21,038 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:36:12,149 INFO [train.py:904] (5/8) Epoch 5, batch 6900, loss[loss=0.2656, simple_loss=0.3473, pruned_loss=0.09195, over 16898.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3316, pruned_loss=0.09001, over 3108383.48 frames. ], batch size: 109, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:36:33,891 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:36:43,070 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-28 08:37:05,031 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0913, 4.0611, 3.9821, 3.3776, 4.0430, 1.7294, 3.8004, 3.8802], device='cuda:5'), covar=tensor([0.0074, 0.0064, 0.0094, 0.0302, 0.0060, 0.1715, 0.0101, 0.0136], device='cuda:5'), in_proj_covar=tensor([0.0089, 0.0076, 0.0118, 0.0123, 0.0087, 0.0139, 0.0104, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:37:29,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3187, 4.3203, 4.1270, 3.2495, 4.2195, 1.5076, 3.8982, 3.9589], device='cuda:5'), covar=tensor([0.0074, 0.0054, 0.0099, 0.0403, 0.0058, 0.2007, 0.0100, 0.0163], device='cuda:5'), in_proj_covar=tensor([0.0089, 0.0076, 0.0118, 0.0123, 0.0088, 0.0139, 0.0104, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:37:30,520 INFO [train.py:904] (5/8) Epoch 5, batch 6950, loss[loss=0.2619, simple_loss=0.3298, pruned_loss=0.09706, over 16680.00 frames. ], tot_loss[loss=0.2594, simple_loss=0.3339, pruned_loss=0.09247, over 3094196.87 frames. ], batch size: 134, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:37:50,873 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 3.906e+02 5.173e+02 7.099e+02 1.028e+03, threshold=1.035e+03, percent-clipped=6.0 2023-04-28 08:37:57,167 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 08:38:18,105 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 08:38:42,859 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:38:48,786 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-28 08:38:49,101 INFO [train.py:904] (5/8) Epoch 5, batch 7000, loss[loss=0.2315, simple_loss=0.3297, pruned_loss=0.06667, over 16394.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3342, pruned_loss=0.09172, over 3101320.18 frames. ], batch size: 68, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:40:06,956 INFO [train.py:904] (5/8) Epoch 5, batch 7050, loss[loss=0.2428, simple_loss=0.3202, pruned_loss=0.08268, over 16644.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3344, pruned_loss=0.09153, over 3104647.77 frames. ], batch size: 57, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:40:26,885 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.583e+02 3.724e+02 4.498e+02 5.703e+02 1.501e+03, threshold=8.996e+02, percent-clipped=4.0 2023-04-28 08:41:26,196 INFO [train.py:904] (5/8) Epoch 5, batch 7100, loss[loss=0.245, simple_loss=0.3248, pruned_loss=0.08261, over 16698.00 frames. ], tot_loss[loss=0.258, simple_loss=0.3333, pruned_loss=0.09129, over 3093118.59 frames. ], batch size: 89, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:41:30,668 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3212, 2.2825, 1.6595, 1.6998, 2.8872, 2.4505, 3.2183, 3.1278], device='cuda:5'), covar=tensor([0.0025, 0.0180, 0.0250, 0.0240, 0.0093, 0.0179, 0.0067, 0.0084], device='cuda:5'), in_proj_covar=tensor([0.0075, 0.0152, 0.0155, 0.0151, 0.0146, 0.0156, 0.0127, 0.0134], device='cuda:5'), out_proj_covar=tensor([9.2441e-05, 1.8511e-04, 1.8481e-04, 1.8092e-04, 1.7951e-04, 1.9134e-04, 1.5032e-04, 1.6392e-04], device='cuda:5') 2023-04-28 08:42:20,350 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:24,181 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:44,686 INFO [train.py:904] (5/8) Epoch 5, batch 7150, loss[loss=0.2839, simple_loss=0.3358, pruned_loss=0.116, over 11402.00 frames. ], tot_loss[loss=0.2564, simple_loss=0.3312, pruned_loss=0.09081, over 3081620.75 frames. ], batch size: 248, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:43:03,971 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.186e+02 3.709e+02 4.695e+02 5.943e+02 1.245e+03, threshold=9.389e+02, percent-clipped=5.0 2023-04-28 08:43:31,012 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 08:43:39,963 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:43:57,722 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:44:00,018 INFO [train.py:904] (5/8) Epoch 5, batch 7200, loss[loss=0.2625, simple_loss=0.3228, pruned_loss=0.1011, over 11565.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3291, pruned_loss=0.08928, over 3064745.52 frames. ], batch size: 246, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:44:13,959 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:45:22,284 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:45:24,168 INFO [train.py:904] (5/8) Epoch 5, batch 7250, loss[loss=0.2175, simple_loss=0.3002, pruned_loss=0.06744, over 16805.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.326, pruned_loss=0.08759, over 3064628.90 frames. ], batch size: 102, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:45:43,494 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 3.368e+02 4.070e+02 5.805e+02 1.027e+03, threshold=8.141e+02, percent-clipped=1.0 2023-04-28 08:45:53,842 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:46:34,218 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:46:40,688 INFO [train.py:904] (5/8) Epoch 5, batch 7300, loss[loss=0.2354, simple_loss=0.3181, pruned_loss=0.07637, over 16738.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3251, pruned_loss=0.08729, over 3066311.16 frames. ], batch size: 83, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:47:13,414 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0092, 3.1788, 3.4731, 3.4580, 3.4350, 3.1409, 3.2903, 3.3271], device='cuda:5'), covar=tensor([0.0325, 0.0468, 0.0323, 0.0400, 0.0392, 0.0442, 0.0640, 0.0394], device='cuda:5'), in_proj_covar=tensor([0.0233, 0.0223, 0.0229, 0.0229, 0.0276, 0.0244, 0.0346, 0.0204], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 08:47:49,458 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:47:58,219 INFO [train.py:904] (5/8) Epoch 5, batch 7350, loss[loss=0.2334, simple_loss=0.3134, pruned_loss=0.07666, over 16657.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3255, pruned_loss=0.08809, over 3030021.64 frames. ], batch size: 124, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:48:17,670 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.639e+02 4.067e+02 5.121e+02 6.309e+02 1.579e+03, threshold=1.024e+03, percent-clipped=9.0 2023-04-28 08:48:20,169 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5660, 3.0459, 2.4067, 4.3984, 3.6303, 4.0605, 1.4215, 2.8732], device='cuda:5'), covar=tensor([0.1403, 0.0588, 0.1214, 0.0092, 0.0284, 0.0307, 0.1510, 0.0799], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0145, 0.0169, 0.0086, 0.0180, 0.0178, 0.0161, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 08:48:43,993 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-04-28 08:49:18,345 INFO [train.py:904] (5/8) Epoch 5, batch 7400, loss[loss=0.2226, simple_loss=0.3055, pruned_loss=0.06985, over 17200.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3266, pruned_loss=0.08953, over 3008687.12 frames. ], batch size: 45, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:49:22,202 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6981, 3.1661, 2.5126, 4.5632, 3.6254, 4.2015, 1.7259, 3.1141], device='cuda:5'), covar=tensor([0.1458, 0.0603, 0.1206, 0.0109, 0.0299, 0.0322, 0.1364, 0.0745], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0146, 0.0171, 0.0087, 0.0182, 0.0180, 0.0162, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 08:50:13,290 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:50:16,639 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1946, 3.6827, 3.8327, 1.5816, 4.0838, 4.1018, 2.8418, 2.9066], device='cuda:5'), covar=tensor([0.0910, 0.0131, 0.0151, 0.1290, 0.0046, 0.0060, 0.0350, 0.0411], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0087, 0.0081, 0.0141, 0.0072, 0.0076, 0.0116, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 08:50:38,176 INFO [train.py:904] (5/8) Epoch 5, batch 7450, loss[loss=0.3187, simple_loss=0.3606, pruned_loss=0.1384, over 11484.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3283, pruned_loss=0.09128, over 3018333.13 frames. ], batch size: 248, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:51:00,815 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.428e+02 4.070e+02 4.599e+02 6.099e+02 1.111e+03, threshold=9.198e+02, percent-clipped=2.0 2023-04-28 08:51:14,545 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:31,925 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:49,938 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:52:00,148 INFO [train.py:904] (5/8) Epoch 5, batch 7500, loss[loss=0.2289, simple_loss=0.3046, pruned_loss=0.07657, over 16894.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.329, pruned_loss=0.09026, over 3039873.64 frames. ], batch size: 116, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:52:36,635 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3891, 1.8846, 1.9992, 4.0152, 1.8918, 2.9374, 2.1329, 2.1489], device='cuda:5'), covar=tensor([0.0644, 0.2210, 0.1251, 0.0303, 0.2934, 0.0968, 0.1959, 0.2174], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0319, 0.0262, 0.0308, 0.0374, 0.0309, 0.0286, 0.0379], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:52:51,053 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:53:02,808 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-28 08:53:08,582 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:53:18,032 INFO [train.py:904] (5/8) Epoch 5, batch 7550, loss[loss=0.2381, simple_loss=0.3124, pruned_loss=0.08187, over 16886.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.328, pruned_loss=0.09038, over 3035676.84 frames. ], batch size: 109, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:53:38,522 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.261e+02 3.568e+02 4.415e+02 5.679e+02 1.356e+03, threshold=8.829e+02, percent-clipped=4.0 2023-04-28 08:53:40,657 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 08:54:00,815 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4999, 3.4780, 2.8628, 2.1829, 2.6114, 2.1858, 3.7901, 3.6242], device='cuda:5'), covar=tensor([0.2367, 0.0766, 0.1330, 0.1588, 0.1914, 0.1467, 0.0399, 0.0609], device='cuda:5'), in_proj_covar=tensor([0.0283, 0.0249, 0.0269, 0.0242, 0.0291, 0.0203, 0.0242, 0.0253], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:54:34,595 INFO [train.py:904] (5/8) Epoch 5, batch 7600, loss[loss=0.2405, simple_loss=0.3168, pruned_loss=0.08215, over 16735.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3269, pruned_loss=0.09002, over 3053314.77 frames. ], batch size: 124, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:54:52,535 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:55:51,252 INFO [train.py:904] (5/8) Epoch 5, batch 7650, loss[loss=0.2482, simple_loss=0.3265, pruned_loss=0.08494, over 16434.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.3279, pruned_loss=0.09093, over 3062439.43 frames. ], batch size: 146, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:56:12,528 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.451e+02 3.891e+02 4.972e+02 6.319e+02 1.654e+03, threshold=9.944e+02, percent-clipped=8.0 2023-04-28 08:56:25,957 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:56:53,155 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6336, 3.8344, 3.9823, 1.7538, 4.1728, 4.2379, 3.2165, 3.0788], device='cuda:5'), covar=tensor([0.0709, 0.0127, 0.0259, 0.1185, 0.0047, 0.0069, 0.0331, 0.0378], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0086, 0.0082, 0.0141, 0.0071, 0.0077, 0.0115, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 08:56:58,026 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3352, 1.9350, 1.5185, 1.6936, 2.1613, 2.0287, 2.2809, 2.3329], device='cuda:5'), covar=tensor([0.0036, 0.0155, 0.0225, 0.0204, 0.0100, 0.0153, 0.0083, 0.0100], device='cuda:5'), in_proj_covar=tensor([0.0073, 0.0148, 0.0152, 0.0148, 0.0144, 0.0152, 0.0124, 0.0133], device='cuda:5'), out_proj_covar=tensor([9.0301e-05, 1.7974e-04, 1.8182e-04, 1.7662e-04, 1.7687e-04, 1.8489e-04, 1.4766e-04, 1.6334e-04], device='cuda:5') 2023-04-28 08:57:08,653 INFO [train.py:904] (5/8) Epoch 5, batch 7700, loss[loss=0.244, simple_loss=0.3166, pruned_loss=0.08573, over 15459.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3284, pruned_loss=0.09155, over 3053007.40 frames. ], batch size: 191, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:57:44,752 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48324.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:58:26,513 INFO [train.py:904] (5/8) Epoch 5, batch 7750, loss[loss=0.2084, simple_loss=0.2965, pruned_loss=0.06017, over 17179.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.3284, pruned_loss=0.09033, over 3066376.81 frames. ], batch size: 46, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:58:47,830 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.847e+02 4.062e+02 4.544e+02 6.017e+02 9.038e+02, threshold=9.088e+02, percent-clipped=0.0 2023-04-28 08:58:55,508 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9363, 1.8682, 2.2117, 3.0957, 2.0432, 2.4724, 2.2345, 1.9357], device='cuda:5'), covar=tensor([0.0597, 0.2024, 0.0950, 0.0365, 0.2586, 0.1050, 0.1696, 0.2195], device='cuda:5'), in_proj_covar=tensor([0.0313, 0.0319, 0.0260, 0.0307, 0.0373, 0.0310, 0.0284, 0.0377], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:59:12,323 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5082, 3.4667, 2.7867, 2.2616, 2.4842, 2.2353, 3.5743, 3.6771], device='cuda:5'), covar=tensor([0.2289, 0.0703, 0.1313, 0.1694, 0.1796, 0.1473, 0.0533, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0248, 0.0269, 0.0244, 0.0290, 0.0202, 0.0242, 0.0254], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 08:59:20,116 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:34,741 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:39,598 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48397.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:45,054 INFO [train.py:904] (5/8) Epoch 5, batch 7800, loss[loss=0.2469, simple_loss=0.3227, pruned_loss=0.0856, over 16911.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3293, pruned_loss=0.09144, over 3066128.60 frames. ], batch size: 109, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:59:47,372 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0760, 2.5178, 2.3995, 3.2778, 2.6279, 3.2487, 1.8476, 2.7351], device='cuda:5'), covar=tensor([0.1050, 0.0419, 0.0912, 0.0094, 0.0220, 0.0360, 0.1133, 0.0596], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0143, 0.0169, 0.0086, 0.0182, 0.0178, 0.0159, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 09:00:21,168 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:28,784 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:49,709 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:52,212 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:01:02,301 INFO [train.py:904] (5/8) Epoch 5, batch 7850, loss[loss=0.2242, simple_loss=0.3037, pruned_loss=0.07231, over 16594.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3299, pruned_loss=0.09117, over 3066974.55 frames. ], batch size: 68, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:01:13,107 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:01:23,614 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.309e+02 3.753e+02 4.508e+02 5.540e+02 1.129e+03, threshold=9.017e+02, percent-clipped=6.0 2023-04-28 09:01:24,108 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:01:46,524 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7791, 3.9803, 3.2202, 2.5384, 3.1189, 2.4794, 4.3449, 4.1775], device='cuda:5'), covar=tensor([0.2179, 0.0644, 0.1213, 0.1503, 0.1914, 0.1302, 0.0344, 0.0591], device='cuda:5'), in_proj_covar=tensor([0.0280, 0.0245, 0.0268, 0.0242, 0.0287, 0.0200, 0.0240, 0.0251], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:01:52,615 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48484.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:02:04,082 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:02:16,631 INFO [train.py:904] (5/8) Epoch 5, batch 7900, loss[loss=0.2423, simple_loss=0.3248, pruned_loss=0.07997, over 16771.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3281, pruned_loss=0.08979, over 3073876.05 frames. ], batch size: 89, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:02:35,283 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:03:36,606 INFO [train.py:904] (5/8) Epoch 5, batch 7950, loss[loss=0.2156, simple_loss=0.2983, pruned_loss=0.06645, over 16764.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3279, pruned_loss=0.08968, over 3072205.69 frames. ], batch size: 83, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:03:56,965 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 3.717e+02 4.429e+02 5.526e+02 1.268e+03, threshold=8.857e+02, percent-clipped=1.0 2023-04-28 09:04:03,068 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 09:04:50,733 INFO [train.py:904] (5/8) Epoch 5, batch 8000, loss[loss=0.283, simple_loss=0.3515, pruned_loss=0.1073, over 17101.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3288, pruned_loss=0.09109, over 3069234.22 frames. ], batch size: 49, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:05:27,049 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 09:06:06,087 INFO [train.py:904] (5/8) Epoch 5, batch 8050, loss[loss=0.2434, simple_loss=0.3227, pruned_loss=0.08208, over 15299.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3273, pruned_loss=0.08911, over 3094637.57 frames. ], batch size: 190, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:06:26,979 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.628e+02 3.970e+02 4.648e+02 5.523e+02 1.617e+03, threshold=9.296e+02, percent-clipped=6.0 2023-04-28 09:06:33,482 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4544, 3.6535, 2.6102, 2.1296, 2.7950, 2.0848, 3.8798, 3.8729], device='cuda:5'), covar=tensor([0.2633, 0.0783, 0.1673, 0.1784, 0.2155, 0.1604, 0.0494, 0.0695], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0245, 0.0270, 0.0243, 0.0291, 0.0201, 0.0241, 0.0253], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:06:49,658 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:07:21,590 INFO [train.py:904] (5/8) Epoch 5, batch 8100, loss[loss=0.2387, simple_loss=0.3174, pruned_loss=0.08, over 16520.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3262, pruned_loss=0.08791, over 3118838.93 frames. ], batch size: 75, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:06,110 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:08:40,610 INFO [train.py:904] (5/8) Epoch 5, batch 8150, loss[loss=0.2779, simple_loss=0.3295, pruned_loss=0.1132, over 11963.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3243, pruned_loss=0.08711, over 3120866.75 frames. ], batch size: 247, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:44,642 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:08:57,270 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 09:09:01,948 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 3.868e+02 4.692e+02 5.916e+02 1.218e+03, threshold=9.385e+02, percent-clipped=3.0 2023-04-28 09:09:20,847 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:25,129 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:10:00,540 INFO [train.py:904] (5/8) Epoch 5, batch 8200, loss[loss=0.2966, simple_loss=0.3456, pruned_loss=0.1238, over 11663.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3218, pruned_loss=0.08625, over 3118387.79 frames. ], batch size: 247, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:10:32,489 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:11:00,970 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:11:22,106 INFO [train.py:904] (5/8) Epoch 5, batch 8250, loss[loss=0.2227, simple_loss=0.3113, pruned_loss=0.06702, over 16856.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3213, pruned_loss=0.08469, over 3094188.55 frames. ], batch size: 102, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:11:44,536 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.248e+02 3.760e+02 4.474e+02 5.646e+02 1.182e+03, threshold=8.947e+02, percent-clipped=2.0 2023-04-28 09:11:50,788 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:12:13,878 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:41,201 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:43,318 INFO [train.py:904] (5/8) Epoch 5, batch 8300, loss[loss=0.1996, simple_loss=0.2945, pruned_loss=0.05232, over 16303.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.318, pruned_loss=0.08135, over 3080253.80 frames. ], batch size: 165, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:13:08,660 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:13:11,048 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 09:13:30,745 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7347, 4.9990, 4.7516, 4.7888, 4.3988, 4.4116, 4.5410, 4.9884], device='cuda:5'), covar=tensor([0.0584, 0.0680, 0.0815, 0.0444, 0.0565, 0.0827, 0.0578, 0.0712], device='cuda:5'), in_proj_covar=tensor([0.0356, 0.0465, 0.0402, 0.0301, 0.0290, 0.0315, 0.0377, 0.0332], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:14:06,505 INFO [train.py:904] (5/8) Epoch 5, batch 8350, loss[loss=0.2199, simple_loss=0.311, pruned_loss=0.06434, over 16374.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3169, pruned_loss=0.07897, over 3080754.32 frames. ], batch size: 146, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:14:30,520 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.953e+02 3.108e+02 3.875e+02 4.541e+02 1.583e+03, threshold=7.750e+02, percent-clipped=2.0 2023-04-28 09:14:45,099 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6403, 4.8162, 4.8039, 4.9020, 4.8116, 5.3627, 4.9754, 4.7308], device='cuda:5'), covar=tensor([0.0754, 0.1317, 0.1197, 0.1347, 0.2263, 0.0776, 0.0955, 0.2078], device='cuda:5'), in_proj_covar=tensor([0.0262, 0.0367, 0.0367, 0.0319, 0.0429, 0.0391, 0.0301, 0.0436], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 09:14:54,758 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:15:07,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9427, 1.6305, 1.3969, 1.3291, 1.7795, 1.6108, 1.8910, 1.8869], device='cuda:5'), covar=tensor([0.0040, 0.0161, 0.0240, 0.0199, 0.0106, 0.0176, 0.0096, 0.0116], device='cuda:5'), in_proj_covar=tensor([0.0074, 0.0149, 0.0151, 0.0146, 0.0145, 0.0152, 0.0122, 0.0131], device='cuda:5'), out_proj_covar=tensor([8.9779e-05, 1.8081e-04, 1.7892e-04, 1.7337e-04, 1.7794e-04, 1.8382e-04, 1.4413e-04, 1.5934e-04], device='cuda:5') 2023-04-28 09:15:30,227 INFO [train.py:904] (5/8) Epoch 5, batch 8400, loss[loss=0.2115, simple_loss=0.2956, pruned_loss=0.06367, over 15323.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3129, pruned_loss=0.07593, over 3070700.11 frames. ], batch size: 190, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:15:40,671 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1490, 4.2193, 4.2235, 4.2873, 4.1971, 4.7381, 4.4076, 4.1567], device='cuda:5'), covar=tensor([0.1349, 0.1543, 0.1546, 0.1922, 0.2607, 0.1031, 0.1376, 0.2583], device='cuda:5'), in_proj_covar=tensor([0.0262, 0.0367, 0.0368, 0.0319, 0.0425, 0.0389, 0.0302, 0.0433], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 09:15:54,263 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3433, 4.3420, 4.2307, 3.6336, 4.2202, 1.5378, 3.9604, 4.0889], device='cuda:5'), covar=tensor([0.0057, 0.0045, 0.0089, 0.0247, 0.0057, 0.1886, 0.0091, 0.0128], device='cuda:5'), in_proj_covar=tensor([0.0090, 0.0077, 0.0121, 0.0121, 0.0089, 0.0143, 0.0105, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:16:11,904 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7397, 3.6690, 3.8457, 3.9963, 4.0298, 3.5740, 4.0011, 4.0303], device='cuda:5'), covar=tensor([0.0815, 0.0642, 0.0954, 0.0477, 0.0403, 0.1330, 0.0494, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0358, 0.0450, 0.0559, 0.0469, 0.0351, 0.0341, 0.0368, 0.0392], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:16:13,367 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 09:16:14,934 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:16:51,012 INFO [train.py:904] (5/8) Epoch 5, batch 8450, loss[loss=0.2079, simple_loss=0.2964, pruned_loss=0.05969, over 16767.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3113, pruned_loss=0.07414, over 3077364.39 frames. ], batch size: 124, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:16:55,263 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:17:13,777 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.140e+02 2.885e+02 3.477e+02 4.145e+02 8.809e+02, threshold=6.955e+02, percent-clipped=1.0 2023-04-28 09:17:38,511 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:07,476 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-04-28 09:18:13,189 INFO [train.py:904] (5/8) Epoch 5, batch 8500, loss[loss=0.1982, simple_loss=0.2716, pruned_loss=0.06238, over 11742.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3067, pruned_loss=0.07092, over 3077883.53 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:18:13,698 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:27,518 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2608, 1.8756, 2.2040, 3.6565, 1.8293, 2.6030, 2.2159, 1.9812], device='cuda:5'), covar=tensor([0.0537, 0.2267, 0.1102, 0.0313, 0.3213, 0.1220, 0.2027, 0.2568], device='cuda:5'), in_proj_covar=tensor([0.0304, 0.0313, 0.0258, 0.0298, 0.0368, 0.0306, 0.0284, 0.0367], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:18:50,732 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6877, 2.7169, 1.8037, 2.7359, 2.0992, 2.7296, 2.0189, 2.4328], device='cuda:5'), covar=tensor([0.0161, 0.0256, 0.1137, 0.0069, 0.0597, 0.0432, 0.1143, 0.0547], device='cuda:5'), in_proj_covar=tensor([0.0115, 0.0145, 0.0174, 0.0075, 0.0156, 0.0172, 0.0182, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 09:18:57,910 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49127.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:19:38,542 INFO [train.py:904] (5/8) Epoch 5, batch 8550, loss[loss=0.2417, simple_loss=0.3283, pruned_loss=0.07759, over 15208.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3037, pruned_loss=0.06941, over 3056706.14 frames. ], batch size: 191, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:20:04,110 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 3.233e+02 3.773e+02 4.691e+02 1.040e+03, threshold=7.547e+02, percent-clipped=6.0 2023-04-28 09:20:10,436 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 09:20:24,696 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 09:20:29,025 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:21:04,548 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:21:17,325 INFO [train.py:904] (5/8) Epoch 5, batch 8600, loss[loss=0.2309, simple_loss=0.3158, pruned_loss=0.07303, over 15401.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3044, pruned_loss=0.06879, over 3036780.56 frames. ], batch size: 191, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:22:28,845 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:22:58,659 INFO [train.py:904] (5/8) Epoch 5, batch 8650, loss[loss=0.1991, simple_loss=0.2928, pruned_loss=0.05271, over 16648.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3017, pruned_loss=0.06656, over 3031273.99 frames. ], batch size: 89, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:23:33,946 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.003e+02 3.105e+02 3.609e+02 4.531e+02 9.141e+02, threshold=7.218e+02, percent-clipped=3.0 2023-04-28 09:24:39,651 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:24:47,026 INFO [train.py:904] (5/8) Epoch 5, batch 8700, loss[loss=0.2048, simple_loss=0.2916, pruned_loss=0.05905, over 16662.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2984, pruned_loss=0.06445, over 3062094.98 frames. ], batch size: 76, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:25:15,021 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7290, 4.7577, 4.5616, 4.4126, 4.1853, 4.6064, 4.4658, 4.3411], device='cuda:5'), covar=tensor([0.0351, 0.0245, 0.0168, 0.0142, 0.0644, 0.0268, 0.0237, 0.0382], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0177, 0.0202, 0.0172, 0.0224, 0.0200, 0.0143, 0.0226], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:25:42,210 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-28 09:26:24,159 INFO [train.py:904] (5/8) Epoch 5, batch 8750, loss[loss=0.2038, simple_loss=0.2808, pruned_loss=0.06345, over 12369.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2974, pruned_loss=0.06353, over 3054749.37 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:26:52,045 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 09:27:05,490 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.935e+02 3.673e+02 5.022e+02 7.467e+02, threshold=7.345e+02, percent-clipped=1.0 2023-04-28 09:28:17,612 INFO [train.py:904] (5/8) Epoch 5, batch 8800, loss[loss=0.2051, simple_loss=0.2948, pruned_loss=0.05773, over 16815.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2961, pruned_loss=0.0626, over 3053674.74 frames. ], batch size: 124, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:28:35,909 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 09:29:02,450 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:29:24,397 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:29:35,247 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0662, 4.0968, 4.1976, 4.2358, 4.1488, 4.6846, 4.4881, 4.1550], device='cuda:5'), covar=tensor([0.1334, 0.1545, 0.1268, 0.1511, 0.2483, 0.0984, 0.0986, 0.2087], device='cuda:5'), in_proj_covar=tensor([0.0257, 0.0358, 0.0358, 0.0310, 0.0414, 0.0388, 0.0298, 0.0425], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:30:04,521 INFO [train.py:904] (5/8) Epoch 5, batch 8850, loss[loss=0.1909, simple_loss=0.2737, pruned_loss=0.05403, over 12364.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2989, pruned_loss=0.0622, over 3047544.83 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:30:09,181 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4436, 5.3289, 5.0685, 4.7189, 5.2483, 1.7620, 5.0244, 5.1436], device='cuda:5'), covar=tensor([0.0031, 0.0040, 0.0077, 0.0165, 0.0037, 0.1623, 0.0057, 0.0081], device='cuda:5'), in_proj_covar=tensor([0.0090, 0.0076, 0.0121, 0.0115, 0.0089, 0.0142, 0.0104, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:30:27,463 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 09:30:38,667 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 3.290e+02 4.056e+02 5.025e+02 1.173e+03, threshold=8.111e+02, percent-clipped=7.0 2023-04-28 09:31:02,593 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:16,932 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:17,238 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 09:31:22,671 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-28 09:31:37,589 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:31:39,930 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:54,252 INFO [train.py:904] (5/8) Epoch 5, batch 8900, loss[loss=0.202, simple_loss=0.2855, pruned_loss=0.0593, over 13010.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2994, pruned_loss=0.06121, over 3065583.08 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:32:48,444 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:39,648 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:59,820 INFO [train.py:904] (5/8) Epoch 5, batch 8950, loss[loss=0.1848, simple_loss=0.2827, pruned_loss=0.04344, over 16872.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2987, pruned_loss=0.06141, over 3075435.22 frames. ], batch size: 102, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:34:03,741 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2872, 3.5486, 3.6439, 1.7887, 3.8096, 3.8597, 3.1570, 2.9688], device='cuda:5'), covar=tensor([0.0775, 0.0122, 0.0138, 0.1205, 0.0047, 0.0051, 0.0284, 0.0391], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0088, 0.0078, 0.0142, 0.0067, 0.0075, 0.0115, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 09:34:35,632 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 2.796e+02 3.504e+02 4.655e+02 7.441e+02, threshold=7.007e+02, percent-clipped=0.0 2023-04-28 09:35:12,818 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1823, 4.9954, 5.6058, 5.5300, 5.5362, 5.1962, 5.1977, 4.7998], device='cuda:5'), covar=tensor([0.0164, 0.0315, 0.0220, 0.0315, 0.0292, 0.0181, 0.0642, 0.0269], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0216, 0.0221, 0.0220, 0.0270, 0.0237, 0.0327, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 09:35:33,165 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:35:44,024 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5311, 3.6878, 2.8061, 2.1978, 2.4706, 2.2019, 3.8928, 3.7079], device='cuda:5'), covar=tensor([0.2189, 0.0526, 0.1278, 0.1548, 0.1806, 0.1431, 0.0325, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0273, 0.0237, 0.0260, 0.0237, 0.0247, 0.0196, 0.0234, 0.0234], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:35:51,815 INFO [train.py:904] (5/8) Epoch 5, batch 9000, loss[loss=0.192, simple_loss=0.2785, pruned_loss=0.05278, over 16416.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2946, pruned_loss=0.05955, over 3095836.03 frames. ], batch size: 146, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:35:51,815 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 09:36:02,104 INFO [train.py:938] (5/8) Epoch 5, validation: loss=0.1735, simple_loss=0.2766, pruned_loss=0.0352, over 944034.00 frames. 2023-04-28 09:36:02,105 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 09:36:24,214 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9872, 1.9755, 2.2507, 3.1989, 1.9647, 2.5090, 2.2747, 1.9306], device='cuda:5'), covar=tensor([0.0532, 0.1993, 0.1014, 0.0364, 0.2863, 0.1170, 0.1768, 0.2535], device='cuda:5'), in_proj_covar=tensor([0.0309, 0.0309, 0.0260, 0.0299, 0.0366, 0.0306, 0.0286, 0.0365], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:36:56,726 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9752, 3.1234, 3.1539, 1.6400, 3.3477, 3.3455, 2.9074, 2.6688], device='cuda:5'), covar=tensor([0.0806, 0.0125, 0.0144, 0.1174, 0.0059, 0.0077, 0.0294, 0.0394], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0088, 0.0078, 0.0142, 0.0068, 0.0076, 0.0115, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 09:37:44,740 INFO [train.py:904] (5/8) Epoch 5, batch 9050, loss[loss=0.2141, simple_loss=0.2904, pruned_loss=0.06895, over 15366.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.296, pruned_loss=0.06025, over 3101248.89 frames. ], batch size: 192, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:38:11,778 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 09:38:18,684 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 3.141e+02 3.840e+02 5.023e+02 8.628e+02, threshold=7.679e+02, percent-clipped=5.0 2023-04-28 09:38:19,829 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-28 09:39:06,974 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3425, 3.8365, 3.8967, 1.8862, 4.1061, 4.1545, 3.1217, 3.1343], device='cuda:5'), covar=tensor([0.0750, 0.0115, 0.0145, 0.1123, 0.0040, 0.0055, 0.0295, 0.0380], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0086, 0.0076, 0.0139, 0.0067, 0.0075, 0.0114, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 09:39:29,267 INFO [train.py:904] (5/8) Epoch 5, batch 9100, loss[loss=0.2143, simple_loss=0.3011, pruned_loss=0.06373, over 16927.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2956, pruned_loss=0.06076, over 3111209.49 frames. ], batch size: 109, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:39:49,524 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2249, 3.2787, 3.5970, 3.5505, 3.5786, 3.3084, 3.3737, 3.4255], device='cuda:5'), covar=tensor([0.0270, 0.0450, 0.0358, 0.0434, 0.0371, 0.0371, 0.0660, 0.0367], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0218, 0.0223, 0.0222, 0.0273, 0.0239, 0.0329, 0.0201], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 09:41:26,346 INFO [train.py:904] (5/8) Epoch 5, batch 9150, loss[loss=0.2058, simple_loss=0.2818, pruned_loss=0.06494, over 12100.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.295, pruned_loss=0.06031, over 3075309.89 frames. ], batch size: 250, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:42:00,604 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:42:01,430 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.296e+02 3.040e+02 3.844e+02 4.947e+02 8.229e+02, threshold=7.688e+02, percent-clipped=4.0 2023-04-28 09:42:25,698 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:42:44,664 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5223, 4.4381, 4.3353, 3.8740, 4.3197, 1.7066, 4.0670, 4.1914], device='cuda:5'), covar=tensor([0.0055, 0.0050, 0.0089, 0.0190, 0.0060, 0.1599, 0.0080, 0.0096], device='cuda:5'), in_proj_covar=tensor([0.0090, 0.0076, 0.0120, 0.0113, 0.0089, 0.0142, 0.0102, 0.0112], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:42:47,442 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 09:42:55,729 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2778, 3.8238, 3.6574, 1.9359, 2.9115, 2.4614, 3.4735, 3.5372], device='cuda:5'), covar=tensor([0.0236, 0.0441, 0.0439, 0.1640, 0.0707, 0.0916, 0.0680, 0.0779], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0116, 0.0148, 0.0138, 0.0129, 0.0126, 0.0134, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 09:42:57,377 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-28 09:42:59,605 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6562, 5.1893, 5.2343, 5.2247, 5.1962, 5.7149, 5.3267, 5.0685], device='cuda:5'), covar=tensor([0.0699, 0.1196, 0.1124, 0.1404, 0.1930, 0.0789, 0.1028, 0.1647], device='cuda:5'), in_proj_covar=tensor([0.0255, 0.0360, 0.0356, 0.0315, 0.0421, 0.0384, 0.0298, 0.0427], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:43:09,761 INFO [train.py:904] (5/8) Epoch 5, batch 9200, loss[loss=0.221, simple_loss=0.3023, pruned_loss=0.06982, over 16745.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2896, pruned_loss=0.0585, over 3090971.23 frames. ], batch size: 134, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:43:22,990 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1586, 4.0539, 4.5014, 4.4933, 4.5235, 4.1596, 4.2586, 4.0538], device='cuda:5'), covar=tensor([0.0212, 0.0436, 0.0331, 0.0365, 0.0321, 0.0305, 0.0635, 0.0353], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0211, 0.0217, 0.0214, 0.0262, 0.0232, 0.0315, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 09:43:58,035 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:44:46,695 INFO [train.py:904] (5/8) Epoch 5, batch 9250, loss[loss=0.1732, simple_loss=0.2501, pruned_loss=0.04809, over 12384.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2896, pruned_loss=0.0585, over 3091571.18 frames. ], batch size: 247, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:45:18,930 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.407e+02 3.305e+02 4.013e+02 4.841e+02 9.780e+02, threshold=8.027e+02, percent-clipped=1.0 2023-04-28 09:46:21,477 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:46:40,058 INFO [train.py:904] (5/8) Epoch 5, batch 9300, loss[loss=0.183, simple_loss=0.2683, pruned_loss=0.04884, over 16573.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2877, pruned_loss=0.05758, over 3073137.73 frames. ], batch size: 68, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:47:58,081 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 09:48:05,872 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:48:26,423 INFO [train.py:904] (5/8) Epoch 5, batch 9350, loss[loss=0.2092, simple_loss=0.2937, pruned_loss=0.06233, over 16806.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2876, pruned_loss=0.05753, over 3071891.13 frames. ], batch size: 124, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:49:00,542 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 3.133e+02 3.801e+02 4.449e+02 7.989e+02, threshold=7.602e+02, percent-clipped=0.0 2023-04-28 09:50:11,387 INFO [train.py:904] (5/8) Epoch 5, batch 9400, loss[loss=0.2218, simple_loss=0.3148, pruned_loss=0.06442, over 16335.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2876, pruned_loss=0.0574, over 3057222.36 frames. ], batch size: 146, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:51:51,665 INFO [train.py:904] (5/8) Epoch 5, batch 9450, loss[loss=0.2098, simple_loss=0.2988, pruned_loss=0.06041, over 16352.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2902, pruned_loss=0.05805, over 3057889.57 frames. ], batch size: 146, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:52:21,843 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.955e+02 3.683e+02 5.023e+02 1.227e+03, threshold=7.366e+02, percent-clipped=5.0 2023-04-28 09:52:43,818 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 09:52:47,323 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:53:09,068 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:53:33,294 INFO [train.py:904] (5/8) Epoch 5, batch 9500, loss[loss=0.1706, simple_loss=0.2638, pruned_loss=0.03871, over 16889.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2893, pruned_loss=0.05703, over 3080941.69 frames. ], batch size: 96, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:54:15,695 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4378, 4.5996, 4.6365, 4.5483, 4.5854, 5.1200, 4.7828, 4.4645], device='cuda:5'), covar=tensor([0.0963, 0.1707, 0.1463, 0.1755, 0.2291, 0.0892, 0.1190, 0.2201], device='cuda:5'), in_proj_covar=tensor([0.0247, 0.0353, 0.0350, 0.0306, 0.0407, 0.0380, 0.0293, 0.0413], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:54:18,293 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:25,653 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:29,340 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:44,419 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:55:18,507 INFO [train.py:904] (5/8) Epoch 5, batch 9550, loss[loss=0.209, simple_loss=0.2981, pruned_loss=0.05995, over 16642.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2894, pruned_loss=0.05761, over 3080989.12 frames. ], batch size: 76, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:55:53,350 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.772e+02 3.459e+02 4.281e+02 6.687e+02, threshold=6.919e+02, percent-clipped=0.0 2023-04-28 09:56:38,106 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:56:52,903 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:56:57,561 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4524, 4.6409, 4.6985, 4.6693, 4.6630, 5.1592, 4.8003, 4.5884], device='cuda:5'), covar=tensor([0.0912, 0.1480, 0.1230, 0.1511, 0.1919, 0.0853, 0.1040, 0.2027], device='cuda:5'), in_proj_covar=tensor([0.0248, 0.0354, 0.0347, 0.0305, 0.0401, 0.0380, 0.0290, 0.0408], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:56:59,049 INFO [train.py:904] (5/8) Epoch 5, batch 9600, loss[loss=0.2192, simple_loss=0.3156, pruned_loss=0.06137, over 16694.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2922, pruned_loss=0.05947, over 3066514.93 frames. ], batch size: 134, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:57:46,527 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7305, 3.5943, 3.8306, 3.9517, 4.0310, 3.5222, 4.0367, 4.0339], device='cuda:5'), covar=tensor([0.0886, 0.0680, 0.0893, 0.0502, 0.0386, 0.1450, 0.0386, 0.0388], device='cuda:5'), in_proj_covar=tensor([0.0345, 0.0429, 0.0534, 0.0444, 0.0330, 0.0330, 0.0349, 0.0366], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 09:58:08,813 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 09:58:47,437 INFO [train.py:904] (5/8) Epoch 5, batch 9650, loss[loss=0.2166, simple_loss=0.3024, pruned_loss=0.06537, over 15161.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2942, pruned_loss=0.05959, over 3064855.47 frames. ], batch size: 190, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:59:05,783 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:59:27,464 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 2.991e+02 3.657e+02 4.626e+02 9.582e+02, threshold=7.315e+02, percent-clipped=7.0 2023-04-28 10:00:35,793 INFO [train.py:904] (5/8) Epoch 5, batch 9700, loss[loss=0.1948, simple_loss=0.2735, pruned_loss=0.05803, over 12272.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2925, pruned_loss=0.05873, over 3078150.85 frames. ], batch size: 247, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:01:56,250 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2991, 1.8333, 1.5291, 1.5653, 2.2112, 1.9473, 2.2182, 2.2872], device='cuda:5'), covar=tensor([0.0029, 0.0210, 0.0249, 0.0261, 0.0122, 0.0183, 0.0082, 0.0111], device='cuda:5'), in_proj_covar=tensor([0.0071, 0.0153, 0.0153, 0.0149, 0.0147, 0.0151, 0.0119, 0.0131], device='cuda:5'), out_proj_covar=tensor([8.4745e-05, 1.8469e-04, 1.8036e-04, 1.7628e-04, 1.7897e-04, 1.8185e-04, 1.3565e-04, 1.5762e-04], device='cuda:5') 2023-04-28 10:02:18,559 INFO [train.py:904] (5/8) Epoch 5, batch 9750, loss[loss=0.2063, simple_loss=0.2845, pruned_loss=0.06405, over 12521.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2907, pruned_loss=0.05884, over 3071106.37 frames. ], batch size: 248, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:02:50,596 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 3.120e+02 3.818e+02 4.621e+02 8.897e+02, threshold=7.636e+02, percent-clipped=1.0 2023-04-28 10:03:32,229 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 10:03:56,541 INFO [train.py:904] (5/8) Epoch 5, batch 9800, loss[loss=0.1896, simple_loss=0.2894, pruned_loss=0.0449, over 16628.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2898, pruned_loss=0.05719, over 3076726.25 frames. ], batch size: 62, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:04:37,008 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:05:26,633 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 10:05:41,983 INFO [train.py:904] (5/8) Epoch 5, batch 9850, loss[loss=0.2036, simple_loss=0.2907, pruned_loss=0.05822, over 16755.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2908, pruned_loss=0.05702, over 3067007.25 frames. ], batch size: 134, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:06:14,680 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.952e+02 3.639e+02 4.340e+02 9.232e+02, threshold=7.278e+02, percent-clipped=1.0 2023-04-28 10:06:21,889 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:06:31,901 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 10:06:52,950 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:07:31,486 INFO [train.py:904] (5/8) Epoch 5, batch 9900, loss[loss=0.2101, simple_loss=0.3021, pruned_loss=0.05906, over 16157.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2912, pruned_loss=0.05764, over 3037121.64 frames. ], batch size: 165, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:08:18,783 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5421, 3.3580, 2.8936, 1.6969, 2.6505, 2.0473, 2.9451, 3.0252], device='cuda:5'), covar=tensor([0.0297, 0.0474, 0.0541, 0.1616, 0.0715, 0.0973, 0.0679, 0.0732], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0115, 0.0151, 0.0139, 0.0129, 0.0127, 0.0134, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 10:09:17,063 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:09:27,684 INFO [train.py:904] (5/8) Epoch 5, batch 9950, loss[loss=0.2166, simple_loss=0.3077, pruned_loss=0.06278, over 16832.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2937, pruned_loss=0.05862, over 3029723.46 frames. ], batch size: 124, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:09:33,885 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:10:04,530 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 3.001e+02 3.652e+02 4.906e+02 2.314e+03, threshold=7.303e+02, percent-clipped=6.0 2023-04-28 10:11:29,341 INFO [train.py:904] (5/8) Epoch 5, batch 10000, loss[loss=0.2053, simple_loss=0.28, pruned_loss=0.06527, over 12157.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2921, pruned_loss=0.05803, over 3048617.82 frames. ], batch size: 246, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:11:44,477 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:13:08,532 INFO [train.py:904] (5/8) Epoch 5, batch 10050, loss[loss=0.2247, simple_loss=0.3082, pruned_loss=0.07064, over 16901.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.292, pruned_loss=0.05786, over 3047012.87 frames. ], batch size: 109, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:13:38,901 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.547e+02 3.217e+02 3.917e+02 1.118e+03, threshold=6.434e+02, percent-clipped=1.0 2023-04-28 10:14:10,143 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0324, 3.9671, 3.8301, 3.4511, 3.8380, 1.5580, 3.6947, 3.6130], device='cuda:5'), covar=tensor([0.0065, 0.0060, 0.0103, 0.0205, 0.0070, 0.1800, 0.0103, 0.0150], device='cuda:5'), in_proj_covar=tensor([0.0088, 0.0076, 0.0119, 0.0109, 0.0088, 0.0143, 0.0101, 0.0111], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:14:28,313 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5733, 3.6316, 2.7698, 2.0287, 2.2834, 2.0253, 3.7618, 3.6103], device='cuda:5'), covar=tensor([0.1970, 0.0550, 0.1180, 0.1618, 0.2156, 0.1598, 0.0336, 0.0520], device='cuda:5'), in_proj_covar=tensor([0.0266, 0.0239, 0.0259, 0.0235, 0.0231, 0.0194, 0.0231, 0.0227], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:14:39,010 INFO [train.py:904] (5/8) Epoch 5, batch 10100, loss[loss=0.2006, simple_loss=0.2855, pruned_loss=0.05781, over 15390.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2928, pruned_loss=0.05804, over 3056337.39 frames. ], batch size: 190, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:14:48,115 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0299, 2.8000, 2.7283, 1.7957, 2.8682, 2.9124, 2.4607, 2.4679], device='cuda:5'), covar=tensor([0.0650, 0.0129, 0.0128, 0.0980, 0.0075, 0.0091, 0.0335, 0.0376], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0089, 0.0077, 0.0141, 0.0067, 0.0075, 0.0114, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 10:14:55,781 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 10:15:30,516 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9433, 2.7494, 2.6754, 1.9693, 2.5701, 2.5991, 2.5754, 1.8601], device='cuda:5'), covar=tensor([0.0260, 0.0028, 0.0040, 0.0207, 0.0063, 0.0058, 0.0049, 0.0264], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0054, 0.0059, 0.0115, 0.0060, 0.0066, 0.0063, 0.0108], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 10:16:20,279 INFO [train.py:904] (5/8) Epoch 6, batch 0, loss[loss=0.2279, simple_loss=0.2999, pruned_loss=0.07797, over 17097.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.2999, pruned_loss=0.07797, over 17097.00 frames. ], batch size: 47, lr: 1.19e-02, grad_scale: 8.0 2023-04-28 10:16:20,279 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 10:16:27,648 INFO [train.py:938] (5/8) Epoch 6, validation: loss=0.1727, simple_loss=0.2755, pruned_loss=0.03501, over 944034.00 frames. 2023-04-28 10:16:27,649 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 10:16:52,399 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 3.546e+02 4.478e+02 5.747e+02 1.222e+03, threshold=8.956e+02, percent-clipped=19.0 2023-04-28 10:17:10,741 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:13,792 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:31,964 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-28 10:17:35,104 INFO [train.py:904] (5/8) Epoch 6, batch 50, loss[loss=0.2739, simple_loss=0.3394, pruned_loss=0.1042, over 16880.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3165, pruned_loss=0.09065, over 750831.92 frames. ], batch size: 116, lr: 1.19e-02, grad_scale: 2.0 2023-04-28 10:17:39,118 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5358, 6.0024, 5.7275, 5.7640, 5.1228, 4.8946, 5.4233, 6.1063], device='cuda:5'), covar=tensor([0.0774, 0.0751, 0.1038, 0.0515, 0.0812, 0.0716, 0.0767, 0.0712], device='cuda:5'), in_proj_covar=tensor([0.0363, 0.0482, 0.0404, 0.0311, 0.0307, 0.0313, 0.0389, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:17:46,547 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:18:02,698 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-28 10:18:20,195 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:34,298 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:47,063 INFO [train.py:904] (5/8) Epoch 6, batch 100, loss[loss=0.2484, simple_loss=0.3164, pruned_loss=0.09014, over 16705.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3071, pruned_loss=0.08197, over 1323425.08 frames. ], batch size: 89, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:18:49,625 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:19:11,790 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.233e+02 3.517e+02 4.205e+02 5.493e+02 1.012e+03, threshold=8.410e+02, percent-clipped=3.0 2023-04-28 10:19:12,321 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:19:24,241 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 10:19:25,080 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7549, 2.8657, 2.2847, 4.1235, 3.6257, 3.9477, 1.7533, 2.9363], device='cuda:5'), covar=tensor([0.1242, 0.0507, 0.1131, 0.0080, 0.0241, 0.0346, 0.1142, 0.0716], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0139, 0.0164, 0.0086, 0.0155, 0.0174, 0.0158, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 10:19:54,858 INFO [train.py:904] (5/8) Epoch 6, batch 150, loss[loss=0.2356, simple_loss=0.298, pruned_loss=0.08658, over 16433.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3051, pruned_loss=0.0806, over 1760675.47 frames. ], batch size: 146, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:19:55,165 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:19:56,161 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:20:20,355 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3871, 3.9873, 3.2009, 5.2658, 4.8012, 4.8864, 2.3097, 3.9638], device='cuda:5'), covar=tensor([0.1218, 0.0434, 0.1005, 0.0114, 0.0312, 0.0275, 0.1224, 0.0535], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0140, 0.0166, 0.0088, 0.0160, 0.0177, 0.0161, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 10:20:24,611 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-28 10:20:57,660 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:21:05,284 INFO [train.py:904] (5/8) Epoch 6, batch 200, loss[loss=0.2017, simple_loss=0.2963, pruned_loss=0.05353, over 17013.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3041, pruned_loss=0.08019, over 2101959.26 frames. ], batch size: 55, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:21:11,827 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-28 10:21:28,631 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 3.434e+02 3.943e+02 4.796e+02 1.132e+03, threshold=7.886e+02, percent-clipped=3.0 2023-04-28 10:22:12,667 INFO [train.py:904] (5/8) Epoch 6, batch 250, loss[loss=0.2073, simple_loss=0.2976, pruned_loss=0.05852, over 17249.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3009, pruned_loss=0.07855, over 2379813.60 frames. ], batch size: 52, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:22:20,797 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:22:24,918 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 10:22:35,874 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0402, 1.7344, 2.2950, 2.8923, 2.6855, 3.1320, 1.7025, 3.1734], device='cuda:5'), covar=tensor([0.0083, 0.0233, 0.0141, 0.0124, 0.0118, 0.0109, 0.0247, 0.0053], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0146, 0.0131, 0.0127, 0.0133, 0.0095, 0.0144, 0.0080], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 10:22:37,426 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 10:23:00,037 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 10:23:20,600 INFO [train.py:904] (5/8) Epoch 6, batch 300, loss[loss=0.1787, simple_loss=0.2679, pruned_loss=0.0447, over 17211.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2967, pruned_loss=0.07437, over 2593215.58 frames. ], batch size: 45, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:23:37,405 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:23:45,702 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.797e+02 3.691e+02 4.478e+02 8.058e+02, threshold=7.381e+02, percent-clipped=1.0 2023-04-28 10:24:30,416 INFO [train.py:904] (5/8) Epoch 6, batch 350, loss[loss=0.2198, simple_loss=0.2914, pruned_loss=0.07412, over 15465.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2933, pruned_loss=0.07202, over 2750227.88 frames. ], batch size: 190, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:25:01,873 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:03,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9443, 3.0051, 3.3100, 2.4399, 3.2268, 3.4326, 3.4094, 1.9475], device='cuda:5'), covar=tensor([0.0332, 0.0085, 0.0037, 0.0191, 0.0045, 0.0047, 0.0037, 0.0262], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0061, 0.0061, 0.0117, 0.0062, 0.0069, 0.0064, 0.0110], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 10:25:19,518 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:37,109 INFO [train.py:904] (5/8) Epoch 6, batch 400, loss[loss=0.1799, simple_loss=0.2666, pruned_loss=0.04663, over 16827.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2911, pruned_loss=0.07117, over 2886002.44 frames. ], batch size: 42, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:25:55,431 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:26:01,252 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.56 vs. limit=5.0 2023-04-28 10:26:01,710 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.109e+02 2.940e+02 3.610e+02 4.239e+02 7.005e+02, threshold=7.220e+02, percent-clipped=1.0 2023-04-28 10:26:05,727 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:26:21,465 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0568, 4.6208, 3.4313, 2.6623, 3.0955, 2.4573, 4.7896, 4.3988], device='cuda:5'), covar=tensor([0.2141, 0.0483, 0.1126, 0.1672, 0.2446, 0.1591, 0.0304, 0.0650], device='cuda:5'), in_proj_covar=tensor([0.0278, 0.0250, 0.0271, 0.0246, 0.0271, 0.0203, 0.0244, 0.0256], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:26:45,078 INFO [train.py:904] (5/8) Epoch 6, batch 450, loss[loss=0.2108, simple_loss=0.2995, pruned_loss=0.06108, over 17038.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2889, pruned_loss=0.06963, over 2976965.62 frames. ], batch size: 53, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:26:47,045 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:28,595 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-04-28 10:27:29,556 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 10:27:52,971 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:53,835 INFO [train.py:904] (5/8) Epoch 6, batch 500, loss[loss=0.2254, simple_loss=0.2927, pruned_loss=0.07905, over 16883.00 frames. ], tot_loss[loss=0.212, simple_loss=0.287, pruned_loss=0.06857, over 3050247.63 frames. ], batch size: 90, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:28:15,933 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-28 10:28:17,361 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 3.090e+02 3.738e+02 4.449e+02 8.454e+02, threshold=7.475e+02, percent-clipped=2.0 2023-04-28 10:29:01,777 INFO [train.py:904] (5/8) Epoch 6, batch 550, loss[loss=0.1977, simple_loss=0.2931, pruned_loss=0.05114, over 17119.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2853, pruned_loss=0.06651, over 3124153.31 frames. ], batch size: 48, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:29:03,268 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:30:13,399 INFO [train.py:904] (5/8) Epoch 6, batch 600, loss[loss=0.2161, simple_loss=0.3052, pruned_loss=0.06347, over 17058.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2858, pruned_loss=0.0663, over 3172370.89 frames. ], batch size: 55, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:30:38,609 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.906e+02 3.488e+02 4.318e+02 1.008e+03, threshold=6.975e+02, percent-clipped=1.0 2023-04-28 10:30:55,693 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:10,937 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:21,423 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6233, 4.8267, 5.1759, 5.2073, 5.2635, 4.8934, 4.6115, 4.6390], device='cuda:5'), covar=tensor([0.0440, 0.0451, 0.0582, 0.0628, 0.0495, 0.0409, 0.1268, 0.0449], device='cuda:5'), in_proj_covar=tensor([0.0254, 0.0255, 0.0254, 0.0249, 0.0302, 0.0268, 0.0377, 0.0224], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 10:31:23,796 INFO [train.py:904] (5/8) Epoch 6, batch 650, loss[loss=0.1897, simple_loss=0.2821, pruned_loss=0.04863, over 17298.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2844, pruned_loss=0.06615, over 3199743.31 frames. ], batch size: 52, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:31:47,313 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:03,803 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6806, 4.4420, 4.7054, 4.9346, 5.0486, 4.5349, 4.9296, 4.9728], device='cuda:5'), covar=tensor([0.0933, 0.0789, 0.1195, 0.0498, 0.0408, 0.0755, 0.0701, 0.0490], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0526, 0.0671, 0.0533, 0.0397, 0.0391, 0.0421, 0.0454], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:32:14,610 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:21,284 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:31,069 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1893, 1.5907, 2.5394, 3.0428, 2.7662, 3.3465, 1.5495, 3.3274], device='cuda:5'), covar=tensor([0.0063, 0.0237, 0.0117, 0.0096, 0.0092, 0.0071, 0.0286, 0.0059], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0145, 0.0131, 0.0128, 0.0131, 0.0095, 0.0144, 0.0080], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 10:32:31,572 INFO [train.py:904] (5/8) Epoch 6, batch 700, loss[loss=0.2203, simple_loss=0.3084, pruned_loss=0.06611, over 17042.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2846, pruned_loss=0.06626, over 3229196.30 frames. ], batch size: 55, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:32:35,197 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:49,294 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:32:57,147 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.941e+02 3.067e+02 3.781e+02 4.816e+02 1.338e+03, threshold=7.562e+02, percent-clipped=6.0 2023-04-28 10:33:19,522 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:32,445 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:39,720 INFO [train.py:904] (5/8) Epoch 6, batch 750, loss[loss=0.2079, simple_loss=0.2767, pruned_loss=0.06961, over 16640.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2846, pruned_loss=0.06593, over 3254554.82 frames. ], batch size: 76, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:33:56,289 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:33:56,529 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7944, 3.7475, 2.7804, 2.3890, 2.6795, 2.2336, 3.7441, 3.6945], device='cuda:5'), covar=tensor([0.1767, 0.0502, 0.1165, 0.1534, 0.1864, 0.1452, 0.0441, 0.0738], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0252, 0.0274, 0.0247, 0.0277, 0.0205, 0.0247, 0.0261], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:34:19,281 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:34:53,036 INFO [train.py:904] (5/8) Epoch 6, batch 800, loss[loss=0.2352, simple_loss=0.2987, pruned_loss=0.08587, over 16723.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2845, pruned_loss=0.06597, over 3261883.44 frames. ], batch size: 134, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:34:59,959 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:35:19,941 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.998e+02 2.921e+02 3.386e+02 4.262e+02 8.289e+02, threshold=6.772e+02, percent-clipped=2.0 2023-04-28 10:35:26,938 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4243, 4.1626, 3.9247, 1.8315, 2.9165, 2.1570, 3.6275, 3.8269], device='cuda:5'), covar=tensor([0.0263, 0.0529, 0.0465, 0.1693, 0.0745, 0.1013, 0.0653, 0.0822], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0129, 0.0153, 0.0141, 0.0130, 0.0125, 0.0137, 0.0139], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 10:36:01,905 INFO [train.py:904] (5/8) Epoch 6, batch 850, loss[loss=0.1993, simple_loss=0.2829, pruned_loss=0.05788, over 16777.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2842, pruned_loss=0.06547, over 3276793.79 frames. ], batch size: 83, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:36:04,210 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:36:35,375 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-28 10:37:10,643 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:37:11,530 INFO [train.py:904] (5/8) Epoch 6, batch 900, loss[loss=0.1712, simple_loss=0.2505, pruned_loss=0.04599, over 16796.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2824, pruned_loss=0.0649, over 3288876.55 frames. ], batch size: 39, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:37:39,521 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 2.905e+02 3.526e+02 4.406e+02 8.198e+02, threshold=7.052e+02, percent-clipped=7.0 2023-04-28 10:38:22,588 INFO [train.py:904] (5/8) Epoch 6, batch 950, loss[loss=0.2076, simple_loss=0.2721, pruned_loss=0.07151, over 12210.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2825, pruned_loss=0.06552, over 3294889.03 frames. ], batch size: 246, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:38:46,346 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:38:55,619 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-28 10:38:59,107 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5348, 4.4395, 4.7141, 2.3249, 4.9378, 4.8614, 3.6446, 4.1329], device='cuda:5'), covar=tensor([0.0466, 0.0109, 0.0114, 0.0970, 0.0038, 0.0063, 0.0249, 0.0204], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0091, 0.0082, 0.0139, 0.0072, 0.0081, 0.0115, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 10:39:10,671 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 10:39:12,964 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:25,715 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:26,807 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:30,783 INFO [train.py:904] (5/8) Epoch 6, batch 1000, loss[loss=0.1931, simple_loss=0.2613, pruned_loss=0.06239, over 16799.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2805, pruned_loss=0.06497, over 3296817.43 frames. ], batch size: 83, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:39:51,037 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:51,422 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 10:39:57,331 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 2.827e+02 3.297e+02 4.150e+02 8.828e+02, threshold=6.593e+02, percent-clipped=4.0 2023-04-28 10:40:12,059 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8760, 3.9877, 4.3306, 3.0734, 3.9013, 4.2247, 4.0979, 2.6768], device='cuda:5'), covar=tensor([0.0291, 0.0032, 0.0023, 0.0208, 0.0036, 0.0037, 0.0029, 0.0255], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0064, 0.0061, 0.0117, 0.0064, 0.0071, 0.0064, 0.0110], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 10:40:27,247 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5340, 3.3186, 3.8207, 2.7605, 3.6782, 3.7754, 3.7682, 2.1965], device='cuda:5'), covar=tensor([0.0287, 0.0125, 0.0032, 0.0202, 0.0043, 0.0058, 0.0039, 0.0289], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0064, 0.0061, 0.0118, 0.0064, 0.0072, 0.0065, 0.0110], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 10:40:39,946 INFO [train.py:904] (5/8) Epoch 6, batch 1050, loss[loss=0.189, simple_loss=0.2643, pruned_loss=0.05681, over 17177.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2807, pruned_loss=0.06552, over 3301620.77 frames. ], batch size: 46, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:40:49,419 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:41:18,030 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:41:49,460 INFO [train.py:904] (5/8) Epoch 6, batch 1100, loss[loss=0.196, simple_loss=0.2773, pruned_loss=0.05734, over 17136.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2795, pruned_loss=0.06486, over 3296299.40 frames. ], batch size: 46, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:41:49,806 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:42:16,639 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.702e+02 3.374e+02 4.258e+02 9.547e+02, threshold=6.748e+02, percent-clipped=3.0 2023-04-28 10:42:25,312 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:42:59,426 INFO [train.py:904] (5/8) Epoch 6, batch 1150, loss[loss=0.1926, simple_loss=0.2859, pruned_loss=0.0496, over 17114.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2793, pruned_loss=0.06402, over 3299633.14 frames. ], batch size: 49, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:43:35,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6558, 4.8278, 4.7893, 4.8695, 4.7179, 5.3428, 4.8669, 4.6370], device='cuda:5'), covar=tensor([0.1026, 0.1538, 0.1527, 0.1601, 0.2776, 0.0976, 0.1241, 0.1954], device='cuda:5'), in_proj_covar=tensor([0.0287, 0.0413, 0.0409, 0.0355, 0.0478, 0.0443, 0.0335, 0.0475], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 10:44:08,005 INFO [train.py:904] (5/8) Epoch 6, batch 1200, loss[loss=0.2079, simple_loss=0.2782, pruned_loss=0.06878, over 15512.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2783, pruned_loss=0.06338, over 3306540.75 frames. ], batch size: 190, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:44:16,550 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9846, 3.2810, 2.6872, 4.6036, 4.1263, 4.3181, 1.6511, 3.1677], device='cuda:5'), covar=tensor([0.1156, 0.0475, 0.0970, 0.0084, 0.0274, 0.0337, 0.1239, 0.0624], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0143, 0.0166, 0.0092, 0.0182, 0.0185, 0.0160, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 10:44:33,634 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.935e+02 3.395e+02 4.046e+02 1.078e+03, threshold=6.791e+02, percent-clipped=4.0 2023-04-28 10:45:20,396 INFO [train.py:904] (5/8) Epoch 6, batch 1250, loss[loss=0.1883, simple_loss=0.2616, pruned_loss=0.05752, over 16796.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2784, pruned_loss=0.0637, over 3303722.92 frames. ], batch size: 39, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:46:13,399 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:15,775 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:26,721 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:30,471 INFO [train.py:904] (5/8) Epoch 6, batch 1300, loss[loss=0.1795, simple_loss=0.2657, pruned_loss=0.04667, over 16819.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2783, pruned_loss=0.06359, over 3305533.84 frames. ], batch size: 42, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:46:42,527 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-28 10:46:52,669 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 10:46:58,326 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.918e+02 3.490e+02 4.202e+02 7.834e+02, threshold=6.979e+02, percent-clipped=4.0 2023-04-28 10:47:20,305 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:30,533 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2206, 4.4279, 3.4260, 2.4983, 3.2794, 2.4453, 4.9019, 4.2725], device='cuda:5'), covar=tensor([0.1907, 0.0576, 0.1147, 0.1569, 0.2247, 0.1495, 0.0255, 0.0662], device='cuda:5'), in_proj_covar=tensor([0.0283, 0.0255, 0.0274, 0.0250, 0.0285, 0.0206, 0.0251, 0.0271], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:47:34,137 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52096.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:41,515 INFO [train.py:904] (5/8) Epoch 6, batch 1350, loss[loss=0.201, simple_loss=0.2899, pruned_loss=0.056, over 17117.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2793, pruned_loss=0.06372, over 3308811.89 frames. ], batch size: 48, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:47:42,061 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:44,914 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:48:51,205 INFO [train.py:904] (5/8) Epoch 6, batch 1400, loss[loss=0.211, simple_loss=0.2756, pruned_loss=0.07324, over 16274.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2791, pruned_loss=0.06327, over 3307095.02 frames. ], batch size: 165, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:48:51,486 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:49:19,188 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 2.727e+02 3.349e+02 4.009e+02 8.330e+02, threshold=6.698e+02, percent-clipped=2.0 2023-04-28 10:49:49,781 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52193.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:49:58,143 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:50:00,567 INFO [train.py:904] (5/8) Epoch 6, batch 1450, loss[loss=0.2199, simple_loss=0.2759, pruned_loss=0.08199, over 16840.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2783, pruned_loss=0.06346, over 3315060.80 frames. ], batch size: 83, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:50:48,903 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9008, 3.4566, 3.0493, 5.1207, 4.6269, 4.7285, 1.6450, 3.6285], device='cuda:5'), covar=tensor([0.1258, 0.0526, 0.0985, 0.0109, 0.0249, 0.0323, 0.1394, 0.0556], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0144, 0.0167, 0.0094, 0.0185, 0.0187, 0.0161, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 10:51:10,836 INFO [train.py:904] (5/8) Epoch 6, batch 1500, loss[loss=0.2271, simple_loss=0.2866, pruned_loss=0.08379, over 16742.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2776, pruned_loss=0.06355, over 3316254.75 frames. ], batch size: 89, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:51:15,526 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:51:38,537 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 3.021e+02 3.666e+02 4.541e+02 9.509e+02, threshold=7.332e+02, percent-clipped=3.0 2023-04-28 10:51:41,137 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8561, 4.2905, 2.2134, 4.7235, 2.8026, 4.6046, 2.3202, 3.3113], device='cuda:5'), covar=tensor([0.0190, 0.0342, 0.1581, 0.0049, 0.0823, 0.0353, 0.1441, 0.0613], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0164, 0.0183, 0.0088, 0.0164, 0.0197, 0.0189, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 10:52:18,642 INFO [train.py:904] (5/8) Epoch 6, batch 1550, loss[loss=0.1624, simple_loss=0.2384, pruned_loss=0.04318, over 15842.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2797, pruned_loss=0.06501, over 3314060.98 frames. ], batch size: 35, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:52:28,402 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 10:52:34,669 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52312.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:52:37,812 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8361, 4.1074, 4.3797, 1.8682, 4.6182, 4.6883, 3.3128, 3.7168], device='cuda:5'), covar=tensor([0.0719, 0.0142, 0.0154, 0.1207, 0.0080, 0.0060, 0.0319, 0.0317], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0093, 0.0083, 0.0140, 0.0072, 0.0084, 0.0117, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 10:53:28,032 INFO [train.py:904] (5/8) Epoch 6, batch 1600, loss[loss=0.2023, simple_loss=0.2664, pruned_loss=0.0691, over 16790.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2827, pruned_loss=0.06612, over 3314769.52 frames. ], batch size: 102, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:53:55,826 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.932e+02 3.332e+02 3.992e+02 7.476e+02, threshold=6.664e+02, percent-clipped=1.0 2023-04-28 10:53:59,292 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:30,470 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:37,133 INFO [train.py:904] (5/8) Epoch 6, batch 1650, loss[loss=0.219, simple_loss=0.2888, pruned_loss=0.07454, over 16904.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2842, pruned_loss=0.06645, over 3312637.25 frames. ], batch size: 90, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:54:40,211 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:55:10,827 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4894, 3.4198, 4.1097, 2.6984, 3.6778, 3.9655, 3.7030, 2.3418], device='cuda:5'), covar=tensor([0.0311, 0.0137, 0.0023, 0.0235, 0.0050, 0.0042, 0.0047, 0.0260], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0062, 0.0060, 0.0113, 0.0062, 0.0071, 0.0065, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 10:55:42,546 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9045, 3.8067, 4.3760, 3.2577, 3.9948, 4.2266, 3.8604, 2.4503], device='cuda:5'), covar=tensor([0.0243, 0.0040, 0.0017, 0.0162, 0.0030, 0.0033, 0.0040, 0.0250], device='cuda:5'), in_proj_covar=tensor([0.0115, 0.0061, 0.0059, 0.0111, 0.0061, 0.0070, 0.0064, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 10:55:45,566 INFO [train.py:904] (5/8) Epoch 6, batch 1700, loss[loss=0.165, simple_loss=0.2498, pruned_loss=0.04013, over 16875.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2863, pruned_loss=0.06675, over 3317087.45 frames. ], batch size: 42, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:55:45,912 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:56:14,198 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.939e+02 3.544e+02 4.398e+02 9.866e+02, threshold=7.088e+02, percent-clipped=4.0 2023-04-28 10:56:57,349 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 10:56:57,701 INFO [train.py:904] (5/8) Epoch 6, batch 1750, loss[loss=0.1874, simple_loss=0.2638, pruned_loss=0.05548, over 16747.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2876, pruned_loss=0.06753, over 3312587.16 frames. ], batch size: 39, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:05,493 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52549.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:58:07,617 INFO [train.py:904] (5/8) Epoch 6, batch 1800, loss[loss=0.2311, simple_loss=0.3127, pruned_loss=0.07475, over 12411.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2885, pruned_loss=0.06767, over 3309239.20 frames. ], batch size: 247, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:14,084 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:58:27,960 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0767, 3.9206, 4.1330, 4.3427, 4.4295, 3.9681, 4.1738, 4.3859], device='cuda:5'), covar=tensor([0.1018, 0.0839, 0.1174, 0.0506, 0.0457, 0.1186, 0.1446, 0.0471], device='cuda:5'), in_proj_covar=tensor([0.0443, 0.0548, 0.0695, 0.0560, 0.0416, 0.0406, 0.0435, 0.0467], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:58:36,301 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 3.092e+02 3.773e+02 4.797e+02 9.614e+02, threshold=7.547e+02, percent-clipped=5.0 2023-04-28 10:58:55,255 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8586, 2.1209, 2.3082, 4.7168, 1.8628, 3.2616, 2.3079, 2.4603], device='cuda:5'), covar=tensor([0.0606, 0.2454, 0.1313, 0.0256, 0.3174, 0.1134, 0.2163, 0.2573], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0331, 0.0274, 0.0318, 0.0374, 0.0348, 0.0301, 0.0403], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:58:58,321 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0239, 4.7152, 4.9503, 5.2536, 5.4016, 4.7244, 5.3391, 5.3393], device='cuda:5'), covar=tensor([0.1016, 0.0832, 0.1338, 0.0466, 0.0355, 0.0691, 0.0398, 0.0386], device='cuda:5'), in_proj_covar=tensor([0.0444, 0.0549, 0.0695, 0.0560, 0.0416, 0.0409, 0.0437, 0.0468], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:59:17,740 INFO [train.py:904] (5/8) Epoch 6, batch 1850, loss[loss=0.1851, simple_loss=0.2739, pruned_loss=0.04811, over 17113.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2903, pruned_loss=0.06835, over 3319375.48 frames. ], batch size: 47, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 10:59:38,423 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 10:59:40,037 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0857, 4.8072, 4.8862, 5.2850, 5.3823, 4.7702, 5.3088, 5.3577], device='cuda:5'), covar=tensor([0.1206, 0.0842, 0.1791, 0.0623, 0.0716, 0.0643, 0.0591, 0.0520], device='cuda:5'), in_proj_covar=tensor([0.0441, 0.0545, 0.0692, 0.0559, 0.0415, 0.0408, 0.0434, 0.0464], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 10:59:55,290 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7551, 4.5471, 4.1057, 2.1057, 3.1838, 2.6913, 3.9887, 4.1599], device='cuda:5'), covar=tensor([0.0301, 0.0439, 0.0398, 0.1487, 0.0657, 0.0856, 0.0616, 0.0894], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0134, 0.0151, 0.0139, 0.0130, 0.0124, 0.0139, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 11:00:25,198 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2457, 4.2118, 4.1894, 3.7222, 4.1913, 1.7220, 4.0096, 3.9905], device='cuda:5'), covar=tensor([0.0076, 0.0059, 0.0087, 0.0231, 0.0061, 0.1652, 0.0087, 0.0121], device='cuda:5'), in_proj_covar=tensor([0.0103, 0.0090, 0.0139, 0.0137, 0.0104, 0.0150, 0.0120, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:00:27,094 INFO [train.py:904] (5/8) Epoch 6, batch 1900, loss[loss=0.1874, simple_loss=0.2822, pruned_loss=0.04626, over 17119.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2892, pruned_loss=0.06718, over 3320171.26 frames. ], batch size: 48, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:00:51,241 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:00:52,624 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8638, 5.0381, 5.5509, 5.5949, 5.5229, 5.1027, 5.0444, 4.7662], device='cuda:5'), covar=tensor([0.0278, 0.0359, 0.0253, 0.0298, 0.0374, 0.0232, 0.0727, 0.0315], device='cuda:5'), in_proj_covar=tensor([0.0269, 0.0267, 0.0265, 0.0263, 0.0322, 0.0280, 0.0395, 0.0233], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 11:00:54,523 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.841e+02 2.667e+02 3.390e+02 4.181e+02 1.051e+03, threshold=6.780e+02, percent-clipped=5.0 2023-04-28 11:01:10,411 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0818, 4.1768, 4.5715, 4.6152, 4.6173, 4.1802, 4.2319, 4.0301], device='cuda:5'), covar=tensor([0.0301, 0.0397, 0.0338, 0.0395, 0.0332, 0.0296, 0.0724, 0.0448], device='cuda:5'), in_proj_covar=tensor([0.0269, 0.0267, 0.0266, 0.0263, 0.0322, 0.0280, 0.0396, 0.0233], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 11:01:26,548 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1109, 3.9381, 4.1257, 4.3339, 4.4540, 3.9820, 4.0840, 4.4143], device='cuda:5'), covar=tensor([0.1000, 0.0829, 0.1254, 0.0555, 0.0430, 0.1067, 0.1443, 0.0471], device='cuda:5'), in_proj_covar=tensor([0.0446, 0.0549, 0.0699, 0.0564, 0.0420, 0.0415, 0.0439, 0.0472], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:01:30,617 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:01:36,640 INFO [train.py:904] (5/8) Epoch 6, batch 1950, loss[loss=0.2113, simple_loss=0.2856, pruned_loss=0.06846, over 16714.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2878, pruned_loss=0.06616, over 3320272.90 frames. ], batch size: 134, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:01:57,117 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9986, 4.3201, 2.0277, 4.6739, 2.7960, 4.6756, 2.3325, 3.1712], device='cuda:5'), covar=tensor([0.0137, 0.0212, 0.1468, 0.0041, 0.0734, 0.0338, 0.1245, 0.0515], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0161, 0.0176, 0.0085, 0.0160, 0.0193, 0.0185, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 11:02:36,911 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52744.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:02:47,598 INFO [train.py:904] (5/8) Epoch 6, batch 2000, loss[loss=0.1933, simple_loss=0.2662, pruned_loss=0.06019, over 16796.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2877, pruned_loss=0.06587, over 3306121.03 frames. ], batch size: 39, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:02:49,201 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52752.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:03:15,697 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 2.923e+02 3.553e+02 4.132e+02 6.210e+02, threshold=7.105e+02, percent-clipped=0.0 2023-04-28 11:03:51,902 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8567, 4.8016, 5.3011, 5.3849, 5.3744, 4.9918, 4.9015, 4.5620], device='cuda:5'), covar=tensor([0.0249, 0.0397, 0.0411, 0.0389, 0.0414, 0.0256, 0.0754, 0.0337], device='cuda:5'), in_proj_covar=tensor([0.0265, 0.0266, 0.0264, 0.0262, 0.0319, 0.0279, 0.0394, 0.0229], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 11:03:57,181 INFO [train.py:904] (5/8) Epoch 6, batch 2050, loss[loss=0.2255, simple_loss=0.285, pruned_loss=0.08296, over 16833.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2872, pruned_loss=0.06563, over 3310725.25 frames. ], batch size: 90, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:04:05,312 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7145, 6.0347, 5.8050, 6.0011, 5.4224, 5.1771, 5.6584, 6.1794], device='cuda:5'), covar=tensor([0.0779, 0.0773, 0.0986, 0.0513, 0.0671, 0.0536, 0.0597, 0.0727], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0551, 0.0452, 0.0354, 0.0336, 0.0345, 0.0447, 0.0389], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:04:15,476 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:05:06,244 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:05:08,351 INFO [train.py:904] (5/8) Epoch 6, batch 2100, loss[loss=0.2051, simple_loss=0.2923, pruned_loss=0.05889, over 17141.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2879, pruned_loss=0.06642, over 3323717.30 frames. ], batch size: 48, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:05:36,281 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.933e+02 3.490e+02 4.297e+02 7.985e+02, threshold=6.979e+02, percent-clipped=3.0 2023-04-28 11:06:12,954 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:06:18,012 INFO [train.py:904] (5/8) Epoch 6, batch 2150, loss[loss=0.2014, simple_loss=0.2941, pruned_loss=0.05434, over 17262.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2892, pruned_loss=0.06705, over 3313000.28 frames. ], batch size: 52, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:06:33,073 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 11:07:29,321 INFO [train.py:904] (5/8) Epoch 6, batch 2200, loss[loss=0.2102, simple_loss=0.2799, pruned_loss=0.07022, over 16793.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2894, pruned_loss=0.06718, over 3308032.54 frames. ], batch size: 102, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:07:29,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9622, 1.8387, 2.4205, 2.9607, 2.6972, 3.4555, 1.7784, 3.2972], device='cuda:5'), covar=tensor([0.0107, 0.0236, 0.0156, 0.0139, 0.0147, 0.0073, 0.0236, 0.0063], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0151, 0.0134, 0.0137, 0.0141, 0.0101, 0.0145, 0.0085], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 11:07:30,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9521, 4.3812, 3.5099, 2.4890, 3.1990, 2.6280, 4.7029, 4.2267], device='cuda:5'), covar=tensor([0.2218, 0.0558, 0.1129, 0.1680, 0.2125, 0.1408, 0.0301, 0.0673], device='cuda:5'), in_proj_covar=tensor([0.0282, 0.0256, 0.0270, 0.0250, 0.0287, 0.0205, 0.0246, 0.0270], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:07:52,280 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:07:56,705 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.110e+02 3.315e+02 3.926e+02 4.640e+02 8.894e+02, threshold=7.853e+02, percent-clipped=3.0 2023-04-28 11:08:38,593 INFO [train.py:904] (5/8) Epoch 6, batch 2250, loss[loss=0.215, simple_loss=0.2869, pruned_loss=0.07158, over 16266.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2911, pruned_loss=0.06823, over 3305994.04 frames. ], batch size: 164, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:08:59,899 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:09:13,444 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.88 vs. limit=5.0 2023-04-28 11:09:21,297 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3486, 2.1748, 1.6432, 1.8876, 2.6030, 2.4188, 2.7382, 2.7548], device='cuda:5'), covar=tensor([0.0058, 0.0186, 0.0233, 0.0242, 0.0093, 0.0148, 0.0093, 0.0099], device='cuda:5'), in_proj_covar=tensor([0.0094, 0.0165, 0.0164, 0.0161, 0.0163, 0.0167, 0.0153, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:09:32,754 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2020, 5.1855, 4.9871, 4.3811, 5.0068, 1.9268, 4.7821, 5.0052], device='cuda:5'), covar=tensor([0.0047, 0.0041, 0.0088, 0.0289, 0.0049, 0.1658, 0.0078, 0.0115], device='cuda:5'), in_proj_covar=tensor([0.0102, 0.0090, 0.0138, 0.0136, 0.0104, 0.0148, 0.0120, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:09:47,159 INFO [train.py:904] (5/8) Epoch 6, batch 2300, loss[loss=0.1787, simple_loss=0.272, pruned_loss=0.0427, over 17114.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.291, pruned_loss=0.06811, over 3311484.31 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:09:52,002 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 11:09:54,551 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:10:15,635 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 3.084e+02 3.786e+02 4.874e+02 1.349e+03, threshold=7.572e+02, percent-clipped=4.0 2023-04-28 11:10:57,589 INFO [train.py:904] (5/8) Epoch 6, batch 2350, loss[loss=0.1704, simple_loss=0.2647, pruned_loss=0.03802, over 17123.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.291, pruned_loss=0.06861, over 3311801.38 frames. ], batch size: 47, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:10:58,925 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 11:11:08,095 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:11:20,272 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:11:38,242 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-28 11:12:08,723 INFO [train.py:904] (5/8) Epoch 6, batch 2400, loss[loss=0.1951, simple_loss=0.2869, pruned_loss=0.05169, over 17016.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.291, pruned_loss=0.06801, over 3321674.19 frames. ], batch size: 50, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:12:36,945 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 2.840e+02 3.422e+02 4.151e+02 8.672e+02, threshold=6.844e+02, percent-clipped=2.0 2023-04-28 11:13:12,788 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6400, 3.8995, 4.2161, 3.0706, 3.7410, 4.0202, 3.7139, 2.1712], device='cuda:5'), covar=tensor([0.0306, 0.0056, 0.0028, 0.0199, 0.0052, 0.0052, 0.0043, 0.0320], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0063, 0.0062, 0.0115, 0.0063, 0.0072, 0.0065, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 11:13:19,231 INFO [train.py:904] (5/8) Epoch 6, batch 2450, loss[loss=0.1937, simple_loss=0.2814, pruned_loss=0.05303, over 17091.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2906, pruned_loss=0.06703, over 3325342.66 frames. ], batch size: 47, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:13:33,748 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:13:37,208 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.11 vs. limit=5.0 2023-04-28 11:13:56,434 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1588, 4.3772, 3.4755, 2.5332, 3.1948, 2.5165, 4.6030, 4.3857], device='cuda:5'), covar=tensor([0.1902, 0.0566, 0.1020, 0.1570, 0.2043, 0.1376, 0.0291, 0.0600], device='cuda:5'), in_proj_covar=tensor([0.0276, 0.0254, 0.0266, 0.0247, 0.0287, 0.0202, 0.0243, 0.0268], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:14:16,171 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:14:29,176 INFO [train.py:904] (5/8) Epoch 6, batch 2500, loss[loss=0.2191, simple_loss=0.309, pruned_loss=0.06461, over 16714.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2907, pruned_loss=0.06692, over 3323195.53 frames. ], batch size: 57, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:14:40,023 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:14:57,076 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.700e+02 3.263e+02 4.236e+02 1.034e+03, threshold=6.525e+02, percent-clipped=4.0 2023-04-28 11:15:38,471 INFO [train.py:904] (5/8) Epoch 6, batch 2550, loss[loss=0.2097, simple_loss=0.2984, pruned_loss=0.06045, over 16776.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2905, pruned_loss=0.06678, over 3327956.08 frames. ], batch size: 62, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:15:40,070 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:42,129 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9025, 4.6015, 4.8475, 5.1108, 5.2375, 4.6306, 5.2458, 5.2167], device='cuda:5'), covar=tensor([0.0939, 0.0884, 0.1404, 0.0494, 0.0388, 0.0601, 0.0382, 0.0383], device='cuda:5'), in_proj_covar=tensor([0.0435, 0.0532, 0.0689, 0.0551, 0.0416, 0.0408, 0.0432, 0.0462], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:15:43,268 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:58,346 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:16:18,515 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4085, 3.9090, 3.3154, 2.0088, 2.9315, 2.2968, 3.7095, 3.7145], device='cuda:5'), covar=tensor([0.0230, 0.0547, 0.0585, 0.1497, 0.0709, 0.0971, 0.0496, 0.0715], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0136, 0.0152, 0.0139, 0.0130, 0.0124, 0.0139, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 11:16:48,723 INFO [train.py:904] (5/8) Epoch 6, batch 2600, loss[loss=0.1732, simple_loss=0.2608, pruned_loss=0.0428, over 17172.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2902, pruned_loss=0.06643, over 3322724.89 frames. ], batch size: 46, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:16:53,831 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8222, 5.2192, 5.3505, 5.2023, 5.1162, 5.7794, 5.2057, 5.1115], device='cuda:5'), covar=tensor([0.0842, 0.1432, 0.1317, 0.1689, 0.2577, 0.0860, 0.1239, 0.2213], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0429, 0.0425, 0.0370, 0.0489, 0.0454, 0.0352, 0.0491], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 11:17:09,396 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:11,647 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:16,417 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.796e+02 3.424e+02 4.100e+02 9.306e+02, threshold=6.847e+02, percent-clipped=3.0 2023-04-28 11:17:17,106 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-28 11:17:24,072 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53376.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:59,457 INFO [train.py:904] (5/8) Epoch 6, batch 2650, loss[loss=0.196, simple_loss=0.2837, pruned_loss=0.05416, over 16888.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2906, pruned_loss=0.06615, over 3324053.33 frames. ], batch size: 96, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:18:10,413 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:14,739 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:37,180 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:18:56,650 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-04-28 11:19:09,533 INFO [train.py:904] (5/8) Epoch 6, batch 2700, loss[loss=0.1961, simple_loss=0.2933, pruned_loss=0.04943, over 17120.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2905, pruned_loss=0.06514, over 3330891.68 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:19:16,986 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:19:38,365 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.946e+02 3.581e+02 4.267e+02 8.371e+02, threshold=7.163e+02, percent-clipped=3.0 2023-04-28 11:20:19,917 INFO [train.py:904] (5/8) Epoch 6, batch 2750, loss[loss=0.1726, simple_loss=0.2573, pruned_loss=0.04397, over 16999.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2902, pruned_loss=0.06462, over 3321814.75 frames. ], batch size: 41, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:20:22,268 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4323, 4.2931, 3.7368, 2.0396, 3.1151, 2.4874, 3.8273, 3.9614], device='cuda:5'), covar=tensor([0.0291, 0.0464, 0.0486, 0.1491, 0.0678, 0.0868, 0.0595, 0.0875], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0137, 0.0155, 0.0140, 0.0130, 0.0124, 0.0140, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 11:20:25,203 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4741, 2.2358, 1.5829, 2.0553, 2.7188, 2.5968, 2.7667, 2.7670], device='cuda:5'), covar=tensor([0.0063, 0.0168, 0.0256, 0.0213, 0.0090, 0.0129, 0.0108, 0.0109], device='cuda:5'), in_proj_covar=tensor([0.0095, 0.0167, 0.0166, 0.0164, 0.0163, 0.0170, 0.0156, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:20:50,388 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9443, 5.2167, 4.9115, 5.0118, 4.6782, 4.6175, 4.7631, 5.3016], device='cuda:5'), covar=tensor([0.0777, 0.0755, 0.0941, 0.0448, 0.0683, 0.0706, 0.0686, 0.0710], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0550, 0.0455, 0.0349, 0.0336, 0.0346, 0.0442, 0.0385], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:21:03,272 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 11:21:31,902 INFO [train.py:904] (5/8) Epoch 6, batch 2800, loss[loss=0.2736, simple_loss=0.3278, pruned_loss=0.1097, over 11980.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2906, pruned_loss=0.0649, over 3313268.91 frames. ], batch size: 246, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:22:01,719 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.830e+02 3.513e+02 4.760e+02 1.218e+03, threshold=7.026e+02, percent-clipped=5.0 2023-04-28 11:22:02,250 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2852, 3.8252, 3.7938, 1.7878, 3.9821, 3.9516, 3.1387, 2.9371], device='cuda:5'), covar=tensor([0.0820, 0.0100, 0.0129, 0.1121, 0.0045, 0.0077, 0.0332, 0.0384], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0094, 0.0085, 0.0140, 0.0071, 0.0085, 0.0119, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 11:22:37,902 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:22:43,370 INFO [train.py:904] (5/8) Epoch 6, batch 2850, loss[loss=0.1897, simple_loss=0.2685, pruned_loss=0.05548, over 15889.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2898, pruned_loss=0.06516, over 3307339.01 frames. ], batch size: 35, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:23:31,787 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4326, 3.3950, 3.4048, 2.9362, 3.3259, 2.1021, 3.1744, 2.9948], device='cuda:5'), covar=tensor([0.0091, 0.0079, 0.0137, 0.0247, 0.0076, 0.1395, 0.0101, 0.0167], device='cuda:5'), in_proj_covar=tensor([0.0104, 0.0094, 0.0145, 0.0142, 0.0108, 0.0152, 0.0124, 0.0136], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:23:51,285 INFO [train.py:904] (5/8) Epoch 6, batch 2900, loss[loss=0.2791, simple_loss=0.3271, pruned_loss=0.1156, over 15584.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2894, pruned_loss=0.06635, over 3305282.90 frames. ], batch size: 191, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:24:04,557 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:24:19,426 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53671.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:24:20,250 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.863e+02 3.412e+02 4.656e+02 8.517e+02, threshold=6.823e+02, percent-clipped=1.0 2023-04-28 11:24:59,798 INFO [train.py:904] (5/8) Epoch 6, batch 2950, loss[loss=0.2221, simple_loss=0.306, pruned_loss=0.06912, over 17134.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2907, pruned_loss=0.06712, over 3306096.39 frames. ], batch size: 49, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:25:15,604 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:25:30,206 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:25:35,055 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2462, 5.1838, 5.0977, 4.7991, 4.6058, 5.0661, 5.0919, 4.7130], device='cuda:5'), covar=tensor([0.0431, 0.0279, 0.0189, 0.0203, 0.1073, 0.0278, 0.0253, 0.0582], device='cuda:5'), in_proj_covar=tensor([0.0212, 0.0230, 0.0249, 0.0221, 0.0285, 0.0251, 0.0176, 0.0280], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 11:26:08,576 INFO [train.py:904] (5/8) Epoch 6, batch 3000, loss[loss=0.2159, simple_loss=0.2884, pruned_loss=0.07174, over 16833.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2903, pruned_loss=0.06739, over 3310110.93 frames. ], batch size: 102, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:26:08,576 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 11:26:17,402 INFO [train.py:938] (5/8) Epoch 6, validation: loss=0.1514, simple_loss=0.258, pruned_loss=0.02246, over 944034.00 frames. 2023-04-28 11:26:17,402 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 11:26:29,474 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:26:45,018 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 11:26:45,969 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.839e+02 3.365e+02 4.010e+02 8.699e+02, threshold=6.731e+02, percent-clipped=2.0 2023-04-28 11:27:25,830 INFO [train.py:904] (5/8) Epoch 6, batch 3050, loss[loss=0.2042, simple_loss=0.2761, pruned_loss=0.0662, over 16730.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2896, pruned_loss=0.06663, over 3319288.15 frames. ], batch size: 102, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:28:14,407 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9596, 4.3168, 2.2039, 4.5407, 2.8189, 4.6002, 2.2472, 3.1732], device='cuda:5'), covar=tensor([0.0128, 0.0221, 0.1390, 0.0058, 0.0732, 0.0312, 0.1334, 0.0548], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0163, 0.0177, 0.0088, 0.0163, 0.0198, 0.0186, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 11:28:26,354 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1588, 1.6549, 2.3263, 2.9967, 2.8491, 3.4796, 2.1198, 3.2112], device='cuda:5'), covar=tensor([0.0079, 0.0254, 0.0154, 0.0120, 0.0117, 0.0068, 0.0194, 0.0085], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0152, 0.0136, 0.0136, 0.0141, 0.0103, 0.0145, 0.0088], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 11:28:35,661 INFO [train.py:904] (5/8) Epoch 6, batch 3100, loss[loss=0.2257, simple_loss=0.2929, pruned_loss=0.07925, over 16690.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2883, pruned_loss=0.06561, over 3318645.61 frames. ], batch size: 134, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:28:43,739 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8322, 3.2366, 2.8221, 5.1257, 4.6248, 4.6556, 1.5270, 3.3843], device='cuda:5'), covar=tensor([0.1242, 0.0554, 0.0970, 0.0097, 0.0237, 0.0307, 0.1401, 0.0636], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0146, 0.0169, 0.0098, 0.0198, 0.0192, 0.0164, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 11:29:05,910 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.841e+02 3.355e+02 4.130e+02 6.611e+02, threshold=6.710e+02, percent-clipped=0.0 2023-04-28 11:29:40,281 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:29:45,140 INFO [train.py:904] (5/8) Epoch 6, batch 3150, loss[loss=0.2193, simple_loss=0.2826, pruned_loss=0.07804, over 16912.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2875, pruned_loss=0.06578, over 3320830.11 frames. ], batch size: 109, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:30:14,243 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8595, 2.6934, 2.6307, 1.9821, 2.3660, 2.6267, 2.5371, 1.8449], device='cuda:5'), covar=tensor([0.0270, 0.0045, 0.0037, 0.0214, 0.0064, 0.0049, 0.0047, 0.0251], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0061, 0.0062, 0.0112, 0.0063, 0.0071, 0.0064, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 11:30:49,057 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:30:56,671 INFO [train.py:904] (5/8) Epoch 6, batch 3200, loss[loss=0.209, simple_loss=0.2991, pruned_loss=0.05943, over 17026.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.286, pruned_loss=0.06505, over 3312835.65 frames. ], batch size: 55, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:31:09,521 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:31:25,871 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53971.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:31:26,626 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.936e+02 3.448e+02 4.117e+02 1.255e+03, threshold=6.896e+02, percent-clipped=5.0 2023-04-28 11:32:09,694 INFO [train.py:904] (5/8) Epoch 6, batch 3250, loss[loss=0.229, simple_loss=0.2907, pruned_loss=0.08365, over 16903.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2861, pruned_loss=0.06488, over 3318423.91 frames. ], batch size: 116, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:32:19,696 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:35,419 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:40,211 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:33:17,051 INFO [train.py:904] (5/8) Epoch 6, batch 3300, loss[loss=0.2308, simple_loss=0.3027, pruned_loss=0.07948, over 16715.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2874, pruned_loss=0.06547, over 3319678.04 frames. ], batch size: 134, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:33:17,599 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1783, 2.2525, 2.4082, 4.8215, 2.0731, 3.2771, 2.4242, 2.5852], device='cuda:5'), covar=tensor([0.0588, 0.2578, 0.1349, 0.0264, 0.3332, 0.1303, 0.2155, 0.2904], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0341, 0.0278, 0.0322, 0.0380, 0.0360, 0.0308, 0.0410], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:33:45,199 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54071.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:33:46,071 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 3.091e+02 3.848e+02 4.464e+02 8.976e+02, threshold=7.696e+02, percent-clipped=3.0 2023-04-28 11:33:48,200 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8020, 4.5210, 4.7348, 4.9574, 5.1610, 4.5750, 5.1111, 5.0973], device='cuda:5'), covar=tensor([0.1145, 0.0849, 0.1522, 0.0544, 0.0423, 0.0657, 0.0439, 0.0450], device='cuda:5'), in_proj_covar=tensor([0.0447, 0.0537, 0.0698, 0.0561, 0.0423, 0.0418, 0.0434, 0.0472], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:34:26,275 INFO [train.py:904] (5/8) Epoch 6, batch 3350, loss[loss=0.1895, simple_loss=0.2767, pruned_loss=0.05115, over 17095.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2889, pruned_loss=0.0664, over 3313768.72 frames. ], batch size: 47, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:35:34,482 INFO [train.py:904] (5/8) Epoch 6, batch 3400, loss[loss=0.191, simple_loss=0.2713, pruned_loss=0.05536, over 17187.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2886, pruned_loss=0.06593, over 3318683.78 frames. ], batch size: 44, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:36:05,330 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.856e+02 2.871e+02 3.468e+02 4.191e+02 6.688e+02, threshold=6.936e+02, percent-clipped=0.0 2023-04-28 11:36:46,920 INFO [train.py:904] (5/8) Epoch 6, batch 3450, loss[loss=0.1613, simple_loss=0.2466, pruned_loss=0.03803, over 15795.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2868, pruned_loss=0.06518, over 3299454.06 frames. ], batch size: 35, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:37:58,632 INFO [train.py:904] (5/8) Epoch 6, batch 3500, loss[loss=0.2074, simple_loss=0.2735, pruned_loss=0.07066, over 16911.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2851, pruned_loss=0.06454, over 3311791.03 frames. ], batch size: 109, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:38:28,916 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.700e+02 3.164e+02 3.995e+02 8.205e+02, threshold=6.329e+02, percent-clipped=2.0 2023-04-28 11:39:10,637 INFO [train.py:904] (5/8) Epoch 6, batch 3550, loss[loss=0.195, simple_loss=0.2791, pruned_loss=0.05543, over 17175.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2837, pruned_loss=0.06406, over 3306190.62 frames. ], batch size: 46, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:39:20,539 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4672, 4.0482, 3.7094, 1.9744, 3.0463, 2.3179, 3.6029, 3.8334], device='cuda:5'), covar=tensor([0.0235, 0.0536, 0.0465, 0.1594, 0.0637, 0.0952, 0.0618, 0.0811], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0137, 0.0153, 0.0138, 0.0130, 0.0122, 0.0138, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 11:39:52,810 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5139, 3.4800, 3.9297, 2.6926, 3.5818, 3.8804, 3.6622, 2.1887], device='cuda:5'), covar=tensor([0.0295, 0.0136, 0.0044, 0.0224, 0.0047, 0.0067, 0.0046, 0.0299], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0059, 0.0062, 0.0112, 0.0062, 0.0070, 0.0065, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 11:40:21,756 INFO [train.py:904] (5/8) Epoch 6, batch 3600, loss[loss=0.2195, simple_loss=0.2903, pruned_loss=0.07437, over 16690.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2833, pruned_loss=0.06322, over 3317224.97 frames. ], batch size: 134, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:40:51,132 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 3.018e+02 3.592e+02 4.498e+02 7.311e+02, threshold=7.184e+02, percent-clipped=7.0 2023-04-28 11:41:33,847 INFO [train.py:904] (5/8) Epoch 6, batch 3650, loss[loss=0.1959, simple_loss=0.2576, pruned_loss=0.06707, over 16421.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2813, pruned_loss=0.0636, over 3316948.03 frames. ], batch size: 75, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:41:34,864 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 11:42:20,337 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 11:42:46,527 INFO [train.py:904] (5/8) Epoch 6, batch 3700, loss[loss=0.1897, simple_loss=0.2618, pruned_loss=0.05881, over 16427.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2804, pruned_loss=0.06567, over 3295838.82 frames. ], batch size: 75, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:43:17,521 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.788e+02 2.928e+02 3.594e+02 4.333e+02 7.725e+02, threshold=7.187e+02, percent-clipped=2.0 2023-04-28 11:43:25,756 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-28 11:43:34,297 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0292, 1.7444, 2.2367, 2.9769, 2.9073, 2.9522, 1.8349, 3.0452], device='cuda:5'), covar=tensor([0.0081, 0.0245, 0.0187, 0.0107, 0.0092, 0.0103, 0.0224, 0.0059], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0152, 0.0138, 0.0137, 0.0142, 0.0103, 0.0145, 0.0090], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 11:43:59,579 INFO [train.py:904] (5/8) Epoch 6, batch 3750, loss[loss=0.2273, simple_loss=0.298, pruned_loss=0.07828, over 15527.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2816, pruned_loss=0.06763, over 3276361.09 frames. ], batch size: 190, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:44:42,862 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6291, 2.9723, 2.7169, 4.4219, 3.8736, 4.3316, 1.2693, 3.5344], device='cuda:5'), covar=tensor([0.1341, 0.0507, 0.0876, 0.0091, 0.0204, 0.0283, 0.1416, 0.0472], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0143, 0.0166, 0.0099, 0.0192, 0.0190, 0.0159, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 11:44:50,494 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5048, 3.4672, 2.4380, 2.2089, 2.6012, 1.9977, 3.2845, 3.4121], device='cuda:5'), covar=tensor([0.2080, 0.0649, 0.1410, 0.1726, 0.1928, 0.1742, 0.0655, 0.0818], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0255, 0.0269, 0.0252, 0.0291, 0.0205, 0.0248, 0.0277], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 11:44:56,345 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 11:45:13,217 INFO [train.py:904] (5/8) Epoch 6, batch 3800, loss[loss=0.2343, simple_loss=0.2993, pruned_loss=0.08462, over 16773.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2822, pruned_loss=0.06904, over 3279231.52 frames. ], batch size: 134, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:45:46,021 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.146e+02 2.814e+02 3.370e+02 4.517e+02 7.386e+02, threshold=6.740e+02, percent-clipped=3.0 2023-04-28 11:46:28,653 INFO [train.py:904] (5/8) Epoch 6, batch 3850, loss[loss=0.2012, simple_loss=0.267, pruned_loss=0.06768, over 16813.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2819, pruned_loss=0.06913, over 3281470.60 frames. ], batch size: 90, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:46:32,756 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 11:46:46,948 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3646, 4.0088, 4.1215, 1.9535, 4.3638, 4.4189, 3.1270, 3.2781], device='cuda:5'), covar=tensor([0.0821, 0.0149, 0.0151, 0.1177, 0.0061, 0.0064, 0.0306, 0.0356], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0094, 0.0081, 0.0137, 0.0072, 0.0082, 0.0114, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 11:47:40,668 INFO [train.py:904] (5/8) Epoch 6, batch 3900, loss[loss=0.248, simple_loss=0.3166, pruned_loss=0.08963, over 12397.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2817, pruned_loss=0.06972, over 3276191.89 frames. ], batch size: 246, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:59,715 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:48:10,540 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 11:48:10,865 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.751e+02 3.451e+02 4.155e+02 1.044e+03, threshold=6.903e+02, percent-clipped=2.0 2023-04-28 11:48:35,418 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1988, 4.2788, 4.6159, 4.6190, 4.6076, 4.1864, 4.2774, 4.1196], device='cuda:5'), covar=tensor([0.0322, 0.0511, 0.0326, 0.0344, 0.0376, 0.0345, 0.0751, 0.0502], device='cuda:5'), in_proj_covar=tensor([0.0263, 0.0264, 0.0256, 0.0259, 0.0313, 0.0276, 0.0386, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 11:48:40,278 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8069, 2.7953, 2.7067, 1.8768, 2.5610, 2.7477, 2.6549, 1.7349], device='cuda:5'), covar=tensor([0.0269, 0.0043, 0.0032, 0.0221, 0.0047, 0.0046, 0.0041, 0.0241], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0058, 0.0061, 0.0112, 0.0060, 0.0069, 0.0064, 0.0104], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 11:48:53,477 INFO [train.py:904] (5/8) Epoch 6, batch 3950, loss[loss=0.2038, simple_loss=0.2669, pruned_loss=0.07032, over 16871.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2817, pruned_loss=0.0702, over 3278899.13 frames. ], batch size: 109, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:49:27,757 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:50:04,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4452, 2.4916, 1.8484, 2.3900, 3.1123, 2.7181, 3.4977, 3.3553], device='cuda:5'), covar=tensor([0.0032, 0.0219, 0.0307, 0.0232, 0.0096, 0.0219, 0.0055, 0.0087], device='cuda:5'), in_proj_covar=tensor([0.0094, 0.0167, 0.0165, 0.0161, 0.0160, 0.0167, 0.0154, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:50:04,726 INFO [train.py:904] (5/8) Epoch 6, batch 4000, loss[loss=0.1889, simple_loss=0.2623, pruned_loss=0.05775, over 17026.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2812, pruned_loss=0.07056, over 3288217.77 frames. ], batch size: 55, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:50:36,655 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.045e+02 2.875e+02 3.507e+02 4.223e+02 8.153e+02, threshold=7.015e+02, percent-clipped=4.0 2023-04-28 11:51:17,350 INFO [train.py:904] (5/8) Epoch 6, batch 4050, loss[loss=0.193, simple_loss=0.2694, pruned_loss=0.05834, over 16781.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2797, pruned_loss=0.06816, over 3292772.63 frames. ], batch size: 39, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:08,450 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 11:52:28,298 INFO [train.py:904] (5/8) Epoch 6, batch 4100, loss[loss=0.2212, simple_loss=0.3045, pruned_loss=0.0689, over 16800.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2814, pruned_loss=0.06758, over 3274570.57 frames. ], batch size: 83, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:45,962 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:53:01,851 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.498e+02 3.136e+02 3.951e+02 8.635e+02, threshold=6.272e+02, percent-clipped=2.0 2023-04-28 11:53:45,451 INFO [train.py:904] (5/8) Epoch 6, batch 4150, loss[loss=0.2848, simple_loss=0.3414, pruned_loss=0.1141, over 11605.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2905, pruned_loss=0.07207, over 3209776.30 frames. ], batch size: 247, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:53:58,081 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-28 11:54:19,609 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:55:01,102 INFO [train.py:904] (5/8) Epoch 6, batch 4200, loss[loss=0.2728, simple_loss=0.3505, pruned_loss=0.09749, over 16677.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.2983, pruned_loss=0.0743, over 3188763.30 frames. ], batch size: 134, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:55:33,719 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 3.214e+02 3.756e+02 4.661e+02 1.168e+03, threshold=7.512e+02, percent-clipped=6.0 2023-04-28 11:55:34,267 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1598, 4.4850, 2.3748, 5.0578, 3.0274, 4.8399, 2.9289, 3.2894], device='cuda:5'), covar=tensor([0.0130, 0.0229, 0.1504, 0.0035, 0.0705, 0.0307, 0.1191, 0.0579], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0161, 0.0181, 0.0085, 0.0165, 0.0194, 0.0187, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 11:56:15,277 INFO [train.py:904] (5/8) Epoch 6, batch 4250, loss[loss=0.2386, simple_loss=0.3017, pruned_loss=0.08775, over 12046.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3012, pruned_loss=0.07463, over 3161312.74 frames. ], batch size: 246, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:56:44,674 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:57:03,274 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6358, 4.9456, 4.6609, 4.6847, 4.3593, 4.3200, 4.4454, 4.9628], device='cuda:5'), covar=tensor([0.0659, 0.0646, 0.0829, 0.0457, 0.0604, 0.0837, 0.0665, 0.0591], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0519, 0.0436, 0.0332, 0.0321, 0.0334, 0.0425, 0.0368], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:57:29,405 INFO [train.py:904] (5/8) Epoch 6, batch 4300, loss[loss=0.2142, simple_loss=0.3116, pruned_loss=0.05839, over 16753.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3026, pruned_loss=0.07322, over 3169406.50 frames. ], batch size: 124, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:58:01,806 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.606e+02 3.247e+02 3.921e+02 6.703e+02, threshold=6.494e+02, percent-clipped=0.0 2023-04-28 11:58:43,136 INFO [train.py:904] (5/8) Epoch 6, batch 4350, loss[loss=0.2214, simple_loss=0.3067, pruned_loss=0.06802, over 16764.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.306, pruned_loss=0.07438, over 3163687.73 frames. ], batch size: 83, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:58:47,764 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7803, 6.0243, 5.6606, 5.8681, 5.4239, 4.9861, 5.5254, 6.1120], device='cuda:5'), covar=tensor([0.0553, 0.0537, 0.0749, 0.0419, 0.0508, 0.0513, 0.0579, 0.0598], device='cuda:5'), in_proj_covar=tensor([0.0396, 0.0524, 0.0440, 0.0336, 0.0322, 0.0339, 0.0428, 0.0372], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 11:59:53,059 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:59:53,505 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-28 11:59:56,594 INFO [train.py:904] (5/8) Epoch 6, batch 4400, loss[loss=0.2495, simple_loss=0.331, pruned_loss=0.08404, over 16659.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3084, pruned_loss=0.07533, over 3174799.21 frames. ], batch size: 62, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:00:27,149 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.770e+02 3.222e+02 4.108e+02 7.043e+02, threshold=6.445e+02, percent-clipped=2.0 2023-04-28 12:01:06,627 INFO [train.py:904] (5/8) Epoch 6, batch 4450, loss[loss=0.229, simple_loss=0.3095, pruned_loss=0.0743, over 16348.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3106, pruned_loss=0.0752, over 3195515.10 frames. ], batch size: 35, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:01:19,213 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:21,583 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:31,366 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:02:16,915 INFO [train.py:904] (5/8) Epoch 6, batch 4500, loss[loss=0.2494, simple_loss=0.3316, pruned_loss=0.08363, over 16796.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3101, pruned_loss=0.07452, over 3214573.68 frames. ], batch size: 124, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:02:47,705 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:02:48,407 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.800e+02 2.322e+02 2.860e+02 3.458e+02 5.535e+02, threshold=5.719e+02, percent-clipped=0.0 2023-04-28 12:03:29,212 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2742, 4.0665, 4.1221, 1.9305, 4.4359, 4.5432, 3.0491, 3.2576], device='cuda:5'), covar=tensor([0.0839, 0.0119, 0.0185, 0.1032, 0.0047, 0.0031, 0.0339, 0.0346], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0095, 0.0082, 0.0140, 0.0073, 0.0080, 0.0117, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 12:03:29,865 INFO [train.py:904] (5/8) Epoch 6, batch 4550, loss[loss=0.245, simple_loss=0.324, pruned_loss=0.08298, over 16607.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3099, pruned_loss=0.07421, over 3228213.65 frames. ], batch size: 57, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:03:38,460 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 12:03:40,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 12:03:57,604 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55320.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:04:41,598 INFO [train.py:904] (5/8) Epoch 6, batch 4600, loss[loss=0.2303, simple_loss=0.3153, pruned_loss=0.0726, over 16951.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3108, pruned_loss=0.07489, over 3191575.95 frames. ], batch size: 109, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:05:07,046 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:05:13,324 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.321e+02 2.807e+02 3.689e+02 1.502e+03, threshold=5.614e+02, percent-clipped=3.0 2023-04-28 12:05:52,756 INFO [train.py:904] (5/8) Epoch 6, batch 4650, loss[loss=0.2589, simple_loss=0.3389, pruned_loss=0.08944, over 15415.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3096, pruned_loss=0.07474, over 3189897.41 frames. ], batch size: 191, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:06:12,026 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5991, 4.9651, 4.9895, 5.0399, 4.9005, 5.5664, 5.0365, 4.8707], device='cuda:5'), covar=tensor([0.0853, 0.1244, 0.1158, 0.1361, 0.2410, 0.0849, 0.0949, 0.1965], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0396, 0.0393, 0.0347, 0.0454, 0.0417, 0.0324, 0.0466], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 12:06:46,846 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-04-28 12:07:03,022 INFO [train.py:904] (5/8) Epoch 6, batch 4700, loss[loss=0.2208, simple_loss=0.2986, pruned_loss=0.0715, over 16724.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3066, pruned_loss=0.07309, over 3194809.54 frames. ], batch size: 134, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:07:34,101 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.601e+02 3.111e+02 3.691e+02 9.251e+02, threshold=6.221e+02, percent-clipped=2.0 2023-04-28 12:08:12,130 INFO [train.py:904] (5/8) Epoch 6, batch 4750, loss[loss=0.2119, simple_loss=0.288, pruned_loss=0.06793, over 16676.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3028, pruned_loss=0.07127, over 3200439.85 frames. ], batch size: 62, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:08:17,381 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:08:36,552 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 12:09:22,947 INFO [train.py:904] (5/8) Epoch 6, batch 4800, loss[loss=0.1974, simple_loss=0.2792, pruned_loss=0.05782, over 16707.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2987, pruned_loss=0.06917, over 3210645.16 frames. ], batch size: 57, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:09:30,577 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 12:09:35,380 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8046, 3.9599, 3.1434, 2.4401, 3.0558, 2.4281, 4.2293, 4.0143], device='cuda:5'), covar=tensor([0.2057, 0.0643, 0.1233, 0.1572, 0.1858, 0.1380, 0.0353, 0.0615], device='cuda:5'), in_proj_covar=tensor([0.0283, 0.0252, 0.0271, 0.0254, 0.0288, 0.0203, 0.0248, 0.0268], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:09:45,466 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:46,592 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:53,994 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.469e+02 2.878e+02 3.568e+02 5.612e+02, threshold=5.756e+02, percent-clipped=0.0 2023-04-28 12:10:35,254 INFO [train.py:904] (5/8) Epoch 6, batch 4850, loss[loss=0.2201, simple_loss=0.3175, pruned_loss=0.06133, over 15570.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3004, pruned_loss=0.0689, over 3206994.69 frames. ], batch size: 190, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:11:29,559 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1697, 4.4598, 2.1933, 5.0243, 2.9199, 4.7502, 2.7170, 3.2101], device='cuda:5'), covar=tensor([0.0144, 0.0197, 0.1583, 0.0018, 0.0695, 0.0224, 0.1175, 0.0624], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0155, 0.0177, 0.0082, 0.0161, 0.0186, 0.0185, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 12:11:48,118 INFO [train.py:904] (5/8) Epoch 6, batch 4900, loss[loss=0.2173, simple_loss=0.3056, pruned_loss=0.06445, over 16886.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2988, pruned_loss=0.06711, over 3204812.09 frames. ], batch size: 116, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:12:15,961 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9017, 1.6662, 1.4913, 1.5399, 1.8282, 1.6646, 1.7347, 1.8355], device='cuda:5'), covar=tensor([0.0051, 0.0157, 0.0214, 0.0200, 0.0100, 0.0162, 0.0103, 0.0134], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0157, 0.0159, 0.0157, 0.0152, 0.0163, 0.0142, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:12:19,091 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.308e+02 2.929e+02 3.604e+02 9.107e+02, threshold=5.857e+02, percent-clipped=2.0 2023-04-28 12:12:59,385 INFO [train.py:904] (5/8) Epoch 6, batch 4950, loss[loss=0.2204, simple_loss=0.3069, pruned_loss=0.06697, over 16492.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.299, pruned_loss=0.06716, over 3200989.14 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:14:05,624 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:14:08,603 INFO [train.py:904] (5/8) Epoch 6, batch 5000, loss[loss=0.2139, simple_loss=0.2915, pruned_loss=0.06812, over 17042.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3009, pruned_loss=0.06744, over 3195174.99 frames. ], batch size: 53, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:14:38,599 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.723e+02 3.199e+02 3.755e+02 7.199e+02, threshold=6.398e+02, percent-clipped=5.0 2023-04-28 12:15:12,949 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8853, 3.2799, 3.3202, 1.5042, 3.5247, 3.5573, 2.6801, 2.5230], device='cuda:5'), covar=tensor([0.0904, 0.0134, 0.0111, 0.1221, 0.0059, 0.0061, 0.0402, 0.0472], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0093, 0.0080, 0.0137, 0.0071, 0.0078, 0.0115, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 12:15:21,204 INFO [train.py:904] (5/8) Epoch 6, batch 5050, loss[loss=0.2115, simple_loss=0.3001, pruned_loss=0.06146, over 15451.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.302, pruned_loss=0.06753, over 3193823.77 frames. ], batch size: 191, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:15:25,496 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:15:33,310 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:28,779 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:31,560 INFO [train.py:904] (5/8) Epoch 6, batch 5100, loss[loss=0.2091, simple_loss=0.288, pruned_loss=0.06507, over 16565.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2999, pruned_loss=0.06674, over 3199558.61 frames. ], batch size: 68, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:16:32,963 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:54,914 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:17:04,487 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.683e+02 3.138e+02 3.930e+02 6.713e+02, threshold=6.277e+02, percent-clipped=2.0 2023-04-28 12:17:45,240 INFO [train.py:904] (5/8) Epoch 6, batch 5150, loss[loss=0.2281, simple_loss=0.3077, pruned_loss=0.07426, over 12091.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2997, pruned_loss=0.06581, over 3195177.79 frames. ], batch size: 247, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:17:57,112 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:06,165 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:26,522 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 12:18:58,946 INFO [train.py:904] (5/8) Epoch 6, batch 5200, loss[loss=0.1843, simple_loss=0.2705, pruned_loss=0.04911, over 16918.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2981, pruned_loss=0.06507, over 3205839.44 frames. ], batch size: 116, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:19:00,366 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 12:19:22,220 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8061, 4.7989, 4.6183, 4.4902, 4.1861, 4.7570, 4.6000, 4.3683], device='cuda:5'), covar=tensor([0.0486, 0.0339, 0.0205, 0.0178, 0.0946, 0.0289, 0.0267, 0.0468], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0212, 0.0231, 0.0203, 0.0259, 0.0229, 0.0158, 0.0254], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:19:30,991 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.992e+02 2.558e+02 3.011e+02 3.644e+02 8.071e+02, threshold=6.021e+02, percent-clipped=1.0 2023-04-28 12:19:52,233 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3717, 4.2373, 4.8419, 4.7198, 4.7490, 4.4277, 4.4481, 4.2328], device='cuda:5'), covar=tensor([0.0229, 0.0379, 0.0277, 0.0404, 0.0399, 0.0237, 0.0682, 0.0363], device='cuda:5'), in_proj_covar=tensor([0.0256, 0.0251, 0.0252, 0.0252, 0.0304, 0.0272, 0.0374, 0.0221], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 12:20:15,828 INFO [train.py:904] (5/8) Epoch 6, batch 5250, loss[loss=0.1879, simple_loss=0.2777, pruned_loss=0.04899, over 16471.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2952, pruned_loss=0.06438, over 3197676.03 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:20:20,552 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7905, 2.5366, 2.4254, 3.7148, 2.7525, 3.6151, 1.3711, 2.8716], device='cuda:5'), covar=tensor([0.1195, 0.0549, 0.1028, 0.0071, 0.0175, 0.0338, 0.1433, 0.0669], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0143, 0.0169, 0.0094, 0.0188, 0.0187, 0.0163, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 12:20:24,862 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9460, 4.1960, 2.0525, 4.7759, 2.7529, 4.6213, 2.4379, 3.0868], device='cuda:5'), covar=tensor([0.0149, 0.0186, 0.1727, 0.0025, 0.0809, 0.0203, 0.1361, 0.0625], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0157, 0.0180, 0.0083, 0.0164, 0.0188, 0.0188, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 12:21:28,019 INFO [train.py:904] (5/8) Epoch 6, batch 5300, loss[loss=0.172, simple_loss=0.2544, pruned_loss=0.04487, over 16753.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2918, pruned_loss=0.06367, over 3207209.17 frames. ], batch size: 83, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:21:59,804 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.725e+02 2.658e+02 3.030e+02 3.663e+02 6.971e+02, threshold=6.060e+02, percent-clipped=3.0 2023-04-28 12:22:40,518 INFO [train.py:904] (5/8) Epoch 6, batch 5350, loss[loss=0.2176, simple_loss=0.2964, pruned_loss=0.06945, over 16625.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2899, pruned_loss=0.06271, over 3207807.05 frames. ], batch size: 57, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:22:45,594 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:44,964 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:52,044 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6868, 2.6466, 2.0600, 3.9276, 3.0351, 3.7445, 1.4616, 2.5612], device='cuda:5'), covar=tensor([0.1138, 0.0499, 0.1187, 0.0091, 0.0182, 0.0323, 0.1270, 0.0771], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0144, 0.0171, 0.0095, 0.0189, 0.0188, 0.0163, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 12:23:52,607 INFO [train.py:904] (5/8) Epoch 6, batch 5400, loss[loss=0.2373, simple_loss=0.3248, pruned_loss=0.07489, over 16260.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2937, pruned_loss=0.06407, over 3201690.58 frames. ], batch size: 165, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:24:26,636 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.610e+02 3.124e+02 3.605e+02 7.209e+02, threshold=6.248e+02, percent-clipped=3.0 2023-04-28 12:25:09,004 INFO [train.py:904] (5/8) Epoch 6, batch 5450, loss[loss=0.2565, simple_loss=0.33, pruned_loss=0.09156, over 16639.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2974, pruned_loss=0.06594, over 3198614.72 frames. ], batch size: 62, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:25:14,957 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:17,620 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:22,963 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 12:26:02,791 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6295, 3.8790, 3.1405, 2.3836, 2.9267, 2.5038, 4.1528, 4.1340], device='cuda:5'), covar=tensor([0.2189, 0.0660, 0.1226, 0.1546, 0.2042, 0.1339, 0.0391, 0.0511], device='cuda:5'), in_proj_covar=tensor([0.0277, 0.0247, 0.0266, 0.0249, 0.0279, 0.0201, 0.0245, 0.0258], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:26:27,336 INFO [train.py:904] (5/8) Epoch 6, batch 5500, loss[loss=0.2116, simple_loss=0.297, pruned_loss=0.06313, over 16315.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3054, pruned_loss=0.0718, over 3162971.66 frames. ], batch size: 35, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:27:01,684 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.186e+02 3.422e+02 4.497e+02 5.690e+02 1.317e+03, threshold=8.994e+02, percent-clipped=17.0 2023-04-28 12:27:43,445 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0032, 3.0910, 1.5398, 3.3077, 2.3134, 3.2549, 1.9067, 2.5378], device='cuda:5'), covar=tensor([0.0178, 0.0360, 0.1466, 0.0066, 0.0695, 0.0433, 0.1259, 0.0600], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0157, 0.0178, 0.0082, 0.0164, 0.0190, 0.0189, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 12:27:46,071 INFO [train.py:904] (5/8) Epoch 6, batch 5550, loss[loss=0.3117, simple_loss=0.3765, pruned_loss=0.1235, over 15152.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3142, pruned_loss=0.07858, over 3151839.21 frames. ], batch size: 190, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:29:07,892 INFO [train.py:904] (5/8) Epoch 6, batch 5600, loss[loss=0.2431, simple_loss=0.3201, pruned_loss=0.08305, over 16498.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3211, pruned_loss=0.08471, over 3113191.00 frames. ], batch size: 68, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:29:45,255 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.502e+02 4.330e+02 5.114e+02 6.903e+02 1.749e+03, threshold=1.023e+03, percent-clipped=9.0 2023-04-28 12:30:29,988 INFO [train.py:904] (5/8) Epoch 6, batch 5650, loss[loss=0.335, simple_loss=0.3739, pruned_loss=0.1481, over 11326.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3281, pruned_loss=0.09053, over 3081444.86 frames. ], batch size: 246, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:30:34,839 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:31:03,084 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3354, 2.4080, 1.7660, 2.2131, 2.8434, 2.4592, 3.2523, 3.1588], device='cuda:5'), covar=tensor([0.0032, 0.0215, 0.0300, 0.0235, 0.0127, 0.0203, 0.0077, 0.0095], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0160, 0.0164, 0.0161, 0.0156, 0.0166, 0.0146, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:31:48,519 INFO [train.py:904] (5/8) Epoch 6, batch 5700, loss[loss=0.3928, simple_loss=0.408, pruned_loss=0.1888, over 11672.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.3308, pruned_loss=0.09367, over 3052628.94 frames. ], batch size: 248, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:31:51,664 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:32:10,999 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 12:32:25,511 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.150e+02 5.062e+02 6.862e+02 1.724e+03, threshold=1.012e+03, percent-clipped=2.0 2023-04-28 12:33:08,914 INFO [train.py:904] (5/8) Epoch 6, batch 5750, loss[loss=0.2718, simple_loss=0.3402, pruned_loss=0.1017, over 16896.00 frames. ], tot_loss[loss=0.2623, simple_loss=0.3339, pruned_loss=0.09537, over 3044035.54 frames. ], batch size: 109, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:33:09,356 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:13,951 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:44,605 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:33:49,670 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7666, 2.0702, 2.3351, 4.6091, 1.9234, 2.9780, 2.2172, 2.2911], device='cuda:5'), covar=tensor([0.0677, 0.2804, 0.1409, 0.0278, 0.3637, 0.1461, 0.2429, 0.2758], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0336, 0.0277, 0.0311, 0.0381, 0.0348, 0.0301, 0.0396], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:33:55,022 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 12:34:21,042 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5613, 3.7939, 3.9182, 1.8676, 4.1205, 4.1161, 3.0278, 3.1621], device='cuda:5'), covar=tensor([0.0682, 0.0109, 0.0139, 0.1091, 0.0051, 0.0071, 0.0285, 0.0341], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0092, 0.0080, 0.0139, 0.0072, 0.0080, 0.0114, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 12:34:30,335 INFO [train.py:904] (5/8) Epoch 6, batch 5800, loss[loss=0.2296, simple_loss=0.3161, pruned_loss=0.0715, over 16930.00 frames. ], tot_loss[loss=0.26, simple_loss=0.3332, pruned_loss=0.09335, over 3057175.82 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:34:32,767 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:35:05,969 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.365e+02 3.759e+02 4.783e+02 6.283e+02 1.634e+03, threshold=9.566e+02, percent-clipped=2.0 2023-04-28 12:35:23,454 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:35:49,151 INFO [train.py:904] (5/8) Epoch 6, batch 5850, loss[loss=0.2706, simple_loss=0.3466, pruned_loss=0.09729, over 16396.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3299, pruned_loss=0.09035, over 3072697.58 frames. ], batch size: 146, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:35:54,363 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7990, 3.7055, 3.2718, 2.5281, 3.1410, 2.4464, 4.5005, 4.0060], device='cuda:5'), covar=tensor([0.2377, 0.0989, 0.1261, 0.1454, 0.1964, 0.1391, 0.0341, 0.0729], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0248, 0.0267, 0.0249, 0.0281, 0.0202, 0.0247, 0.0258], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:35:57,899 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2700, 4.2274, 4.7804, 4.7387, 4.6882, 4.3654, 4.3129, 4.1145], device='cuda:5'), covar=tensor([0.0258, 0.0372, 0.0263, 0.0361, 0.0428, 0.0259, 0.0841, 0.0443], device='cuda:5'), in_proj_covar=tensor([0.0257, 0.0249, 0.0254, 0.0254, 0.0307, 0.0274, 0.0374, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 12:36:19,040 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4070, 1.9523, 1.4773, 1.7925, 2.4183, 2.0802, 2.4573, 2.5367], device='cuda:5'), covar=tensor([0.0058, 0.0226, 0.0301, 0.0249, 0.0123, 0.0203, 0.0099, 0.0120], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0159, 0.0164, 0.0161, 0.0156, 0.0165, 0.0146, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:37:08,263 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 12:37:11,653 INFO [train.py:904] (5/8) Epoch 6, batch 5900, loss[loss=0.2333, simple_loss=0.3073, pruned_loss=0.07968, over 16172.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3287, pruned_loss=0.08889, over 3101769.43 frames. ], batch size: 165, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:37:52,139 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.523e+02 3.504e+02 4.467e+02 5.504e+02 1.096e+03, threshold=8.934e+02, percent-clipped=2.0 2023-04-28 12:38:33,957 INFO [train.py:904] (5/8) Epoch 6, batch 5950, loss[loss=0.2409, simple_loss=0.3369, pruned_loss=0.07243, over 16847.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3293, pruned_loss=0.0877, over 3089997.43 frames. ], batch size: 102, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:38:36,184 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 12:38:59,794 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7519, 4.7378, 4.6166, 4.4193, 4.1734, 4.6404, 4.5855, 4.3615], device='cuda:5'), covar=tensor([0.0427, 0.0263, 0.0188, 0.0186, 0.0756, 0.0292, 0.0254, 0.0423], device='cuda:5'), in_proj_covar=tensor([0.0189, 0.0209, 0.0222, 0.0196, 0.0250, 0.0225, 0.0158, 0.0252], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:39:28,396 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6902, 2.6760, 1.7473, 2.7921, 2.1890, 2.7504, 1.9871, 2.3665], device='cuda:5'), covar=tensor([0.0179, 0.0356, 0.1125, 0.0084, 0.0619, 0.0446, 0.1063, 0.0509], device='cuda:5'), in_proj_covar=tensor([0.0121, 0.0151, 0.0172, 0.0080, 0.0159, 0.0184, 0.0183, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 12:39:52,047 INFO [train.py:904] (5/8) Epoch 6, batch 6000, loss[loss=0.2263, simple_loss=0.3033, pruned_loss=0.0746, over 16683.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3281, pruned_loss=0.08684, over 3101597.91 frames. ], batch size: 57, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:52,047 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 12:40:01,518 INFO [train.py:938] (5/8) Epoch 6, validation: loss=0.18, simple_loss=0.2922, pruned_loss=0.03386, over 944034.00 frames. 2023-04-28 12:40:01,519 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 12:40:36,515 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.218e+02 3.558e+02 4.306e+02 5.339e+02 1.394e+03, threshold=8.611e+02, percent-clipped=5.0 2023-04-28 12:41:04,616 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 12:41:19,701 INFO [train.py:904] (5/8) Epoch 6, batch 6050, loss[loss=0.2265, simple_loss=0.3131, pruned_loss=0.06995, over 16851.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3258, pruned_loss=0.08569, over 3113450.18 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:41:20,624 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:41:23,947 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:42:13,287 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8789, 2.6352, 2.6056, 1.8532, 2.4016, 2.5419, 2.5061, 1.8726], device='cuda:5'), covar=tensor([0.0263, 0.0034, 0.0043, 0.0205, 0.0059, 0.0062, 0.0043, 0.0228], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0054, 0.0058, 0.0113, 0.0061, 0.0071, 0.0064, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 12:42:18,184 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0725, 5.3849, 5.0857, 5.1808, 4.7133, 4.5925, 4.8445, 5.4311], device='cuda:5'), covar=tensor([0.0693, 0.0675, 0.0950, 0.0523, 0.0656, 0.0688, 0.0691, 0.0765], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0531, 0.0453, 0.0349, 0.0325, 0.0342, 0.0437, 0.0386], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:42:33,680 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:42:36,668 INFO [train.py:904] (5/8) Epoch 6, batch 6100, loss[loss=0.2339, simple_loss=0.3166, pruned_loss=0.07562, over 16868.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3255, pruned_loss=0.08482, over 3117942.54 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:42:58,696 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:43:13,032 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.259e+02 3.305e+02 4.282e+02 5.206e+02 1.267e+03, threshold=8.564e+02, percent-clipped=3.0 2023-04-28 12:43:20,812 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:43:56,263 INFO [train.py:904] (5/8) Epoch 6, batch 6150, loss[loss=0.2012, simple_loss=0.2828, pruned_loss=0.05987, over 16289.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3239, pruned_loss=0.08475, over 3103416.09 frames. ], batch size: 35, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:44:08,558 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:44:55,639 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 12:45:06,414 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-28 12:45:17,265 INFO [train.py:904] (5/8) Epoch 6, batch 6200, loss[loss=0.2878, simple_loss=0.3382, pruned_loss=0.1187, over 11570.00 frames. ], tot_loss[loss=0.245, simple_loss=0.3216, pruned_loss=0.08421, over 3086515.86 frames. ], batch size: 247, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:45:46,243 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:45:53,733 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 3.479e+02 4.387e+02 5.634e+02 1.000e+03, threshold=8.774e+02, percent-clipped=2.0 2023-04-28 12:46:22,796 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:26,551 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:27,831 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8414, 2.6687, 2.5916, 1.9167, 2.4453, 2.5294, 2.5408, 1.8145], device='cuda:5'), covar=tensor([0.0289, 0.0033, 0.0042, 0.0208, 0.0051, 0.0057, 0.0040, 0.0262], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0053, 0.0058, 0.0113, 0.0060, 0.0070, 0.0063, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 12:46:34,331 INFO [train.py:904] (5/8) Epoch 6, batch 6250, loss[loss=0.2965, simple_loss=0.3479, pruned_loss=0.1225, over 11611.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3208, pruned_loss=0.08419, over 3084410.68 frames. ], batch size: 248, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:20,957 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7448, 3.7650, 4.2245, 4.1880, 4.2112, 3.8451, 3.9060, 3.8295], device='cuda:5'), covar=tensor([0.0280, 0.0471, 0.0374, 0.0435, 0.0390, 0.0326, 0.0795, 0.0453], device='cuda:5'), in_proj_covar=tensor([0.0261, 0.0253, 0.0258, 0.0253, 0.0306, 0.0274, 0.0377, 0.0226], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 12:47:50,600 INFO [train.py:904] (5/8) Epoch 6, batch 6300, loss[loss=0.2185, simple_loss=0.3036, pruned_loss=0.06674, over 16928.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3204, pruned_loss=0.08333, over 3095520.72 frames. ], batch size: 109, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:56,720 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:47:59,845 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:48:29,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.167e+02 3.660e+02 4.420e+02 5.766e+02 1.393e+03, threshold=8.841e+02, percent-clipped=5.0 2023-04-28 12:48:51,862 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1035, 3.9565, 3.8345, 2.3301, 3.4183, 3.8031, 3.7064, 2.1274], device='cuda:5'), covar=tensor([0.0364, 0.0018, 0.0031, 0.0256, 0.0050, 0.0060, 0.0029, 0.0293], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0054, 0.0059, 0.0114, 0.0062, 0.0071, 0.0064, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 12:49:09,385 INFO [train.py:904] (5/8) Epoch 6, batch 6350, loss[loss=0.2331, simple_loss=0.3114, pruned_loss=0.07743, over 16872.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3211, pruned_loss=0.08431, over 3087254.75 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:49:20,907 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4830, 4.0694, 3.7395, 2.0056, 2.9775, 2.5754, 3.7790, 3.9174], device='cuda:5'), covar=tensor([0.0180, 0.0351, 0.0482, 0.1504, 0.0708, 0.0778, 0.0521, 0.0663], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0126, 0.0152, 0.0139, 0.0132, 0.0124, 0.0138, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 12:50:18,356 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1652, 3.8427, 3.7248, 2.2196, 3.3592, 3.6529, 3.6469, 1.9773], device='cuda:5'), covar=tensor([0.0338, 0.0018, 0.0026, 0.0259, 0.0050, 0.0073, 0.0027, 0.0298], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0054, 0.0059, 0.0114, 0.0061, 0.0071, 0.0064, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 12:50:26,696 INFO [train.py:904] (5/8) Epoch 6, batch 6400, loss[loss=0.254, simple_loss=0.3254, pruned_loss=0.0913, over 16966.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3213, pruned_loss=0.08481, over 3097325.58 frames. ], batch size: 109, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:50:38,787 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:50:58,321 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8358, 2.0972, 1.5873, 1.8182, 2.5672, 2.3719, 2.8606, 2.8704], device='cuda:5'), covar=tensor([0.0058, 0.0236, 0.0317, 0.0302, 0.0166, 0.0208, 0.0137, 0.0116], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0159, 0.0163, 0.0161, 0.0155, 0.0165, 0.0148, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:51:01,323 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.642e+02 4.267e+02 5.177e+02 6.433e+02 1.516e+03, threshold=1.035e+03, percent-clipped=6.0 2023-04-28 12:51:08,489 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:51:19,367 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:35,434 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 12:51:36,849 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:42,181 INFO [train.py:904] (5/8) Epoch 6, batch 6450, loss[loss=0.2274, simple_loss=0.3023, pruned_loss=0.0762, over 16217.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3201, pruned_loss=0.08304, over 3119194.91 frames. ], batch size: 165, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:52:26,738 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:52:35,153 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2345, 1.8871, 2.1774, 3.7780, 1.8772, 2.6181, 2.1174, 2.0896], device='cuda:5'), covar=tensor([0.0642, 0.2546, 0.1320, 0.0346, 0.3075, 0.1341, 0.2196, 0.2288], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0337, 0.0278, 0.0311, 0.0383, 0.0348, 0.0305, 0.0397], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:52:52,973 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:52:57,253 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:00,874 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6505, 4.9502, 5.0184, 5.0016, 5.0024, 5.4732, 5.1464, 4.9274], device='cuda:5'), covar=tensor([0.0787, 0.1389, 0.1286, 0.1445, 0.1935, 0.0791, 0.0978, 0.1854], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0405, 0.0412, 0.0357, 0.0467, 0.0441, 0.0338, 0.0482], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 12:53:03,835 INFO [train.py:904] (5/8) Epoch 6, batch 6500, loss[loss=0.2233, simple_loss=0.3038, pruned_loss=0.07144, over 16744.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.318, pruned_loss=0.08233, over 3115262.87 frames. ], batch size: 83, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:53:15,536 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:53:25,370 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:39,383 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3095, 4.3337, 4.8342, 4.7676, 4.7972, 4.4103, 4.2062, 4.2930], device='cuda:5'), covar=tensor([0.0404, 0.0543, 0.0390, 0.0489, 0.0527, 0.0401, 0.1239, 0.0469], device='cuda:5'), in_proj_covar=tensor([0.0256, 0.0250, 0.0255, 0.0248, 0.0302, 0.0270, 0.0372, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 12:53:41,780 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 3.641e+02 4.251e+02 5.070e+02 1.044e+03, threshold=8.501e+02, percent-clipped=1.0 2023-04-28 12:54:20,424 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2121, 1.6748, 2.3979, 3.1275, 2.8772, 3.3274, 1.5203, 3.2839], device='cuda:5'), covar=tensor([0.0050, 0.0283, 0.0154, 0.0087, 0.0102, 0.0085, 0.0350, 0.0056], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0147, 0.0130, 0.0130, 0.0134, 0.0100, 0.0146, 0.0087], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 12:54:25,252 INFO [train.py:904] (5/8) Epoch 6, batch 6550, loss[loss=0.2942, simple_loss=0.3471, pruned_loss=0.1206, over 11645.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3211, pruned_loss=0.08336, over 3102453.39 frames. ], batch size: 248, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:54:26,297 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0742, 3.2303, 3.5150, 3.4793, 3.4732, 3.2051, 3.2710, 3.3949], device='cuda:5'), covar=tensor([0.0394, 0.0616, 0.0402, 0.0475, 0.0535, 0.0471, 0.0783, 0.0439], device='cuda:5'), in_proj_covar=tensor([0.0256, 0.0250, 0.0255, 0.0248, 0.0301, 0.0270, 0.0371, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 12:54:29,274 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5583, 2.6081, 1.6146, 2.6581, 2.0920, 2.7145, 1.8475, 2.2657], device='cuda:5'), covar=tensor([0.0156, 0.0292, 0.1125, 0.0078, 0.0594, 0.0374, 0.1029, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0153, 0.0176, 0.0082, 0.0161, 0.0188, 0.0187, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 12:54:31,820 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:55:05,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3935, 4.3500, 4.2613, 4.1229, 3.8348, 4.3323, 4.1450, 4.0302], device='cuda:5'), covar=tensor([0.0482, 0.0390, 0.0208, 0.0186, 0.0803, 0.0305, 0.0412, 0.0476], device='cuda:5'), in_proj_covar=tensor([0.0190, 0.0211, 0.0223, 0.0196, 0.0249, 0.0227, 0.0161, 0.0256], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:55:40,135 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57349.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:55:42,173 INFO [train.py:904] (5/8) Epoch 6, batch 6600, loss[loss=0.334, simple_loss=0.373, pruned_loss=0.1475, over 11517.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3238, pruned_loss=0.08488, over 3089784.58 frames. ], batch size: 247, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:55:43,212 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:56:18,770 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.503e+02 3.780e+02 4.683e+02 5.878e+02 1.016e+03, threshold=9.365e+02, percent-clipped=5.0 2023-04-28 12:56:47,425 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7900, 4.1020, 3.8521, 3.9003, 3.5637, 3.7132, 3.8370, 4.0378], device='cuda:5'), covar=tensor([0.0874, 0.0766, 0.0903, 0.0524, 0.0684, 0.1270, 0.0682, 0.0847], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0530, 0.0451, 0.0347, 0.0329, 0.0347, 0.0434, 0.0380], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:56:56,499 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:57:00,535 INFO [train.py:904] (5/8) Epoch 6, batch 6650, loss[loss=0.237, simple_loss=0.3126, pruned_loss=0.08074, over 16642.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.3248, pruned_loss=0.08678, over 3069610.48 frames. ], batch size: 134, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:57:58,044 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2719, 3.1914, 3.2270, 3.3995, 3.3840, 3.1955, 3.3691, 3.4266], device='cuda:5'), covar=tensor([0.0837, 0.0703, 0.1048, 0.0523, 0.0655, 0.1740, 0.0874, 0.0643], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0500, 0.0644, 0.0520, 0.0391, 0.0385, 0.0406, 0.0437], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:58:15,927 INFO [train.py:904] (5/8) Epoch 6, batch 6700, loss[loss=0.2414, simple_loss=0.3164, pruned_loss=0.08315, over 16538.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3236, pruned_loss=0.08639, over 3085522.02 frames. ], batch size: 75, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:58:19,993 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0191, 3.9832, 3.8490, 3.2811, 3.8773, 1.6840, 3.7015, 3.5867], device='cuda:5'), covar=tensor([0.0069, 0.0061, 0.0103, 0.0275, 0.0064, 0.1955, 0.0084, 0.0138], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0085, 0.0130, 0.0129, 0.0098, 0.0148, 0.0113, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 12:58:20,081 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7383, 2.5240, 2.2924, 3.3737, 2.5490, 3.6432, 1.3581, 2.7581], device='cuda:5'), covar=tensor([0.1339, 0.0586, 0.1144, 0.0092, 0.0185, 0.0357, 0.1556, 0.0796], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0145, 0.0172, 0.0096, 0.0197, 0.0192, 0.0166, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 12:58:29,601 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:29,639 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:34,642 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:54,097 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.152e+02 3.718e+02 4.462e+02 5.305e+02 7.989e+02, threshold=8.923e+02, percent-clipped=0.0 2023-04-28 12:59:24,863 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:59:34,111 INFO [train.py:904] (5/8) Epoch 6, batch 6750, loss[loss=0.2275, simple_loss=0.3091, pruned_loss=0.07293, over 16874.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3235, pruned_loss=0.08708, over 3083267.89 frames. ], batch size: 116, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:59:43,095 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:08,117 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:36,612 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:49,721 INFO [train.py:904] (5/8) Epoch 6, batch 6800, loss[loss=0.2386, simple_loss=0.3167, pruned_loss=0.0803, over 16664.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3233, pruned_loss=0.08682, over 3081479.30 frames. ], batch size: 134, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:00:54,410 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 13:00:58,805 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:10,621 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:27,507 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.272e+02 3.624e+02 4.504e+02 5.898e+02 1.165e+03, threshold=9.008e+02, percent-clipped=3.0 2023-04-28 13:01:43,263 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-28 13:02:03,639 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6251, 2.5934, 1.7234, 2.7186, 2.1653, 2.7365, 1.9365, 2.3326], device='cuda:5'), covar=tensor([0.0210, 0.0439, 0.1274, 0.0102, 0.0701, 0.0610, 0.1187, 0.0555], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0154, 0.0177, 0.0083, 0.0161, 0.0189, 0.0188, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 13:02:06,903 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:02:07,776 INFO [train.py:904] (5/8) Epoch 6, batch 6850, loss[loss=0.2449, simple_loss=0.3331, pruned_loss=0.07837, over 17004.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3241, pruned_loss=0.08738, over 3074427.21 frames. ], batch size: 55, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:02:18,833 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-28 13:02:24,718 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:02:29,344 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 13:02:30,207 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2566, 3.9048, 3.4284, 1.6575, 2.9064, 2.4378, 3.5792, 3.9534], device='cuda:5'), covar=tensor([0.0204, 0.0426, 0.0548, 0.1721, 0.0748, 0.0875, 0.0588, 0.0496], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0127, 0.0153, 0.0140, 0.0133, 0.0125, 0.0139, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 13:03:22,394 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:03:24,927 INFO [train.py:904] (5/8) Epoch 6, batch 6900, loss[loss=0.2442, simple_loss=0.3266, pruned_loss=0.08088, over 16936.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3258, pruned_loss=0.08599, over 3096256.47 frames. ], batch size: 109, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:03:25,325 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:03:33,546 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4194, 4.1941, 4.3797, 4.5723, 4.7219, 4.2916, 4.6792, 4.7024], device='cuda:5'), covar=tensor([0.0924, 0.0796, 0.1183, 0.0506, 0.0426, 0.0749, 0.0427, 0.0372], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0500, 0.0642, 0.0517, 0.0393, 0.0384, 0.0406, 0.0432], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:04:02,716 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.992e+02 3.443e+02 4.296e+02 5.667e+02 1.150e+03, threshold=8.592e+02, percent-clipped=3.0 2023-04-28 13:04:11,479 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3595, 2.9025, 2.5319, 2.3214, 2.3151, 2.1990, 2.8404, 2.9668], device='cuda:5'), covar=tensor([0.1604, 0.0633, 0.1056, 0.1265, 0.1450, 0.1294, 0.0383, 0.0627], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0249, 0.0267, 0.0251, 0.0280, 0.0202, 0.0246, 0.0259], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:04:30,464 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 13:04:32,139 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5348, 1.8841, 2.1483, 4.1793, 1.8984, 2.6453, 2.1447, 2.1301], device='cuda:5'), covar=tensor([0.0644, 0.2698, 0.1430, 0.0305, 0.3296, 0.1522, 0.2240, 0.2520], device='cuda:5'), in_proj_covar=tensor([0.0325, 0.0336, 0.0278, 0.0311, 0.0383, 0.0350, 0.0303, 0.0398], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:04:37,919 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:39,809 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6028, 4.6367, 4.4049, 3.8017, 4.4613, 1.5493, 4.2003, 4.3369], device='cuda:5'), covar=tensor([0.0060, 0.0043, 0.0106, 0.0281, 0.0056, 0.1922, 0.0085, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0097, 0.0084, 0.0129, 0.0128, 0.0097, 0.0147, 0.0112, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:04:40,819 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:43,536 INFO [train.py:904] (5/8) Epoch 6, batch 6950, loss[loss=0.2475, simple_loss=0.3251, pruned_loss=0.08493, over 16605.00 frames. ], tot_loss[loss=0.2524, simple_loss=0.3281, pruned_loss=0.08837, over 3077879.52 frames. ], batch size: 57, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:05:34,122 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4021, 4.3870, 4.2679, 4.1017, 3.8182, 4.3393, 4.2264, 3.9808], device='cuda:5'), covar=tensor([0.0572, 0.0368, 0.0239, 0.0218, 0.0878, 0.0358, 0.0366, 0.0594], device='cuda:5'), in_proj_covar=tensor([0.0189, 0.0210, 0.0221, 0.0193, 0.0247, 0.0226, 0.0160, 0.0254], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:05:34,155 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:05:40,027 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 13:05:44,570 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0778, 4.8826, 5.0679, 5.3064, 5.4871, 4.8026, 5.4060, 5.3863], device='cuda:5'), covar=tensor([0.1293, 0.0780, 0.1150, 0.0462, 0.0463, 0.0560, 0.0423, 0.0398], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0494, 0.0633, 0.0508, 0.0387, 0.0380, 0.0402, 0.0428], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:05:51,957 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9407, 3.8022, 3.9589, 4.1355, 4.2235, 3.7773, 4.1627, 4.2063], device='cuda:5'), covar=tensor([0.1022, 0.0739, 0.1003, 0.0501, 0.0486, 0.1171, 0.0524, 0.0426], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0493, 0.0632, 0.0507, 0.0385, 0.0380, 0.0401, 0.0427], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:06:01,334 INFO [train.py:904] (5/8) Epoch 6, batch 7000, loss[loss=0.2352, simple_loss=0.329, pruned_loss=0.07068, over 17137.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.328, pruned_loss=0.08746, over 3083874.63 frames. ], batch size: 47, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:06:06,526 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:06:10,808 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7091, 3.7069, 3.8032, 3.7832, 3.7451, 4.1850, 3.9734, 3.7734], device='cuda:5'), covar=tensor([0.1808, 0.1863, 0.1831, 0.2035, 0.2696, 0.1454, 0.1136, 0.2295], device='cuda:5'), in_proj_covar=tensor([0.0295, 0.0399, 0.0404, 0.0353, 0.0460, 0.0437, 0.0333, 0.0477], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 13:06:38,242 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.231e+02 4.685e+02 6.535e+02 1.606e+03, threshold=9.370e+02, percent-clipped=8.0 2023-04-28 13:06:55,744 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3857, 4.2938, 4.8744, 4.8352, 4.8295, 4.4277, 4.4679, 4.2061], device='cuda:5'), covar=tensor([0.0278, 0.0428, 0.0332, 0.0398, 0.0384, 0.0307, 0.0818, 0.0438], device='cuda:5'), in_proj_covar=tensor([0.0260, 0.0255, 0.0262, 0.0259, 0.0308, 0.0277, 0.0380, 0.0227], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 13:07:04,828 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-28 13:07:07,591 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:07:17,567 INFO [train.py:904] (5/8) Epoch 6, batch 7050, loss[loss=0.2275, simple_loss=0.3053, pruned_loss=0.07484, over 16235.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3281, pruned_loss=0.08665, over 3086075.22 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:07:44,512 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:07,263 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-04-28 13:08:19,121 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7829, 1.2371, 1.5613, 1.6129, 1.7033, 1.8138, 1.3585, 1.6745], device='cuda:5'), covar=tensor([0.0101, 0.0184, 0.0102, 0.0117, 0.0100, 0.0074, 0.0187, 0.0051], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0148, 0.0129, 0.0128, 0.0133, 0.0099, 0.0145, 0.0086], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 13:08:20,874 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:34,158 INFO [train.py:904] (5/8) Epoch 6, batch 7100, loss[loss=0.2982, simple_loss=0.3437, pruned_loss=0.1263, over 11484.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.3269, pruned_loss=0.08682, over 3080116.87 frames. ], batch size: 247, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:08:35,324 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:37,939 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 13:08:42,773 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8616, 5.0968, 4.8330, 4.7995, 4.5798, 4.4570, 4.6158, 5.1459], device='cuda:5'), covar=tensor([0.0687, 0.0685, 0.0833, 0.0566, 0.0573, 0.0795, 0.0761, 0.0732], device='cuda:5'), in_proj_covar=tensor([0.0401, 0.0517, 0.0446, 0.0338, 0.0317, 0.0343, 0.0422, 0.0372], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:09:10,534 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 3.825e+02 4.803e+02 6.106e+02 1.425e+03, threshold=9.606e+02, percent-clipped=4.0 2023-04-28 13:09:32,898 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:48,241 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:49,606 INFO [train.py:904] (5/8) Epoch 6, batch 7150, loss[loss=0.2367, simple_loss=0.3242, pruned_loss=0.07459, over 16433.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.3248, pruned_loss=0.08643, over 3086388.18 frames. ], batch size: 146, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:09:49,971 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:10:12,199 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 13:10:17,519 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5206, 4.5601, 5.0965, 5.0554, 5.0255, 4.5580, 4.5805, 4.2225], device='cuda:5'), covar=tensor([0.0255, 0.0337, 0.0235, 0.0309, 0.0345, 0.0293, 0.0829, 0.0480], device='cuda:5'), in_proj_covar=tensor([0.0254, 0.0248, 0.0254, 0.0252, 0.0298, 0.0272, 0.0371, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 13:10:19,964 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1373, 3.4855, 3.6159, 1.6461, 3.8181, 3.8167, 2.9113, 2.8280], device='cuda:5'), covar=tensor([0.0848, 0.0140, 0.0135, 0.1228, 0.0050, 0.0080, 0.0334, 0.0403], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0091, 0.0082, 0.0140, 0.0073, 0.0081, 0.0115, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 13:10:35,971 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1292, 4.1101, 3.9980, 3.2705, 4.0012, 1.4835, 3.8201, 3.7738], device='cuda:5'), covar=tensor([0.0069, 0.0071, 0.0110, 0.0318, 0.0076, 0.2147, 0.0102, 0.0174], device='cuda:5'), in_proj_covar=tensor([0.0097, 0.0084, 0.0129, 0.0129, 0.0097, 0.0148, 0.0113, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:11:00,230 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57948.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:11:00,404 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3934, 5.3248, 5.1790, 5.0104, 4.6628, 5.2206, 5.1098, 4.9343], device='cuda:5'), covar=tensor([0.0405, 0.0219, 0.0183, 0.0153, 0.0914, 0.0281, 0.0194, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0211, 0.0220, 0.0190, 0.0248, 0.0226, 0.0159, 0.0254], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:11:04,898 INFO [train.py:904] (5/8) Epoch 6, batch 7200, loss[loss=0.2488, simple_loss=0.3164, pruned_loss=0.09058, over 11663.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3226, pruned_loss=0.08477, over 3059209.26 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:11:41,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.198e+02 3.384e+02 4.240e+02 5.535e+02 1.102e+03, threshold=8.480e+02, percent-clipped=1.0 2023-04-28 13:12:14,212 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7024, 2.1929, 1.6660, 1.9986, 2.5981, 2.3433, 2.8894, 2.8683], device='cuda:5'), covar=tensor([0.0053, 0.0178, 0.0271, 0.0240, 0.0122, 0.0190, 0.0088, 0.0101], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0161, 0.0165, 0.0163, 0.0157, 0.0166, 0.0146, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:12:28,072 INFO [train.py:904] (5/8) Epoch 6, batch 7250, loss[loss=0.2499, simple_loss=0.3181, pruned_loss=0.09086, over 15322.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3201, pruned_loss=0.08357, over 3049175.03 frames. ], batch size: 191, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:12:33,680 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4928, 3.4839, 3.4315, 2.8651, 3.3641, 2.0679, 3.1658, 2.8805], device='cuda:5'), covar=tensor([0.0090, 0.0065, 0.0111, 0.0195, 0.0060, 0.1543, 0.0093, 0.0137], device='cuda:5'), in_proj_covar=tensor([0.0096, 0.0083, 0.0129, 0.0128, 0.0096, 0.0147, 0.0112, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:13:18,670 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8107, 4.5506, 4.7905, 4.9807, 5.1342, 4.5528, 5.1002, 5.0638], device='cuda:5'), covar=tensor([0.1155, 0.0762, 0.1262, 0.0482, 0.0450, 0.0607, 0.0451, 0.0412], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0488, 0.0624, 0.0503, 0.0385, 0.0375, 0.0398, 0.0423], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:13:44,380 INFO [train.py:904] (5/8) Epoch 6, batch 7300, loss[loss=0.2462, simple_loss=0.3287, pruned_loss=0.08182, over 16510.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3199, pruned_loss=0.08308, over 3049971.32 frames. ], batch size: 75, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:13:49,562 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:14:21,944 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.373e+02 3.804e+02 4.653e+02 5.753e+02 1.061e+03, threshold=9.306e+02, percent-clipped=1.0 2023-04-28 13:14:43,470 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5395, 4.7939, 4.8368, 4.8039, 4.7883, 5.2902, 4.8186, 4.6388], device='cuda:5'), covar=tensor([0.0977, 0.1316, 0.1319, 0.1455, 0.1935, 0.0838, 0.1326, 0.2225], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0405, 0.0410, 0.0356, 0.0464, 0.0440, 0.0337, 0.0480], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 13:14:43,484 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:02,127 INFO [train.py:904] (5/8) Epoch 6, batch 7350, loss[loss=0.2378, simple_loss=0.3187, pruned_loss=0.07847, over 16175.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.319, pruned_loss=0.08262, over 3055470.01 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:15:03,788 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:28,238 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:36,729 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 13:15:44,789 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-28 13:16:04,923 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:09,176 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:19,326 INFO [train.py:904] (5/8) Epoch 6, batch 7400, loss[loss=0.2472, simple_loss=0.3283, pruned_loss=0.08307, over 16269.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3207, pruned_loss=0.08439, over 3042207.65 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:16:19,753 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:38,470 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0191, 3.7763, 3.8454, 2.5995, 3.4690, 3.7745, 3.5695, 2.0437], device='cuda:5'), covar=tensor([0.0378, 0.0018, 0.0024, 0.0236, 0.0047, 0.0070, 0.0040, 0.0305], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0055, 0.0059, 0.0116, 0.0063, 0.0075, 0.0065, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 13:16:43,943 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:57,167 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.315e+02 3.701e+02 4.363e+02 5.153e+02 1.099e+03, threshold=8.726e+02, percent-clipped=1.0 2023-04-28 13:17:13,475 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0405, 1.8568, 2.2565, 3.0566, 2.7557, 3.4172, 1.8217, 3.0909], device='cuda:5'), covar=tensor([0.0075, 0.0237, 0.0181, 0.0105, 0.0113, 0.0065, 0.0234, 0.0057], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0146, 0.0128, 0.0129, 0.0133, 0.0098, 0.0143, 0.0084], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 13:17:34,942 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:37,757 INFO [train.py:904] (5/8) Epoch 6, batch 7450, loss[loss=0.2727, simple_loss=0.3505, pruned_loss=0.09742, over 15224.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3225, pruned_loss=0.08572, over 3062401.31 frames. ], batch size: 190, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:17:39,834 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:45,764 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:18:59,288 INFO [train.py:904] (5/8) Epoch 6, batch 7500, loss[loss=0.2223, simple_loss=0.2971, pruned_loss=0.07373, over 17111.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3231, pruned_loss=0.08506, over 3062454.44 frames. ], batch size: 47, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:19:39,440 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.280e+02 3.913e+02 4.700e+02 5.850e+02 1.352e+03, threshold=9.401e+02, percent-clipped=2.0 2023-04-28 13:20:17,505 INFO [train.py:904] (5/8) Epoch 6, batch 7550, loss[loss=0.2885, simple_loss=0.3417, pruned_loss=0.1177, over 11537.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3213, pruned_loss=0.08449, over 3064013.08 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:21:03,970 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6509, 4.5643, 5.1763, 5.1175, 5.1121, 4.6936, 4.7517, 4.3474], device='cuda:5'), covar=tensor([0.0254, 0.0481, 0.0290, 0.0349, 0.0427, 0.0319, 0.0744, 0.0405], device='cuda:5'), in_proj_covar=tensor([0.0251, 0.0248, 0.0253, 0.0247, 0.0297, 0.0268, 0.0367, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 13:21:33,626 INFO [train.py:904] (5/8) Epoch 6, batch 7600, loss[loss=0.2443, simple_loss=0.3258, pruned_loss=0.08138, over 16247.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3206, pruned_loss=0.08506, over 3058567.80 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:22:14,299 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 3.777e+02 4.488e+02 5.757e+02 1.119e+03, threshold=8.975e+02, percent-clipped=2.0 2023-04-28 13:22:33,705 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:22:52,234 INFO [train.py:904] (5/8) Epoch 6, batch 7650, loss[loss=0.2373, simple_loss=0.3154, pruned_loss=0.0796, over 16606.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3215, pruned_loss=0.08576, over 3072772.86 frames. ], batch size: 57, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:23:37,782 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 13:23:49,829 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:24:11,093 INFO [train.py:904] (5/8) Epoch 6, batch 7700, loss[loss=0.2322, simple_loss=0.3126, pruned_loss=0.07594, over 16665.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3219, pruned_loss=0.08674, over 3060292.90 frames. ], batch size: 76, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:24:45,831 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 13:24:51,997 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 3.672e+02 4.554e+02 5.723e+02 8.873e+02, threshold=9.107e+02, percent-clipped=1.0 2023-04-28 13:25:01,028 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 13:25:01,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5644, 4.5357, 4.3959, 4.2620, 3.9385, 4.4533, 4.3549, 4.1281], device='cuda:5'), covar=tensor([0.0561, 0.0426, 0.0237, 0.0214, 0.0904, 0.0441, 0.0320, 0.0616], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0208, 0.0217, 0.0189, 0.0246, 0.0225, 0.0160, 0.0253], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:25:23,662 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:25:28,766 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58500.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:25:29,692 INFO [train.py:904] (5/8) Epoch 6, batch 7750, loss[loss=0.2508, simple_loss=0.3271, pruned_loss=0.08724, over 16734.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3219, pruned_loss=0.08614, over 3074711.80 frames. ], batch size: 134, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:26:09,024 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 13:26:46,074 INFO [train.py:904] (5/8) Epoch 6, batch 7800, loss[loss=0.2298, simple_loss=0.3114, pruned_loss=0.07409, over 16758.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.322, pruned_loss=0.08617, over 3089094.20 frames. ], batch size: 76, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:27:26,212 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.190e+02 3.955e+02 4.761e+02 5.914e+02 1.211e+03, threshold=9.523e+02, percent-clipped=2.0 2023-04-28 13:28:01,703 INFO [train.py:904] (5/8) Epoch 6, batch 7850, loss[loss=0.2293, simple_loss=0.3185, pruned_loss=0.07012, over 16823.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3233, pruned_loss=0.08647, over 3069881.60 frames. ], batch size: 90, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:28:12,051 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.11 vs. limit=5.0 2023-04-28 13:29:14,714 INFO [train.py:904] (5/8) Epoch 6, batch 7900, loss[loss=0.2456, simple_loss=0.3184, pruned_loss=0.08642, over 16866.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3223, pruned_loss=0.0853, over 3076608.92 frames. ], batch size: 116, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:29:53,311 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 3.363e+02 4.237e+02 5.568e+02 1.033e+03, threshold=8.474e+02, percent-clipped=1.0 2023-04-28 13:30:02,319 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:30:31,228 INFO [train.py:904] (5/8) Epoch 6, batch 7950, loss[loss=0.2195, simple_loss=0.2964, pruned_loss=0.07128, over 17184.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3224, pruned_loss=0.0855, over 3084780.09 frames. ], batch size: 46, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:31:33,558 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:31:46,452 INFO [train.py:904] (5/8) Epoch 6, batch 8000, loss[loss=0.2157, simple_loss=0.3002, pruned_loss=0.06556, over 17138.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.322, pruned_loss=0.08533, over 3084996.37 frames. ], batch size: 47, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:32:25,096 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58775.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:32:27,147 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.294e+02 3.253e+02 3.971e+02 4.697e+02 8.095e+02, threshold=7.943e+02, percent-clipped=0.0 2023-04-28 13:32:57,922 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:33:02,621 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58800.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:33:03,406 INFO [train.py:904] (5/8) Epoch 6, batch 8050, loss[loss=0.2292, simple_loss=0.3113, pruned_loss=0.07354, over 16770.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3215, pruned_loss=0.08519, over 3067261.38 frames. ], batch size: 124, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:33:58,090 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:13,205 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:14,435 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3075, 3.2164, 3.2968, 3.4455, 3.4506, 3.2067, 3.4323, 3.4658], device='cuda:5'), covar=tensor([0.0823, 0.0735, 0.1079, 0.0538, 0.0642, 0.1883, 0.0757, 0.0613], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0499, 0.0638, 0.0504, 0.0391, 0.0383, 0.0411, 0.0436], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:34:16,885 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:20,755 INFO [train.py:904] (5/8) Epoch 6, batch 8100, loss[loss=0.2749, simple_loss=0.3552, pruned_loss=0.09731, over 16345.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3216, pruned_loss=0.08541, over 3054898.77 frames. ], batch size: 146, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:34:31,938 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4525, 2.0507, 1.5913, 1.7788, 2.4206, 2.1721, 2.5222, 2.6230], device='cuda:5'), covar=tensor([0.0054, 0.0186, 0.0262, 0.0247, 0.0111, 0.0196, 0.0094, 0.0123], device='cuda:5'), in_proj_covar=tensor([0.0087, 0.0161, 0.0166, 0.0164, 0.0159, 0.0166, 0.0149, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:35:03,718 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.302e+02 3.498e+02 4.445e+02 5.357e+02 1.535e+03, threshold=8.891e+02, percent-clipped=10.0 2023-04-28 13:35:16,829 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5534, 5.8740, 5.5899, 5.6708, 5.1108, 4.9679, 5.4388, 5.9832], device='cuda:5'), covar=tensor([0.0757, 0.0714, 0.0865, 0.0484, 0.0672, 0.0595, 0.0624, 0.0691], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0521, 0.0448, 0.0337, 0.0322, 0.0346, 0.0426, 0.0374], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:35:38,308 INFO [train.py:904] (5/8) Epoch 6, batch 8150, loss[loss=0.2236, simple_loss=0.2986, pruned_loss=0.07429, over 16498.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3181, pruned_loss=0.08336, over 3071491.19 frames. ], batch size: 68, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:46,427 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:35:56,584 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 13:36:12,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7402, 4.7038, 4.5154, 3.8275, 4.5777, 1.6028, 4.3441, 4.4850], device='cuda:5'), covar=tensor([0.0052, 0.0051, 0.0095, 0.0304, 0.0053, 0.1898, 0.0087, 0.0129], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0085, 0.0132, 0.0129, 0.0099, 0.0152, 0.0115, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:36:24,296 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-28 13:36:54,139 INFO [train.py:904] (5/8) Epoch 6, batch 8200, loss[loss=0.2316, simple_loss=0.3167, pruned_loss=0.0732, over 15298.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3166, pruned_loss=0.08306, over 3073822.97 frames. ], batch size: 191, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:36:56,763 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9993, 2.4563, 2.3193, 3.0081, 2.4523, 3.2602, 1.7461, 2.6972], device='cuda:5'), covar=tensor([0.1103, 0.0401, 0.0982, 0.0104, 0.0190, 0.0351, 0.1192, 0.0674], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0145, 0.0172, 0.0097, 0.0200, 0.0192, 0.0167, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 13:37:21,775 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:37:40,738 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.295e+02 3.597e+02 4.719e+02 5.797e+02 1.131e+03, threshold=9.438e+02, percent-clipped=6.0 2023-04-28 13:38:06,484 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.82 vs. limit=5.0 2023-04-28 13:38:17,128 INFO [train.py:904] (5/8) Epoch 6, batch 8250, loss[loss=0.2171, simple_loss=0.2839, pruned_loss=0.07516, over 11827.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3156, pruned_loss=0.08084, over 3068228.90 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:39:07,652 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.26 vs. limit=5.0 2023-04-28 13:39:16,129 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:39:37,607 INFO [train.py:904] (5/8) Epoch 6, batch 8300, loss[loss=0.202, simple_loss=0.2803, pruned_loss=0.06183, over 12164.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3124, pruned_loss=0.07732, over 3064643.11 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:39:42,541 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9128, 3.7260, 3.3382, 5.0755, 4.2333, 4.9230, 1.8687, 3.5366], device='cuda:5'), covar=tensor([0.1442, 0.0509, 0.0868, 0.0098, 0.0268, 0.0275, 0.1466, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0140, 0.0166, 0.0093, 0.0191, 0.0185, 0.0162, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 13:40:10,824 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 13:40:22,350 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.786e+02 3.488e+02 4.487e+02 8.597e+02, threshold=6.977e+02, percent-clipped=0.0 2023-04-28 13:40:39,827 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 13:40:59,224 INFO [train.py:904] (5/8) Epoch 6, batch 8350, loss[loss=0.2207, simple_loss=0.3073, pruned_loss=0.06704, over 15225.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3101, pruned_loss=0.07449, over 3055558.48 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:41:04,626 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0696, 1.7259, 1.5611, 1.4753, 1.8204, 1.6270, 1.7510, 1.8668], device='cuda:5'), covar=tensor([0.0050, 0.0149, 0.0186, 0.0202, 0.0105, 0.0158, 0.0096, 0.0100], device='cuda:5'), in_proj_covar=tensor([0.0086, 0.0162, 0.0164, 0.0164, 0.0160, 0.0165, 0.0146, 0.0145], device='cuda:5'), out_proj_covar=tensor([9.9809e-05, 1.8778e-04, 1.8607e-04, 1.8649e-04, 1.8731e-04, 1.9180e-04, 1.6551e-04, 1.6762e-04], device='cuda:5') 2023-04-28 13:41:49,276 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:42:15,013 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1847, 4.1270, 4.0302, 3.4865, 4.0472, 1.6018, 3.8601, 3.8469], device='cuda:5'), covar=tensor([0.0068, 0.0061, 0.0104, 0.0251, 0.0070, 0.1981, 0.0097, 0.0147], device='cuda:5'), in_proj_covar=tensor([0.0095, 0.0082, 0.0129, 0.0124, 0.0097, 0.0150, 0.0111, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:42:21,186 INFO [train.py:904] (5/8) Epoch 6, batch 8400, loss[loss=0.2266, simple_loss=0.304, pruned_loss=0.07462, over 12073.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3062, pruned_loss=0.07112, over 3055172.96 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:42:42,356 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:42:42,408 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1767, 1.6790, 2.4173, 3.0942, 2.7648, 3.1355, 1.8975, 2.9953], device='cuda:5'), covar=tensor([0.0061, 0.0262, 0.0147, 0.0095, 0.0107, 0.0117, 0.0249, 0.0103], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0147, 0.0129, 0.0128, 0.0135, 0.0098, 0.0146, 0.0084], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 13:43:05,302 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.850e+02 3.501e+02 4.655e+02 7.905e+02, threshold=7.001e+02, percent-clipped=2.0 2023-04-28 13:43:40,735 INFO [train.py:904] (5/8) Epoch 6, batch 8450, loss[loss=0.2034, simple_loss=0.2934, pruned_loss=0.0567, over 16949.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3041, pruned_loss=0.06872, over 3073605.81 frames. ], batch size: 109, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:44:19,999 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:45:00,578 INFO [train.py:904] (5/8) Epoch 6, batch 8500, loss[loss=0.17, simple_loss=0.2474, pruned_loss=0.04629, over 11772.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2995, pruned_loss=0.06592, over 3065207.83 frames. ], batch size: 246, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:45:19,658 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:45:22,698 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.51 vs. limit=5.0 2023-04-28 13:45:46,052 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.033e+02 2.693e+02 3.373e+02 4.229e+02 1.183e+03, threshold=6.745e+02, percent-clipped=1.0 2023-04-28 13:46:25,538 INFO [train.py:904] (5/8) Epoch 6, batch 8550, loss[loss=0.212, simple_loss=0.2844, pruned_loss=0.06983, over 11858.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2969, pruned_loss=0.06446, over 3055924.48 frames. ], batch size: 246, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:47:38,391 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:47:41,392 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1223, 3.8989, 4.1212, 4.2955, 4.4505, 4.0027, 4.4244, 4.4031], device='cuda:5'), covar=tensor([0.1000, 0.0876, 0.1299, 0.0543, 0.0444, 0.0912, 0.0406, 0.0399], device='cuda:5'), in_proj_covar=tensor([0.0392, 0.0475, 0.0607, 0.0490, 0.0376, 0.0369, 0.0384, 0.0418], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:48:07,976 INFO [train.py:904] (5/8) Epoch 6, batch 8600, loss[loss=0.1957, simple_loss=0.2797, pruned_loss=0.05583, over 12712.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2974, pruned_loss=0.06389, over 3040917.48 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:49:03,382 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.797e+02 2.892e+02 3.852e+02 5.028e+02 1.593e+03, threshold=7.704e+02, percent-clipped=11.0 2023-04-28 13:49:14,847 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.58 vs. limit=5.0 2023-04-28 13:49:16,116 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:49:46,468 INFO [train.py:904] (5/8) Epoch 6, batch 8650, loss[loss=0.1992, simple_loss=0.2952, pruned_loss=0.05158, over 16240.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2956, pruned_loss=0.06234, over 3043625.12 frames. ], batch size: 165, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:50:55,835 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:51:31,176 INFO [train.py:904] (5/8) Epoch 6, batch 8700, loss[loss=0.2065, simple_loss=0.2959, pruned_loss=0.05857, over 15323.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2926, pruned_loss=0.06063, over 3042351.66 frames. ], batch size: 191, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:52:21,221 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 2.911e+02 3.776e+02 4.464e+02 8.360e+02, threshold=7.553e+02, percent-clipped=2.0 2023-04-28 13:52:22,279 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:05,575 INFO [train.py:904] (5/8) Epoch 6, batch 8750, loss[loss=0.2271, simple_loss=0.3177, pruned_loss=0.06829, over 16685.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2924, pruned_loss=0.06014, over 3036484.68 frames. ], batch size: 134, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:53:08,775 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:54,441 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:56,401 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1582, 1.3726, 1.7647, 2.0896, 2.0671, 2.1740, 1.5835, 2.0890], device='cuda:5'), covar=tensor([0.0110, 0.0294, 0.0166, 0.0146, 0.0155, 0.0121, 0.0268, 0.0069], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0148, 0.0131, 0.0128, 0.0137, 0.0096, 0.0146, 0.0082], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 13:54:55,925 INFO [train.py:904] (5/8) Epoch 6, batch 8800, loss[loss=0.1814, simple_loss=0.2794, pruned_loss=0.04174, over 16895.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2904, pruned_loss=0.05853, over 3042399.10 frames. ], batch size: 96, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:55:17,687 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:20,046 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:52,198 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.649e+02 3.204e+02 3.980e+02 7.645e+02, threshold=6.408e+02, percent-clipped=1.0 2023-04-28 13:56:37,977 INFO [train.py:904] (5/8) Epoch 6, batch 8850, loss[loss=0.1957, simple_loss=0.29, pruned_loss=0.05064, over 15252.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2925, pruned_loss=0.05768, over 3029620.55 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:56:56,205 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:57:32,563 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 13:57:40,478 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4366, 1.4206, 1.7883, 2.4678, 2.2397, 2.4229, 1.6375, 2.3774], device='cuda:5'), covar=tensor([0.0091, 0.0314, 0.0176, 0.0132, 0.0153, 0.0101, 0.0262, 0.0077], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0147, 0.0130, 0.0127, 0.0135, 0.0094, 0.0144, 0.0081], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 13:58:21,078 INFO [train.py:904] (5/8) Epoch 6, batch 8900, loss[loss=0.201, simple_loss=0.29, pruned_loss=0.05604, over 17005.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2922, pruned_loss=0.05738, over 3020864.70 frames. ], batch size: 109, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:58:24,474 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9097, 1.9392, 2.2014, 3.2763, 1.9826, 2.4088, 2.2086, 1.9603], device='cuda:5'), covar=tensor([0.0626, 0.2402, 0.1326, 0.0367, 0.3078, 0.1497, 0.2044, 0.2718], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0327, 0.0274, 0.0300, 0.0380, 0.0341, 0.0299, 0.0386], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 13:59:05,308 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3756, 3.5712, 1.6236, 3.7497, 2.2959, 3.6722, 1.8496, 2.6550], device='cuda:5'), covar=tensor([0.0158, 0.0212, 0.1662, 0.0055, 0.0877, 0.0385, 0.1516, 0.0615], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0145, 0.0175, 0.0081, 0.0155, 0.0178, 0.0185, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 13:59:22,390 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.981e+02 2.897e+02 3.325e+02 4.292e+02 6.901e+02, threshold=6.651e+02, percent-clipped=1.0 2023-04-28 13:59:47,744 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 14:00:23,145 INFO [train.py:904] (5/8) Epoch 6, batch 8950, loss[loss=0.1864, simple_loss=0.2732, pruned_loss=0.04979, over 15366.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2921, pruned_loss=0.0576, over 3014915.42 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:02:00,485 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 14:02:12,312 INFO [train.py:904] (5/8) Epoch 6, batch 9000, loss[loss=0.2008, simple_loss=0.281, pruned_loss=0.06031, over 16513.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2888, pruned_loss=0.05603, over 3017324.04 frames. ], batch size: 147, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:02:12,312 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 14:02:22,191 INFO [train.py:938] (5/8) Epoch 6, validation: loss=0.1682, simple_loss=0.2716, pruned_loss=0.03235, over 944034.00 frames. 2023-04-28 14:02:22,192 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 14:03:21,586 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.787e+02 3.475e+02 4.351e+02 1.064e+03, threshold=6.950e+02, percent-clipped=4.0 2023-04-28 14:04:06,173 INFO [train.py:904] (5/8) Epoch 6, batch 9050, loss[loss=0.1755, simple_loss=0.2667, pruned_loss=0.04217, over 17113.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2904, pruned_loss=0.05693, over 3030683.01 frames. ], batch size: 49, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:04:44,710 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:05:24,271 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1098, 4.1004, 4.2354, 4.1381, 4.1430, 4.7017, 4.4231, 4.1310], device='cuda:5'), covar=tensor([0.1477, 0.1825, 0.1886, 0.2000, 0.2742, 0.1150, 0.1195, 0.2294], device='cuda:5'), in_proj_covar=tensor([0.0274, 0.0384, 0.0392, 0.0335, 0.0440, 0.0422, 0.0319, 0.0453], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:05:45,907 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:05:50,595 INFO [train.py:904] (5/8) Epoch 6, batch 9100, loss[loss=0.1922, simple_loss=0.2732, pruned_loss=0.05554, over 12104.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2899, pruned_loss=0.05703, over 3043939.12 frames. ], batch size: 247, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:06:04,403 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:24,657 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:52,683 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4298, 1.4727, 1.8496, 2.4110, 2.2881, 2.5297, 1.4596, 2.4964], device='cuda:5'), covar=tensor([0.0135, 0.0293, 0.0197, 0.0145, 0.0154, 0.0112, 0.0326, 0.0081], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0150, 0.0132, 0.0131, 0.0139, 0.0097, 0.0149, 0.0083], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 14:06:55,733 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.998e+02 3.675e+02 4.654e+02 1.190e+03, threshold=7.350e+02, percent-clipped=4.0 2023-04-28 14:07:14,459 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2006, 3.3284, 1.6859, 3.5216, 2.2844, 3.4463, 1.7945, 2.6046], device='cuda:5'), covar=tensor([0.0216, 0.0257, 0.1548, 0.0074, 0.0834, 0.0388, 0.1475, 0.0599], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0146, 0.0174, 0.0081, 0.0154, 0.0178, 0.0185, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 14:07:19,882 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:07:47,019 INFO [train.py:904] (5/8) Epoch 6, batch 9150, loss[loss=0.1744, simple_loss=0.2696, pruned_loss=0.03959, over 16879.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2905, pruned_loss=0.05648, over 3050394.45 frames. ], batch size: 96, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:08:06,831 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:09:30,402 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:09:30,687 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 14:09:31,028 INFO [train.py:904] (5/8) Epoch 6, batch 9200, loss[loss=0.1997, simple_loss=0.2843, pruned_loss=0.0575, over 16836.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.286, pruned_loss=0.05525, over 3067601.75 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:10:22,280 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 2.993e+02 3.687e+02 4.643e+02 6.569e+02, threshold=7.374e+02, percent-clipped=0.0 2023-04-28 14:11:09,076 INFO [train.py:904] (5/8) Epoch 6, batch 9250, loss[loss=0.184, simple_loss=0.2743, pruned_loss=0.04682, over 16748.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2854, pruned_loss=0.05512, over 3066204.86 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:12:34,014 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:12:58,489 INFO [train.py:904] (5/8) Epoch 6, batch 9300, loss[loss=0.187, simple_loss=0.2627, pruned_loss=0.05564, over 12480.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2835, pruned_loss=0.05428, over 3054351.87 frames. ], batch size: 246, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:13:01,520 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:13:33,253 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:14:01,688 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:14:05,092 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.894e+02 3.605e+02 4.509e+02 1.051e+03, threshold=7.210e+02, percent-clipped=5.0 2023-04-28 14:14:23,784 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 14:14:34,258 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4856, 4.2262, 4.0124, 2.1557, 3.1981, 2.4154, 3.7893, 3.8720], device='cuda:5'), covar=tensor([0.0203, 0.0501, 0.0382, 0.1436, 0.0661, 0.0941, 0.0561, 0.0673], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0121, 0.0151, 0.0139, 0.0132, 0.0125, 0.0136, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 14:14:44,856 INFO [train.py:904] (5/8) Epoch 6, batch 9350, loss[loss=0.2016, simple_loss=0.2899, pruned_loss=0.05667, over 16953.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2836, pruned_loss=0.05476, over 3051408.45 frames. ], batch size: 109, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:14:46,382 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 14:15:13,213 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 14:15:37,242 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:15:48,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5537, 3.2356, 3.0881, 1.7605, 2.7310, 2.1529, 3.0135, 3.1414], device='cuda:5'), covar=tensor([0.0290, 0.0573, 0.0507, 0.1641, 0.0726, 0.0976, 0.0739, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0122, 0.0151, 0.0139, 0.0132, 0.0125, 0.0136, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 14:16:03,907 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:16:05,591 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-04-28 14:16:11,506 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9870, 3.1052, 3.1030, 2.3264, 2.8476, 3.0012, 3.0171, 1.8187], device='cuda:5'), covar=tensor([0.0315, 0.0019, 0.0028, 0.0208, 0.0065, 0.0055, 0.0048, 0.0317], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0053, 0.0056, 0.0113, 0.0062, 0.0071, 0.0063, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 14:16:25,957 INFO [train.py:904] (5/8) Epoch 6, batch 9400, loss[loss=0.1675, simple_loss=0.2544, pruned_loss=0.0403, over 12379.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2846, pruned_loss=0.05468, over 3051870.65 frames. ], batch size: 246, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:16:41,498 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:17:25,097 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 3.031e+02 3.474e+02 4.387e+02 1.018e+03, threshold=6.948e+02, percent-clipped=2.0 2023-04-28 14:17:31,345 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2296, 3.2778, 1.5581, 3.5152, 2.1646, 3.4042, 1.7740, 2.5430], device='cuda:5'), covar=tensor([0.0219, 0.0355, 0.1835, 0.0091, 0.0938, 0.0522, 0.1740, 0.0772], device='cuda:5'), in_proj_covar=tensor([0.0121, 0.0147, 0.0174, 0.0082, 0.0155, 0.0178, 0.0186, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 14:17:55,029 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3806, 2.9787, 2.6438, 2.3344, 2.1602, 2.1469, 2.9833, 2.9774], device='cuda:5'), covar=tensor([0.1855, 0.0640, 0.1102, 0.1518, 0.1839, 0.1446, 0.0344, 0.0681], device='cuda:5'), in_proj_covar=tensor([0.0273, 0.0238, 0.0261, 0.0244, 0.0241, 0.0199, 0.0237, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:18:05,576 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:09,623 INFO [train.py:904] (5/8) Epoch 6, batch 9450, loss[loss=0.2011, simple_loss=0.29, pruned_loss=0.05609, over 17201.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2862, pruned_loss=0.05516, over 3055113.43 frames. ], batch size: 45, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:18:15,340 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:19,419 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:26,164 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9677, 2.4410, 2.2721, 3.1384, 2.4895, 3.3684, 1.6507, 2.8051], device='cuda:5'), covar=tensor([0.1188, 0.0468, 0.0899, 0.0105, 0.0163, 0.0383, 0.1235, 0.0597], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0142, 0.0166, 0.0093, 0.0166, 0.0186, 0.0163, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 14:19:38,361 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2934, 4.1901, 4.1427, 3.6458, 4.0951, 1.5638, 3.9273, 3.9191], device='cuda:5'), covar=tensor([0.0059, 0.0058, 0.0096, 0.0218, 0.0068, 0.1955, 0.0088, 0.0147], device='cuda:5'), in_proj_covar=tensor([0.0093, 0.0079, 0.0123, 0.0114, 0.0094, 0.0148, 0.0109, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:19:39,948 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:19:50,605 INFO [train.py:904] (5/8) Epoch 6, batch 9500, loss[loss=0.1885, simple_loss=0.2821, pruned_loss=0.04745, over 16844.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2853, pruned_loss=0.05455, over 3056270.63 frames. ], batch size: 102, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:20:11,484 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:20:50,226 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-28 14:20:51,256 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.861e+02 2.773e+02 3.351e+02 4.137e+02 1.111e+03, threshold=6.703e+02, percent-clipped=2.0 2023-04-28 14:21:39,664 INFO [train.py:904] (5/8) Epoch 6, batch 9550, loss[loss=0.2079, simple_loss=0.2913, pruned_loss=0.06227, over 12484.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2848, pruned_loss=0.05456, over 3062648.39 frames. ], batch size: 247, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:23:21,511 INFO [train.py:904] (5/8) Epoch 6, batch 9600, loss[loss=0.2161, simple_loss=0.2936, pruned_loss=0.06937, over 12606.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.287, pruned_loss=0.05588, over 3061477.62 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:23:27,197 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2493, 4.5819, 4.9570, 4.9635, 4.9161, 4.5357, 4.2077, 4.1715], device='cuda:5'), covar=tensor([0.0606, 0.0425, 0.0498, 0.0558, 0.0629, 0.0507, 0.1289, 0.0537], device='cuda:5'), in_proj_covar=tensor([0.0237, 0.0232, 0.0239, 0.0231, 0.0274, 0.0253, 0.0340, 0.0206], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 14:24:17,217 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.768e+02 3.406e+02 4.197e+02 1.161e+03, threshold=6.812e+02, percent-clipped=3.0 2023-04-28 14:24:57,311 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 14:25:09,708 INFO [train.py:904] (5/8) Epoch 6, batch 9650, loss[loss=0.2089, simple_loss=0.2861, pruned_loss=0.06588, over 12468.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2887, pruned_loss=0.05631, over 3055831.72 frames. ], batch size: 250, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:25:28,087 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 14:25:57,587 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:13,629 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1598, 3.2550, 3.6105, 3.5709, 3.5561, 3.3126, 3.3862, 3.3778], device='cuda:5'), covar=tensor([0.0326, 0.0539, 0.0359, 0.0406, 0.0457, 0.0402, 0.0703, 0.0363], device='cuda:5'), in_proj_covar=tensor([0.0242, 0.0237, 0.0243, 0.0236, 0.0281, 0.0258, 0.0349, 0.0210], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 14:26:22,175 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:58,006 INFO [train.py:904] (5/8) Epoch 6, batch 9700, loss[loss=0.2079, simple_loss=0.2923, pruned_loss=0.06179, over 16955.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2875, pruned_loss=0.0562, over 3045057.15 frames. ], batch size: 109, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:27:59,429 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.972e+02 2.947e+02 3.794e+02 4.727e+02 9.838e+02, threshold=7.588e+02, percent-clipped=5.0 2023-04-28 14:28:41,238 INFO [train.py:904] (5/8) Epoch 6, batch 9750, loss[loss=0.1804, simple_loss=0.2767, pruned_loss=0.04209, over 16867.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2861, pruned_loss=0.05612, over 3030871.43 frames. ], batch size: 90, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:28:47,807 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:11,616 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:19,580 INFO [train.py:904] (5/8) Epoch 6, batch 9800, loss[loss=0.2163, simple_loss=0.3119, pruned_loss=0.06041, over 16635.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2862, pruned_loss=0.05462, over 3057071.82 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:30:22,403 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:27,295 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:31:14,372 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.802e+02 3.388e+02 4.163e+02 9.547e+02, threshold=6.775e+02, percent-clipped=1.0 2023-04-28 14:31:33,106 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9073, 3.9753, 3.7997, 3.6777, 3.5386, 3.9061, 3.6052, 3.6510], device='cuda:5'), covar=tensor([0.0509, 0.0311, 0.0221, 0.0181, 0.0656, 0.0298, 0.0730, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0199, 0.0212, 0.0185, 0.0235, 0.0215, 0.0150, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:31:48,143 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:32:03,907 INFO [train.py:904] (5/8) Epoch 6, batch 9850, loss[loss=0.2292, simple_loss=0.3108, pruned_loss=0.07378, over 16944.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2877, pruned_loss=0.05456, over 3066656.85 frames. ], batch size: 109, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:33:38,904 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 14:33:56,864 INFO [train.py:904] (5/8) Epoch 6, batch 9900, loss[loss=0.2162, simple_loss=0.3033, pruned_loss=0.06452, over 16947.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2874, pruned_loss=0.05389, over 3059318.32 frames. ], batch size: 109, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:35:04,552 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.757e+02 3.371e+02 4.013e+02 6.843e+02, threshold=6.741e+02, percent-clipped=1.0 2023-04-28 14:35:19,024 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 14:35:42,542 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:35:52,859 INFO [train.py:904] (5/8) Epoch 6, batch 9950, loss[loss=0.1924, simple_loss=0.2901, pruned_loss=0.04732, over 16865.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2897, pruned_loss=0.05411, over 3074056.42 frames. ], batch size: 124, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:36:10,192 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:29,197 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7640, 3.4021, 2.7369, 5.0566, 4.2825, 4.6624, 1.5889, 3.2056], device='cuda:5'), covar=tensor([0.1270, 0.0474, 0.1003, 0.0053, 0.0107, 0.0220, 0.1327, 0.0674], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0139, 0.0165, 0.0090, 0.0156, 0.0184, 0.0160, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 14:36:34,865 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:43,582 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:13,880 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:21,117 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-04-28 14:37:38,497 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60744.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:54,662 INFO [train.py:904] (5/8) Epoch 6, batch 10000, loss[loss=0.2122, simple_loss=0.3018, pruned_loss=0.06129, over 15322.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2873, pruned_loss=0.05296, over 3101764.74 frames. ], batch size: 190, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:38:06,645 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:29,854 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:50,655 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:51,224 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.599e+02 3.122e+02 4.143e+02 8.746e+02, threshold=6.243e+02, percent-clipped=4.0 2023-04-28 14:38:52,650 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6946, 1.9643, 1.4076, 1.6733, 2.3366, 2.0758, 2.6109, 2.5826], device='cuda:5'), covar=tensor([0.0039, 0.0228, 0.0339, 0.0288, 0.0145, 0.0228, 0.0100, 0.0117], device='cuda:5'), in_proj_covar=tensor([0.0082, 0.0162, 0.0162, 0.0161, 0.0158, 0.0161, 0.0139, 0.0143], device='cuda:5'), out_proj_covar=tensor([9.3449e-05, 1.8742e-04, 1.8285e-04, 1.8114e-04, 1.8409e-04, 1.8567e-04, 1.5341e-04, 1.6451e-04], device='cuda:5') 2023-04-28 14:38:57,420 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:39:12,744 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5639, 4.6061, 4.4094, 4.2355, 3.9891, 4.5057, 4.4312, 4.1957], device='cuda:5'), covar=tensor([0.0527, 0.0427, 0.0266, 0.0206, 0.0898, 0.0363, 0.0304, 0.0498], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0201, 0.0215, 0.0187, 0.0236, 0.0217, 0.0151, 0.0239], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:39:33,420 INFO [train.py:904] (5/8) Epoch 6, batch 10050, loss[loss=0.2306, simple_loss=0.3194, pruned_loss=0.07093, over 16411.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2883, pruned_loss=0.05343, over 3086873.39 frames. ], batch size: 146, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:41:05,285 INFO [train.py:904] (5/8) Epoch 6, batch 10100, loss[loss=0.1967, simple_loss=0.2826, pruned_loss=0.05545, over 16226.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2885, pruned_loss=0.05365, over 3084066.04 frames. ], batch size: 165, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:41:11,006 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:41:21,643 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 14:42:01,267 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.889e+02 3.366e+02 3.931e+02 7.986e+02, threshold=6.733e+02, percent-clipped=7.0 2023-04-28 14:42:04,910 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8444, 1.8033, 2.2231, 3.0668, 1.8995, 2.1087, 2.0802, 1.7251], device='cuda:5'), covar=tensor([0.0808, 0.3109, 0.1410, 0.0523, 0.3873, 0.1936, 0.2576, 0.3562], device='cuda:5'), in_proj_covar=tensor([0.0309, 0.0320, 0.0274, 0.0298, 0.0367, 0.0336, 0.0294, 0.0373], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:42:22,739 INFO [train.py:904] (5/8) Epoch 6, batch 10150, loss[loss=0.2223, simple_loss=0.2894, pruned_loss=0.07756, over 12443.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2867, pruned_loss=0.05371, over 3066214.16 frames. ], batch size: 250, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:42:48,316 INFO [train.py:904] (5/8) Epoch 7, batch 0, loss[loss=0.2173, simple_loss=0.2964, pruned_loss=0.06913, over 17221.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2964, pruned_loss=0.06913, over 17221.00 frames. ], batch size: 45, lr: 1.02e-02, grad_scale: 8.0 2023-04-28 14:42:48,316 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 14:42:55,783 INFO [train.py:938] (5/8) Epoch 7, validation: loss=0.1665, simple_loss=0.2702, pruned_loss=0.03141, over 944034.00 frames. 2023-04-28 14:42:55,784 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 14:42:55,984 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:42:57,993 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2853, 5.2285, 5.0887, 4.6680, 5.0177, 1.8609, 4.9128, 5.1302], device='cuda:5'), covar=tensor([0.0048, 0.0043, 0.0095, 0.0166, 0.0047, 0.1805, 0.0069, 0.0089], device='cuda:5'), in_proj_covar=tensor([0.0094, 0.0080, 0.0125, 0.0112, 0.0093, 0.0149, 0.0109, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:44:05,499 INFO [train.py:904] (5/8) Epoch 7, batch 50, loss[loss=0.2338, simple_loss=0.3021, pruned_loss=0.08269, over 16819.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3057, pruned_loss=0.07777, over 760651.50 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:44:49,474 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 3.058e+02 3.620e+02 4.703e+02 1.122e+03, threshold=7.241e+02, percent-clipped=6.0 2023-04-28 14:45:10,371 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4977, 4.5391, 4.6488, 4.6543, 4.5461, 5.1931, 4.8636, 4.5610], device='cuda:5'), covar=tensor([0.1321, 0.1825, 0.1990, 0.1998, 0.3234, 0.1212, 0.1486, 0.2668], device='cuda:5'), in_proj_covar=tensor([0.0291, 0.0408, 0.0413, 0.0350, 0.0467, 0.0445, 0.0339, 0.0472], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 14:45:15,349 INFO [train.py:904] (5/8) Epoch 7, batch 100, loss[loss=0.2289, simple_loss=0.3109, pruned_loss=0.07346, over 17066.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.2983, pruned_loss=0.07271, over 1325022.58 frames. ], batch size: 53, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:45:19,400 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0646, 4.9727, 4.8692, 4.2897, 4.8589, 1.9648, 4.6343, 4.8268], device='cuda:5'), covar=tensor([0.0048, 0.0050, 0.0094, 0.0238, 0.0060, 0.1773, 0.0088, 0.0116], device='cuda:5'), in_proj_covar=tensor([0.0098, 0.0083, 0.0132, 0.0119, 0.0098, 0.0154, 0.0114, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:46:24,675 INFO [train.py:904] (5/8) Epoch 7, batch 150, loss[loss=0.2256, simple_loss=0.302, pruned_loss=0.07463, over 15494.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2955, pruned_loss=0.07075, over 1769453.37 frames. ], batch size: 190, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:46:25,066 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1998, 2.9605, 3.6250, 2.4044, 3.3923, 3.5472, 3.5144, 2.1574], device='cuda:5'), covar=tensor([0.0289, 0.0121, 0.0027, 0.0217, 0.0045, 0.0051, 0.0040, 0.0267], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0058, 0.0059, 0.0116, 0.0062, 0.0073, 0.0065, 0.0112], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 14:46:38,007 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61061.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:46:56,471 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:47:07,164 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 2.994e+02 3.585e+02 4.192e+02 7.211e+02, threshold=7.170e+02, percent-clipped=0.0 2023-04-28 14:47:34,208 INFO [train.py:904] (5/8) Epoch 7, batch 200, loss[loss=0.1813, simple_loss=0.2721, pruned_loss=0.04521, over 17201.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.2968, pruned_loss=0.07182, over 2110860.00 frames. ], batch size: 46, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:48:01,662 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:48:44,073 INFO [train.py:904] (5/8) Epoch 7, batch 250, loss[loss=0.2156, simple_loss=0.3078, pruned_loss=0.06168, over 17049.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2943, pruned_loss=0.07077, over 2382058.51 frames. ], batch size: 53, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:48:50,878 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0787, 4.7319, 4.9882, 5.2428, 5.4865, 4.7859, 5.3811, 5.3746], device='cuda:5'), covar=tensor([0.1252, 0.0926, 0.1540, 0.0578, 0.0461, 0.0626, 0.0457, 0.0399], device='cuda:5'), in_proj_covar=tensor([0.0428, 0.0525, 0.0657, 0.0523, 0.0398, 0.0396, 0.0417, 0.0449], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:49:24,843 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 3.071e+02 3.589e+02 4.451e+02 1.172e+03, threshold=7.179e+02, percent-clipped=5.0 2023-04-28 14:49:28,761 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5965, 5.9546, 5.6521, 5.7724, 5.1510, 5.1744, 5.4492, 6.0723], device='cuda:5'), covar=tensor([0.0799, 0.0735, 0.1041, 0.0526, 0.0740, 0.0532, 0.0674, 0.0765], device='cuda:5'), in_proj_covar=tensor([0.0431, 0.0561, 0.0468, 0.0369, 0.0348, 0.0367, 0.0465, 0.0409], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:49:32,043 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7847, 4.1583, 2.1434, 4.3505, 2.7685, 4.4296, 2.1668, 3.0454], device='cuda:5'), covar=tensor([0.0163, 0.0250, 0.1470, 0.0071, 0.0847, 0.0356, 0.1491, 0.0648], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0155, 0.0180, 0.0089, 0.0158, 0.0191, 0.0190, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 14:49:51,556 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-28 14:49:52,030 INFO [train.py:904] (5/8) Epoch 7, batch 300, loss[loss=0.2534, simple_loss=0.3045, pruned_loss=0.1011, over 16852.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2909, pruned_loss=0.0691, over 2586722.35 frames. ], batch size: 116, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:50:14,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7527, 4.0577, 4.2742, 2.0817, 4.5673, 4.5279, 3.0628, 3.3049], device='cuda:5'), covar=tensor([0.0724, 0.0136, 0.0134, 0.1086, 0.0056, 0.0078, 0.0407, 0.0369], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0092, 0.0081, 0.0141, 0.0069, 0.0086, 0.0118, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 14:50:59,338 INFO [train.py:904] (5/8) Epoch 7, batch 350, loss[loss=0.1866, simple_loss=0.2622, pruned_loss=0.05551, over 16496.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.287, pruned_loss=0.0666, over 2756604.25 frames. ], batch size: 75, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:51:01,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8198, 3.8623, 3.0837, 2.4771, 2.7482, 2.1787, 3.9422, 3.7551], device='cuda:5'), covar=tensor([0.1954, 0.0528, 0.1144, 0.1718, 0.2204, 0.1575, 0.0416, 0.0879], device='cuda:5'), in_proj_covar=tensor([0.0280, 0.0249, 0.0267, 0.0252, 0.0261, 0.0206, 0.0248, 0.0263], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:51:41,730 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.825e+02 3.324e+02 4.016e+02 8.512e+02, threshold=6.649e+02, percent-clipped=2.0 2023-04-28 14:51:50,647 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9682, 1.7759, 2.2113, 2.8086, 2.5785, 3.3496, 1.9629, 3.0917], device='cuda:5'), covar=tensor([0.0106, 0.0251, 0.0187, 0.0158, 0.0145, 0.0098, 0.0246, 0.0083], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0150, 0.0134, 0.0134, 0.0137, 0.0100, 0.0146, 0.0085], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 14:52:08,575 INFO [train.py:904] (5/8) Epoch 7, batch 400, loss[loss=0.2016, simple_loss=0.2839, pruned_loss=0.05964, over 16763.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2858, pruned_loss=0.06644, over 2892363.52 frames. ], batch size: 62, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:52:15,280 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-28 14:52:27,166 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7587, 2.7229, 2.4224, 3.5726, 2.9537, 3.6827, 1.4629, 2.8145], device='cuda:5'), covar=tensor([0.1220, 0.0478, 0.0985, 0.0123, 0.0230, 0.0330, 0.1295, 0.0715], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0143, 0.0168, 0.0097, 0.0176, 0.0192, 0.0163, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 14:52:32,467 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8015, 1.7453, 2.2009, 2.7264, 2.7622, 2.7286, 1.7432, 2.8924], device='cuda:5'), covar=tensor([0.0091, 0.0249, 0.0196, 0.0142, 0.0115, 0.0137, 0.0251, 0.0069], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0152, 0.0135, 0.0135, 0.0138, 0.0101, 0.0147, 0.0085], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 14:53:09,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4464, 2.0163, 2.2254, 3.9881, 1.9877, 2.6969, 2.1439, 2.1935], device='cuda:5'), covar=tensor([0.0657, 0.2563, 0.1473, 0.0368, 0.2930, 0.1514, 0.2360, 0.2372], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0346, 0.0288, 0.0318, 0.0387, 0.0369, 0.0312, 0.0410], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:53:17,210 INFO [train.py:904] (5/8) Epoch 7, batch 450, loss[loss=0.2083, simple_loss=0.2947, pruned_loss=0.06097, over 16993.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2833, pruned_loss=0.06457, over 2994285.06 frames. ], batch size: 53, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:53:48,656 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:53:49,915 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8972, 4.3783, 3.2731, 2.3524, 3.0558, 2.5657, 4.5431, 4.0414], device='cuda:5'), covar=tensor([0.2207, 0.0518, 0.1242, 0.1843, 0.2180, 0.1449, 0.0347, 0.0765], device='cuda:5'), in_proj_covar=tensor([0.0283, 0.0252, 0.0272, 0.0255, 0.0268, 0.0209, 0.0250, 0.0267], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:54:00,438 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.768e+02 3.236e+02 4.177e+02 7.606e+02, threshold=6.472e+02, percent-clipped=3.0 2023-04-28 14:54:27,760 INFO [train.py:904] (5/8) Epoch 7, batch 500, loss[loss=0.1695, simple_loss=0.2552, pruned_loss=0.0419, over 17213.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2821, pruned_loss=0.06346, over 3066494.74 frames. ], batch size: 44, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:54:47,755 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61417.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:54:54,990 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:55:35,823 INFO [train.py:904] (5/8) Epoch 7, batch 550, loss[loss=0.2113, simple_loss=0.2773, pruned_loss=0.07262, over 16806.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2812, pruned_loss=0.06339, over 3115148.76 frames. ], batch size: 102, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:56:17,396 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.888e+02 3.437e+02 4.263e+02 9.933e+02, threshold=6.875e+02, percent-clipped=6.0 2023-04-28 14:56:43,958 INFO [train.py:904] (5/8) Epoch 7, batch 600, loss[loss=0.2259, simple_loss=0.2817, pruned_loss=0.08506, over 16860.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.281, pruned_loss=0.06394, over 3156296.54 frames. ], batch size: 116, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:56:46,180 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 14:57:27,423 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-04-28 14:57:36,572 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0506, 4.7588, 4.9980, 5.2488, 5.4403, 4.8248, 5.4122, 5.4111], device='cuda:5'), covar=tensor([0.1161, 0.1011, 0.1561, 0.0534, 0.0423, 0.0635, 0.0367, 0.0416], device='cuda:5'), in_proj_covar=tensor([0.0452, 0.0549, 0.0694, 0.0547, 0.0423, 0.0413, 0.0437, 0.0474], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:57:39,379 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 14:57:53,649 INFO [train.py:904] (5/8) Epoch 7, batch 650, loss[loss=0.1914, simple_loss=0.2625, pruned_loss=0.06016, over 16759.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2798, pruned_loss=0.06342, over 3194068.45 frames. ], batch size: 102, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:58:02,510 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1832, 3.5493, 3.3098, 2.0474, 2.8219, 2.1816, 3.5286, 3.5198], device='cuda:5'), covar=tensor([0.0207, 0.0579, 0.0529, 0.1564, 0.0698, 0.1019, 0.0444, 0.0814], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0129, 0.0154, 0.0140, 0.0132, 0.0124, 0.0137, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 14:58:31,633 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5977, 4.3399, 4.5413, 4.7791, 4.9427, 4.4750, 4.8268, 4.8932], device='cuda:5'), covar=tensor([0.1182, 0.1026, 0.1524, 0.0680, 0.0535, 0.0774, 0.0936, 0.0590], device='cuda:5'), in_proj_covar=tensor([0.0450, 0.0551, 0.0693, 0.0548, 0.0421, 0.0412, 0.0439, 0.0474], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 14:58:35,765 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.778e+02 2.890e+02 3.388e+02 4.314e+02 1.492e+03, threshold=6.776e+02, percent-clipped=7.0 2023-04-28 14:59:01,945 INFO [train.py:904] (5/8) Epoch 7, batch 700, loss[loss=0.2378, simple_loss=0.3074, pruned_loss=0.08411, over 16462.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2793, pruned_loss=0.06276, over 3223783.53 frames. ], batch size: 146, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:59:47,335 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4429, 3.3883, 2.7133, 2.2560, 2.3886, 2.1104, 3.3728, 3.3034], device='cuda:5'), covar=tensor([0.2141, 0.0682, 0.1255, 0.1750, 0.2262, 0.1618, 0.0519, 0.0932], device='cuda:5'), in_proj_covar=tensor([0.0285, 0.0255, 0.0273, 0.0256, 0.0275, 0.0211, 0.0252, 0.0272], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:00:08,136 INFO [train.py:904] (5/8) Epoch 7, batch 750, loss[loss=0.1966, simple_loss=0.2718, pruned_loss=0.06067, over 16809.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.28, pruned_loss=0.06329, over 3240750.38 frames. ], batch size: 39, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 15:00:51,556 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.695e+02 2.818e+02 3.177e+02 3.782e+02 5.959e+02, threshold=6.355e+02, percent-clipped=0.0 2023-04-28 15:01:16,016 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3035, 4.2062, 4.2376, 4.3472, 4.2951, 4.9126, 4.4870, 4.1922], device='cuda:5'), covar=tensor([0.1515, 0.1967, 0.1760, 0.1906, 0.3123, 0.1176, 0.1331, 0.2493], device='cuda:5'), in_proj_covar=tensor([0.0308, 0.0442, 0.0439, 0.0374, 0.0503, 0.0469, 0.0360, 0.0505], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:01:16,768 INFO [train.py:904] (5/8) Epoch 7, batch 800, loss[loss=0.1987, simple_loss=0.27, pruned_loss=0.06375, over 16263.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2801, pruned_loss=0.06346, over 3257083.71 frames. ], batch size: 165, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:01:37,350 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:01:57,096 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:00,312 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 15:02:13,201 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:24,042 INFO [train.py:904] (5/8) Epoch 7, batch 850, loss[loss=0.188, simple_loss=0.2763, pruned_loss=0.04979, over 17130.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.279, pruned_loss=0.06242, over 3259133.70 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:02:42,593 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:51,642 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6461, 2.6708, 2.3060, 2.5373, 3.0812, 2.8993, 3.6707, 3.3352], device='cuda:5'), covar=tensor([0.0040, 0.0199, 0.0244, 0.0218, 0.0133, 0.0182, 0.0099, 0.0114], device='cuda:5'), in_proj_covar=tensor([0.0099, 0.0173, 0.0173, 0.0172, 0.0168, 0.0175, 0.0163, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:03:06,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.876e+02 3.503e+02 4.271e+02 9.516e+02, threshold=7.005e+02, percent-clipped=8.0 2023-04-28 15:03:19,367 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:33,702 INFO [train.py:904] (5/8) Epoch 7, batch 900, loss[loss=0.1743, simple_loss=0.2475, pruned_loss=0.05057, over 16976.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2774, pruned_loss=0.06154, over 3270471.98 frames. ], batch size: 41, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:03:36,354 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:04:15,047 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1968, 2.3349, 1.7145, 2.0795, 2.8041, 2.4669, 3.1452, 2.9789], device='cuda:5'), covar=tensor([0.0060, 0.0261, 0.0381, 0.0299, 0.0139, 0.0242, 0.0161, 0.0163], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0174, 0.0173, 0.0172, 0.0168, 0.0175, 0.0163, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:04:40,621 INFO [train.py:904] (5/8) Epoch 7, batch 950, loss[loss=0.1602, simple_loss=0.2435, pruned_loss=0.03846, over 17022.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2776, pruned_loss=0.06147, over 3285371.86 frames. ], batch size: 41, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:05:20,759 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.637e+02 3.238e+02 3.897e+02 7.916e+02, threshold=6.477e+02, percent-clipped=2.0 2023-04-28 15:05:46,850 INFO [train.py:904] (5/8) Epoch 7, batch 1000, loss[loss=0.2035, simple_loss=0.287, pruned_loss=0.05999, over 17013.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2767, pruned_loss=0.06165, over 3289337.54 frames. ], batch size: 55, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:06:05,355 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:06:40,602 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5460, 2.4978, 2.1212, 2.2553, 2.9004, 2.7432, 3.5782, 3.1929], device='cuda:5'), covar=tensor([0.0040, 0.0225, 0.0281, 0.0252, 0.0138, 0.0201, 0.0108, 0.0137], device='cuda:5'), in_proj_covar=tensor([0.0103, 0.0175, 0.0174, 0.0173, 0.0170, 0.0177, 0.0166, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:06:56,256 INFO [train.py:904] (5/8) Epoch 7, batch 1050, loss[loss=0.1979, simple_loss=0.2664, pruned_loss=0.06466, over 16453.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2763, pruned_loss=0.06084, over 3297731.85 frames. ], batch size: 146, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:07:28,155 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61975.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:07:37,641 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.768e+02 3.346e+02 4.170e+02 9.065e+02, threshold=6.692e+02, percent-clipped=2.0 2023-04-28 15:08:07,097 INFO [train.py:904] (5/8) Epoch 7, batch 1100, loss[loss=0.2235, simple_loss=0.2786, pruned_loss=0.0842, over 16888.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2752, pruned_loss=0.06076, over 3296773.06 frames. ], batch size: 90, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:09:15,549 INFO [train.py:904] (5/8) Epoch 7, batch 1150, loss[loss=0.181, simple_loss=0.2578, pruned_loss=0.05207, over 16298.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2743, pruned_loss=0.06002, over 3304833.93 frames. ], batch size: 165, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:09:57,386 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.657e+02 3.127e+02 3.921e+02 1.096e+03, threshold=6.253e+02, percent-clipped=5.0 2023-04-28 15:10:04,080 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:10:18,791 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62099.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:10:22,595 INFO [train.py:904] (5/8) Epoch 7, batch 1200, loss[loss=0.1937, simple_loss=0.286, pruned_loss=0.05069, over 17116.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2735, pruned_loss=0.05966, over 3309950.40 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:10:45,988 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0158, 4.3829, 4.6419, 3.4247, 3.8892, 4.4700, 4.1317, 2.8605], device='cuda:5'), covar=tensor([0.0268, 0.0025, 0.0020, 0.0210, 0.0054, 0.0048, 0.0038, 0.0287], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0063, 0.0063, 0.0119, 0.0066, 0.0078, 0.0069, 0.0115], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:10:51,372 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:11:30,501 INFO [train.py:904] (5/8) Epoch 7, batch 1250, loss[loss=0.2191, simple_loss=0.279, pruned_loss=0.07958, over 16800.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2737, pruned_loss=0.06013, over 3307270.89 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:12:13,503 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.700e+02 3.234e+02 4.045e+02 7.667e+02, threshold=6.469e+02, percent-clipped=6.0 2023-04-28 15:12:13,879 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:12:39,194 INFO [train.py:904] (5/8) Epoch 7, batch 1300, loss[loss=0.1469, simple_loss=0.228, pruned_loss=0.03294, over 16807.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.273, pruned_loss=0.05964, over 3316932.50 frames. ], batch size: 39, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:13:49,060 INFO [train.py:904] (5/8) Epoch 7, batch 1350, loss[loss=0.1985, simple_loss=0.2856, pruned_loss=0.05573, over 17073.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2731, pruned_loss=0.05928, over 3318766.38 frames. ], batch size: 53, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:14:15,372 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:14:17,826 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9528, 4.8670, 4.7111, 4.1773, 4.7788, 2.0009, 4.5632, 4.6823], device='cuda:5'), covar=tensor([0.0065, 0.0055, 0.0119, 0.0293, 0.0066, 0.1938, 0.0092, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0110, 0.0093, 0.0145, 0.0138, 0.0111, 0.0158, 0.0128, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:14:19,047 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:14:31,595 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.487e+02 3.145e+02 3.874e+02 9.778e+02, threshold=6.290e+02, percent-clipped=3.0 2023-04-28 15:14:56,761 INFO [train.py:904] (5/8) Epoch 7, batch 1400, loss[loss=0.1978, simple_loss=0.2847, pruned_loss=0.05544, over 16720.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2738, pruned_loss=0.05958, over 3319089.86 frames. ], batch size: 57, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:15:42,136 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:16:06,309 INFO [train.py:904] (5/8) Epoch 7, batch 1450, loss[loss=0.2078, simple_loss=0.2972, pruned_loss=0.05921, over 17045.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2728, pruned_loss=0.05905, over 3321780.14 frames. ], batch size: 50, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:16:12,907 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 15:16:47,646 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.994e+02 2.891e+02 3.326e+02 4.338e+02 8.815e+02, threshold=6.653e+02, percent-clipped=8.0 2023-04-28 15:16:56,066 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:00,008 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:11,101 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:14,686 INFO [train.py:904] (5/8) Epoch 7, batch 1500, loss[loss=0.2169, simple_loss=0.2807, pruned_loss=0.07656, over 16513.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2724, pruned_loss=0.05926, over 3312064.37 frames. ], batch size: 75, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:17:15,107 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5822, 3.5103, 4.1418, 2.7882, 3.7243, 4.1142, 3.8256, 2.6636], device='cuda:5'), covar=tensor([0.0322, 0.0147, 0.0032, 0.0244, 0.0049, 0.0053, 0.0038, 0.0268], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0063, 0.0062, 0.0115, 0.0065, 0.0077, 0.0067, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:17:41,082 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 15:18:01,119 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:12,712 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5937, 2.0852, 2.3945, 4.2414, 2.0384, 2.7337, 2.1609, 2.2469], device='cuda:5'), covar=tensor([0.0711, 0.2621, 0.1415, 0.0336, 0.3147, 0.1691, 0.2566, 0.2542], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0350, 0.0292, 0.0323, 0.0385, 0.0382, 0.0318, 0.0419], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:18:16,546 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:23,659 INFO [train.py:904] (5/8) Epoch 7, batch 1550, loss[loss=0.1926, simple_loss=0.2779, pruned_loss=0.05366, over 17034.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2748, pruned_loss=0.06154, over 3309403.07 frames. ], batch size: 53, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:18:24,179 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:28,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7830, 4.1456, 4.4497, 3.2522, 3.8678, 4.4338, 4.0216, 2.8606], device='cuda:5'), covar=tensor([0.0272, 0.0037, 0.0021, 0.0176, 0.0053, 0.0029, 0.0033, 0.0238], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0063, 0.0062, 0.0115, 0.0065, 0.0076, 0.0067, 0.0112], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:18:59,842 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:19:04,165 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5009, 2.5230, 2.0616, 2.3198, 2.8734, 2.7361, 3.5810, 3.1520], device='cuda:5'), covar=tensor([0.0055, 0.0233, 0.0307, 0.0278, 0.0152, 0.0232, 0.0112, 0.0160], device='cuda:5'), in_proj_covar=tensor([0.0106, 0.0175, 0.0174, 0.0176, 0.0172, 0.0178, 0.0168, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:19:06,022 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.773e+02 3.245e+02 3.904e+02 9.637e+02, threshold=6.490e+02, percent-clipped=1.0 2023-04-28 15:19:16,804 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5547, 4.4884, 4.5501, 4.4847, 4.4535, 5.1092, 4.7917, 4.4001], device='cuda:5'), covar=tensor([0.1272, 0.1836, 0.1689, 0.2078, 0.2745, 0.1043, 0.1283, 0.2643], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0446, 0.0450, 0.0377, 0.0503, 0.0478, 0.0356, 0.0508], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:19:32,106 INFO [train.py:904] (5/8) Epoch 7, batch 1600, loss[loss=0.2362, simple_loss=0.3136, pruned_loss=0.07942, over 15627.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2778, pruned_loss=0.0626, over 3310689.31 frames. ], batch size: 191, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:19:35,536 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7794, 4.8468, 5.4499, 5.3853, 5.3546, 4.9537, 4.9589, 4.7162], device='cuda:5'), covar=tensor([0.0253, 0.0344, 0.0258, 0.0318, 0.0304, 0.0263, 0.0715, 0.0350], device='cuda:5'), in_proj_covar=tensor([0.0286, 0.0287, 0.0294, 0.0277, 0.0336, 0.0306, 0.0408, 0.0249], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 15:20:01,679 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8538, 2.0004, 2.2814, 3.2008, 2.1027, 2.3916, 2.2844, 2.0762], device='cuda:5'), covar=tensor([0.0679, 0.2326, 0.1240, 0.0453, 0.2565, 0.1428, 0.1949, 0.2497], device='cuda:5'), in_proj_covar=tensor([0.0333, 0.0347, 0.0290, 0.0319, 0.0381, 0.0377, 0.0316, 0.0415], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:20:31,530 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:20:40,976 INFO [train.py:904] (5/8) Epoch 7, batch 1650, loss[loss=0.2153, simple_loss=0.2923, pruned_loss=0.06915, over 16431.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2795, pruned_loss=0.06331, over 3304287.71 frames. ], batch size: 68, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:20:47,948 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 15:21:04,913 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62570.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:21:22,704 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.048e+02 2.793e+02 3.410e+02 4.977e+02 9.439e+02, threshold=6.821e+02, percent-clipped=5.0 2023-04-28 15:21:25,427 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:21:49,316 INFO [train.py:904] (5/8) Epoch 7, batch 1700, loss[loss=0.1945, simple_loss=0.2851, pruned_loss=0.05195, over 17113.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2816, pruned_loss=0.06432, over 3309145.06 frames. ], batch size: 47, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:21:55,118 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:12,980 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:27,929 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:51,307 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:58,553 INFO [train.py:904] (5/8) Epoch 7, batch 1750, loss[loss=0.2225, simple_loss=0.3084, pruned_loss=0.06827, over 16692.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2812, pruned_loss=0.06299, over 3319372.76 frames. ], batch size: 57, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:23:18,133 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:23:21,072 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8640, 4.8983, 5.5351, 5.4685, 5.4531, 4.9753, 4.9848, 4.7944], device='cuda:5'), covar=tensor([0.0295, 0.0403, 0.0285, 0.0333, 0.0338, 0.0323, 0.0860, 0.0354], device='cuda:5'), in_proj_covar=tensor([0.0288, 0.0289, 0.0296, 0.0279, 0.0336, 0.0310, 0.0416, 0.0252], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 15:23:41,547 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.853e+02 3.319e+02 4.250e+02 7.879e+02, threshold=6.638e+02, percent-clipped=1.0 2023-04-28 15:24:08,276 INFO [train.py:904] (5/8) Epoch 7, batch 1800, loss[loss=0.1671, simple_loss=0.2549, pruned_loss=0.03963, over 17233.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2821, pruned_loss=0.06306, over 3306203.14 frames. ], batch size: 45, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:24:20,582 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 15:24:42,453 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:10,943 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:17,220 INFO [train.py:904] (5/8) Epoch 7, batch 1850, loss[loss=0.2455, simple_loss=0.3282, pruned_loss=0.08143, over 16660.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2828, pruned_loss=0.06279, over 3311987.93 frames. ], batch size: 62, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:25:45,802 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:52,380 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:59,967 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.724e+02 3.242e+02 3.816e+02 7.402e+02, threshold=6.484e+02, percent-clipped=2.0 2023-04-28 15:26:26,227 INFO [train.py:904] (5/8) Epoch 7, batch 1900, loss[loss=0.2114, simple_loss=0.2811, pruned_loss=0.07084, over 16725.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2818, pruned_loss=0.06153, over 3317232.85 frames. ], batch size: 134, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:27:00,818 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:08,904 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 15:27:11,579 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:36,785 INFO [train.py:904] (5/8) Epoch 7, batch 1950, loss[loss=0.2105, simple_loss=0.2806, pruned_loss=0.07025, over 16798.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2811, pruned_loss=0.0609, over 3321233.76 frames. ], batch size: 102, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:27:43,402 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 15:28:04,106 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4412, 2.1195, 2.1998, 4.2017, 2.0620, 2.8714, 2.1638, 2.2734], device='cuda:5'), covar=tensor([0.0738, 0.2615, 0.1559, 0.0320, 0.3067, 0.1492, 0.2493, 0.2479], device='cuda:5'), in_proj_covar=tensor([0.0336, 0.0348, 0.0290, 0.0318, 0.0383, 0.0378, 0.0316, 0.0415], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:28:19,261 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.715e+02 3.351e+02 4.043e+02 9.306e+02, threshold=6.703e+02, percent-clipped=2.0 2023-04-28 15:28:44,335 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:28:45,199 INFO [train.py:904] (5/8) Epoch 7, batch 2000, loss[loss=0.1726, simple_loss=0.2586, pruned_loss=0.04323, over 17201.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2804, pruned_loss=0.06026, over 3322139.93 frames. ], batch size: 44, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:29:19,120 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-28 15:29:24,416 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:39,666 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:39,846 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1180, 3.6983, 3.3715, 2.1915, 2.8349, 2.4521, 3.4619, 3.6478], device='cuda:5'), covar=tensor([0.0311, 0.0593, 0.0505, 0.1327, 0.0680, 0.0844, 0.0605, 0.0694], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0138, 0.0155, 0.0140, 0.0134, 0.0124, 0.0139, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 15:29:55,700 INFO [train.py:904] (5/8) Epoch 7, batch 2050, loss[loss=0.1972, simple_loss=0.2846, pruned_loss=0.05493, over 17213.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2806, pruned_loss=0.06095, over 3312554.41 frames. ], batch size: 44, lr: 9.99e-03, grad_scale: 16.0 2023-04-28 15:30:17,154 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6798, 2.8507, 2.5716, 4.2242, 3.7562, 4.1953, 1.3270, 2.9413], device='cuda:5'), covar=tensor([0.1357, 0.0551, 0.1045, 0.0101, 0.0224, 0.0329, 0.1469, 0.0719], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0148, 0.0171, 0.0104, 0.0198, 0.0200, 0.0166, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 15:30:31,335 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:30:39,216 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.104e+02 2.968e+02 3.465e+02 4.226e+02 9.943e+02, threshold=6.931e+02, percent-clipped=3.0 2023-04-28 15:31:05,142 INFO [train.py:904] (5/8) Epoch 7, batch 2100, loss[loss=0.2028, simple_loss=0.2898, pruned_loss=0.05794, over 17135.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2817, pruned_loss=0.06146, over 3309739.72 frames. ], batch size: 47, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:31:33,201 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:32:09,403 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:32:16,878 INFO [train.py:904] (5/8) Epoch 7, batch 2150, loss[loss=0.1872, simple_loss=0.2741, pruned_loss=0.05014, over 17155.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2828, pruned_loss=0.06167, over 3312971.73 frames. ], batch size: 46, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:33:00,562 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.945e+02 2.751e+02 3.323e+02 4.118e+02 9.925e+02, threshold=6.647e+02, percent-clipped=2.0 2023-04-28 15:33:17,098 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63095.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:33:25,764 INFO [train.py:904] (5/8) Epoch 7, batch 2200, loss[loss=0.2185, simple_loss=0.2935, pruned_loss=0.07179, over 16502.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.284, pruned_loss=0.06267, over 3309383.79 frames. ], batch size: 68, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:34:03,308 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:34:33,876 INFO [train.py:904] (5/8) Epoch 7, batch 2250, loss[loss=0.2251, simple_loss=0.2885, pruned_loss=0.08082, over 16718.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2846, pruned_loss=0.0634, over 3311371.76 frames. ], batch size: 124, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:35:18,531 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.937e+02 3.503e+02 4.061e+02 7.762e+02, threshold=7.006e+02, percent-clipped=5.0 2023-04-28 15:35:42,445 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63201.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:35:43,098 INFO [train.py:904] (5/8) Epoch 7, batch 2300, loss[loss=0.2432, simple_loss=0.3163, pruned_loss=0.08505, over 16684.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.285, pruned_loss=0.06363, over 3306194.70 frames. ], batch size: 89, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:36:10,180 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:22,913 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6227, 3.2543, 2.6363, 4.9790, 4.3852, 4.6868, 1.5679, 3.4514], device='cuda:5'), covar=tensor([0.1346, 0.0555, 0.1122, 0.0130, 0.0269, 0.0331, 0.1404, 0.0579], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0147, 0.0169, 0.0104, 0.0199, 0.0201, 0.0166, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 15:36:37,707 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:48,221 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:51,523 INFO [train.py:904] (5/8) Epoch 7, batch 2350, loss[loss=0.214, simple_loss=0.2821, pruned_loss=0.0729, over 16882.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2844, pruned_loss=0.06305, over 3314685.75 frames. ], batch size: 96, lr: 9.96e-03, grad_scale: 4.0 2023-04-28 15:37:01,886 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7133, 3.2912, 2.8031, 5.1220, 4.4557, 4.7168, 1.5379, 3.4704], device='cuda:5'), covar=tensor([0.1304, 0.0543, 0.1010, 0.0090, 0.0246, 0.0324, 0.1416, 0.0588], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0147, 0.0169, 0.0104, 0.0200, 0.0201, 0.0166, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 15:37:34,010 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:37:38,008 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.053e+02 2.699e+02 3.274e+02 4.399e+02 8.377e+02, threshold=6.549e+02, percent-clipped=2.0 2023-04-28 15:37:43,610 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:38:00,173 INFO [train.py:904] (5/8) Epoch 7, batch 2400, loss[loss=0.2223, simple_loss=0.2914, pruned_loss=0.07657, over 16787.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2865, pruned_loss=0.06383, over 3314459.40 frames. ], batch size: 83, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:38:27,639 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:39:10,361 INFO [train.py:904] (5/8) Epoch 7, batch 2450, loss[loss=0.2255, simple_loss=0.3129, pruned_loss=0.06904, over 16683.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2857, pruned_loss=0.06285, over 3318811.25 frames. ], batch size: 57, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:39:34,203 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:39:52,730 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5559, 3.8408, 1.9952, 3.9425, 2.7153, 3.9540, 2.1220, 2.9278], device='cuda:5'), covar=tensor([0.0159, 0.0227, 0.1321, 0.0115, 0.0676, 0.0420, 0.1288, 0.0568], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0164, 0.0181, 0.0100, 0.0161, 0.0205, 0.0189, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 15:39:55,100 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.805e+02 3.565e+02 4.384e+02 1.263e+03, threshold=7.129e+02, percent-clipped=5.0 2023-04-28 15:39:56,478 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 15:40:19,249 INFO [train.py:904] (5/8) Epoch 7, batch 2500, loss[loss=0.18, simple_loss=0.2676, pruned_loss=0.04615, over 16869.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.285, pruned_loss=0.06233, over 3319908.26 frames. ], batch size: 42, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:40:44,245 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:46,539 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:55,631 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:41:26,930 INFO [train.py:904] (5/8) Epoch 7, batch 2550, loss[loss=0.1899, simple_loss=0.2648, pruned_loss=0.05745, over 16754.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2854, pruned_loss=0.06332, over 3320378.09 frames. ], batch size: 83, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:42:02,626 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:08,290 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:11,101 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63483.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:14,138 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 3.238e+02 3.870e+02 4.521e+02 8.304e+02, threshold=7.740e+02, percent-clipped=2.0 2023-04-28 15:42:35,903 INFO [train.py:904] (5/8) Epoch 7, batch 2600, loss[loss=0.2143, simple_loss=0.287, pruned_loss=0.0708, over 16837.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2846, pruned_loss=0.06309, over 3323602.31 frames. ], batch size: 96, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:43:26,858 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3557, 4.7032, 5.1161, 5.0705, 5.0872, 4.7537, 4.2026, 4.4469], device='cuda:5'), covar=tensor([0.0585, 0.0547, 0.0524, 0.0650, 0.0561, 0.0511, 0.1480, 0.0551], device='cuda:5'), in_proj_covar=tensor([0.0286, 0.0290, 0.0292, 0.0276, 0.0334, 0.0308, 0.0415, 0.0248], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 15:43:42,048 INFO [train.py:904] (5/8) Epoch 7, batch 2650, loss[loss=0.1782, simple_loss=0.2669, pruned_loss=0.04481, over 16959.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2846, pruned_loss=0.06202, over 3329521.20 frames. ], batch size: 41, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:44:06,158 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:15,588 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63577.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:27,538 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.764e+02 3.348e+02 3.954e+02 9.438e+02, threshold=6.696e+02, percent-clipped=1.0 2023-04-28 15:44:27,943 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:48,761 INFO [train.py:904] (5/8) Epoch 7, batch 2700, loss[loss=0.2205, simple_loss=0.2865, pruned_loss=0.07725, over 16433.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2849, pruned_loss=0.06156, over 3335477.68 frames. ], batch size: 146, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:45:15,839 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8009, 4.0853, 4.2221, 3.0712, 3.6173, 4.0957, 3.9364, 2.6089], device='cuda:5'), covar=tensor([0.0274, 0.0050, 0.0029, 0.0226, 0.0063, 0.0046, 0.0038, 0.0264], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0063, 0.0061, 0.0116, 0.0065, 0.0076, 0.0067, 0.0112], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:45:28,066 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:45:52,004 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63647.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:45:57,997 INFO [train.py:904] (5/8) Epoch 7, batch 2750, loss[loss=0.2032, simple_loss=0.277, pruned_loss=0.06465, over 16825.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2849, pruned_loss=0.06142, over 3334979.78 frames. ], batch size: 116, lr: 9.93e-03, grad_scale: 4.0 2023-04-28 15:46:26,074 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 15:46:45,846 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.696e+02 3.173e+02 3.925e+02 8.532e+02, threshold=6.347e+02, percent-clipped=1.0 2023-04-28 15:46:47,423 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6142, 6.0071, 5.6686, 5.7050, 5.2655, 5.1371, 5.4683, 6.0322], device='cuda:5'), covar=tensor([0.0873, 0.0719, 0.1042, 0.0576, 0.0683, 0.0609, 0.0743, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0454, 0.0587, 0.0491, 0.0387, 0.0369, 0.0379, 0.0486, 0.0432], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:46:53,616 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.62 vs. limit=5.0 2023-04-28 15:46:57,058 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3512, 3.7515, 3.8636, 1.9745, 4.0496, 4.0037, 3.0034, 3.0347], device='cuda:5'), covar=tensor([0.0727, 0.0113, 0.0097, 0.0991, 0.0047, 0.0080, 0.0334, 0.0348], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0097, 0.0084, 0.0138, 0.0071, 0.0090, 0.0119, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 15:47:08,537 INFO [train.py:904] (5/8) Epoch 7, batch 2800, loss[loss=0.1985, simple_loss=0.2801, pruned_loss=0.05846, over 16608.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2854, pruned_loss=0.06135, over 3337166.31 frames. ], batch size: 68, lr: 9.93e-03, grad_scale: 8.0 2023-04-28 15:48:12,844 INFO [train.py:904] (5/8) Epoch 7, batch 2850, loss[loss=0.2006, simple_loss=0.288, pruned_loss=0.05658, over 17130.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2845, pruned_loss=0.06103, over 3333776.21 frames. ], batch size: 47, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:48:15,583 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0187, 5.5164, 5.6186, 5.4311, 5.5233, 6.0497, 5.6298, 5.4268], device='cuda:5'), covar=tensor([0.0739, 0.1776, 0.1577, 0.2008, 0.2432, 0.1039, 0.1160, 0.2222], device='cuda:5'), in_proj_covar=tensor([0.0319, 0.0459, 0.0458, 0.0389, 0.0524, 0.0488, 0.0370, 0.0524], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:48:47,359 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:48:50,242 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:49:02,041 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.714e+02 3.034e+02 3.568e+02 4.214e+02 8.361e+02, threshold=7.135e+02, percent-clipped=4.0 2023-04-28 15:49:23,203 INFO [train.py:904] (5/8) Epoch 7, batch 2900, loss[loss=0.2358, simple_loss=0.3216, pruned_loss=0.07501, over 16726.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2845, pruned_loss=0.06201, over 3323214.35 frames. ], batch size: 57, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:49:30,788 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 15:49:35,903 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:50:22,289 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:50:24,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5197, 5.9000, 5.5951, 5.6468, 5.2578, 5.0771, 5.3366, 5.9863], device='cuda:5'), covar=tensor([0.0780, 0.0688, 0.0848, 0.0516, 0.0631, 0.0589, 0.0713, 0.0689], device='cuda:5'), in_proj_covar=tensor([0.0459, 0.0596, 0.0496, 0.0391, 0.0372, 0.0383, 0.0487, 0.0436], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:50:32,273 INFO [train.py:904] (5/8) Epoch 7, batch 2950, loss[loss=0.2343, simple_loss=0.2924, pruned_loss=0.08813, over 16464.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2844, pruned_loss=0.06271, over 3331489.39 frames. ], batch size: 75, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:50:58,348 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:03,067 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3801, 4.3356, 4.2480, 4.1177, 3.9165, 4.3215, 4.0988, 4.0263], device='cuda:5'), covar=tensor([0.0475, 0.0339, 0.0220, 0.0187, 0.0748, 0.0302, 0.0505, 0.0481], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0252, 0.0260, 0.0232, 0.0294, 0.0260, 0.0183, 0.0296], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:51:05,408 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:17,103 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 3.004e+02 3.573e+02 4.238e+02 9.088e+02, threshold=7.147e+02, percent-clipped=1.0 2023-04-28 15:51:26,328 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3549, 5.3016, 5.1336, 4.4913, 5.1353, 2.0327, 4.8880, 5.2984], device='cuda:5'), covar=tensor([0.0050, 0.0048, 0.0109, 0.0305, 0.0057, 0.1742, 0.0090, 0.0113], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0099, 0.0151, 0.0148, 0.0116, 0.0160, 0.0133, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:51:38,108 INFO [train.py:904] (5/8) Epoch 7, batch 3000, loss[loss=0.2007, simple_loss=0.2808, pruned_loss=0.06025, over 16995.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2836, pruned_loss=0.06255, over 3329514.77 frames. ], batch size: 41, lr: 9.91e-03, grad_scale: 8.0 2023-04-28 15:51:38,108 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 15:51:46,841 INFO [train.py:938] (5/8) Epoch 7, validation: loss=0.1489, simple_loss=0.2553, pruned_loss=0.02124, over 944034.00 frames. 2023-04-28 15:51:46,842 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 15:51:54,309 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:07,315 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:16,563 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6192, 3.8330, 2.0960, 4.0008, 2.6971, 4.0236, 2.2371, 3.1048], device='cuda:5'), covar=tensor([0.0156, 0.0271, 0.1349, 0.0126, 0.0677, 0.0403, 0.1253, 0.0534], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0163, 0.0179, 0.0101, 0.0161, 0.0205, 0.0189, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 15:52:20,151 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:20,171 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:42,099 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:56,536 INFO [train.py:904] (5/8) Epoch 7, batch 3050, loss[loss=0.2424, simple_loss=0.2995, pruned_loss=0.09265, over 16752.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2824, pruned_loss=0.06199, over 3336811.85 frames. ], batch size: 134, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:53:29,257 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7714, 2.1072, 2.2929, 4.3791, 2.0608, 2.8466, 2.3091, 2.3461], device='cuda:5'), covar=tensor([0.0655, 0.2738, 0.1531, 0.0336, 0.3294, 0.1678, 0.2238, 0.2698], device='cuda:5'), in_proj_covar=tensor([0.0343, 0.0356, 0.0296, 0.0327, 0.0389, 0.0389, 0.0321, 0.0421], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 15:53:30,188 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:53:43,714 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.789e+02 3.428e+02 3.925e+02 6.614e+02, threshold=6.856e+02, percent-clipped=0.0 2023-04-28 15:54:06,476 INFO [train.py:904] (5/8) Epoch 7, batch 3100, loss[loss=0.2402, simple_loss=0.3088, pruned_loss=0.08578, over 12376.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2828, pruned_loss=0.06332, over 3319352.52 frames. ], batch size: 246, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:54:36,617 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:54:43,594 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1250, 4.1512, 4.4549, 2.1168, 4.6977, 4.6359, 3.3186, 3.7347], device='cuda:5'), covar=tensor([0.0596, 0.0165, 0.0148, 0.1067, 0.0041, 0.0077, 0.0298, 0.0296], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0100, 0.0086, 0.0141, 0.0073, 0.0093, 0.0122, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 15:55:06,574 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8699, 3.3090, 2.8530, 4.6249, 4.0939, 4.5309, 1.7220, 3.2297], device='cuda:5'), covar=tensor([0.1273, 0.0501, 0.0901, 0.0130, 0.0277, 0.0323, 0.1322, 0.0653], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0148, 0.0169, 0.0108, 0.0202, 0.0201, 0.0167, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 15:55:16,385 INFO [train.py:904] (5/8) Epoch 7, batch 3150, loss[loss=0.1872, simple_loss=0.2538, pruned_loss=0.06027, over 17011.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2813, pruned_loss=0.06238, over 3321386.17 frames. ], batch size: 41, lr: 9.90e-03, grad_scale: 4.0 2023-04-28 15:55:18,201 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 15:55:49,727 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:55:52,155 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:55:54,557 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1627, 1.8285, 2.4415, 3.0370, 2.6484, 3.4360, 2.0314, 3.3872], device='cuda:5'), covar=tensor([0.0121, 0.0283, 0.0181, 0.0151, 0.0179, 0.0112, 0.0262, 0.0074], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0157, 0.0140, 0.0143, 0.0148, 0.0105, 0.0149, 0.0096], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 15:56:00,132 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:03,682 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.613e+02 3.331e+02 4.706e+02 9.304e+02, threshold=6.662e+02, percent-clipped=4.0 2023-04-28 15:56:24,292 INFO [train.py:904] (5/8) Epoch 7, batch 3200, loss[loss=0.2389, simple_loss=0.3042, pruned_loss=0.08684, over 16867.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2808, pruned_loss=0.06218, over 3326065.95 frames. ], batch size: 109, lr: 9.90e-03, grad_scale: 8.0 2023-04-28 15:56:34,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6672, 4.6319, 4.7177, 4.7526, 4.6286, 5.2367, 4.8345, 4.5379], device='cuda:5'), covar=tensor([0.1266, 0.1703, 0.1649, 0.1975, 0.2912, 0.1149, 0.1326, 0.2753], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0458, 0.0454, 0.0380, 0.0519, 0.0489, 0.0368, 0.0524], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 15:56:54,790 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:57,816 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:57:33,495 INFO [train.py:904] (5/8) Epoch 7, batch 3250, loss[loss=0.2458, simple_loss=0.3091, pruned_loss=0.09127, over 16317.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2815, pruned_loss=0.06261, over 3327407.16 frames. ], batch size: 165, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:57:53,164 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:58:01,564 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 15:58:21,283 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.809e+02 3.257e+02 3.904e+02 7.986e+02, threshold=6.515e+02, percent-clipped=1.0 2023-04-28 15:58:42,191 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64201.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:58:43,066 INFO [train.py:904] (5/8) Epoch 7, batch 3300, loss[loss=0.1946, simple_loss=0.2712, pruned_loss=0.05902, over 16853.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2818, pruned_loss=0.0621, over 3338757.16 frames. ], batch size: 96, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:59:16,291 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:25,417 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:34,074 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8259, 4.2356, 2.1349, 4.5168, 2.9007, 4.5106, 2.2142, 3.3098], device='cuda:5'), covar=tensor([0.0158, 0.0222, 0.1411, 0.0090, 0.0692, 0.0365, 0.1418, 0.0545], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0162, 0.0177, 0.0101, 0.0162, 0.0206, 0.0190, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 15:59:41,231 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:51,977 INFO [train.py:904] (5/8) Epoch 7, batch 3350, loss[loss=0.186, simple_loss=0.2604, pruned_loss=0.05582, over 16692.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2829, pruned_loss=0.06256, over 3335878.64 frames. ], batch size: 89, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:59:53,379 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.65 vs. limit=5.0 2023-04-28 15:59:59,770 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5918, 2.6871, 1.7252, 2.6474, 2.1875, 2.7346, 1.9646, 2.3546], device='cuda:5'), covar=tensor([0.0206, 0.0364, 0.1302, 0.0176, 0.0668, 0.0445, 0.1149, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0164, 0.0179, 0.0101, 0.0163, 0.0208, 0.0191, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:00:20,557 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:22,691 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:42,050 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.871e+02 3.440e+02 4.131e+02 8.469e+02, threshold=6.880e+02, percent-clipped=3.0 2023-04-28 16:00:45,938 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:49,518 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:01:01,640 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-04-28 16:01:02,163 INFO [train.py:904] (5/8) Epoch 7, batch 3400, loss[loss=0.1999, simple_loss=0.2825, pruned_loss=0.05861, over 17179.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2826, pruned_loss=0.06212, over 3323108.64 frames. ], batch size: 46, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:01:59,792 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 16:02:02,953 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:02:10,861 INFO [train.py:904] (5/8) Epoch 7, batch 3450, loss[loss=0.215, simple_loss=0.2783, pruned_loss=0.0758, over 16876.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.281, pruned_loss=0.0616, over 3321038.86 frames. ], batch size: 116, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:02:50,063 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:03:01,785 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.661e+02 3.216e+02 3.896e+02 8.084e+02, threshold=6.431e+02, percent-clipped=1.0 2023-04-28 16:03:13,338 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9611, 2.6491, 2.6121, 1.8288, 2.8284, 2.8005, 2.4552, 2.3738], device='cuda:5'), covar=tensor([0.0804, 0.0190, 0.0187, 0.0883, 0.0101, 0.0149, 0.0423, 0.0449], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0099, 0.0085, 0.0139, 0.0073, 0.0092, 0.0121, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 16:03:21,655 INFO [train.py:904] (5/8) Epoch 7, batch 3500, loss[loss=0.1709, simple_loss=0.2518, pruned_loss=0.04503, over 16812.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2792, pruned_loss=0.0606, over 3314438.46 frames. ], batch size: 42, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:03:28,778 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:04:30,075 INFO [train.py:904] (5/8) Epoch 7, batch 3550, loss[loss=0.1959, simple_loss=0.2805, pruned_loss=0.05569, over 16742.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2782, pruned_loss=0.05965, over 3324339.60 frames. ], batch size: 57, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:04:48,211 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:04:50,576 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:05:18,415 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.738e+02 2.678e+02 3.341e+02 4.465e+02 1.003e+03, threshold=6.682e+02, percent-clipped=6.0 2023-04-28 16:05:26,377 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 16:05:38,529 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:05:39,302 INFO [train.py:904] (5/8) Epoch 7, batch 3600, loss[loss=0.2034, simple_loss=0.2742, pruned_loss=0.06624, over 16243.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2778, pruned_loss=0.05991, over 3322442.14 frames. ], batch size: 165, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:05:56,509 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:14,304 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:20,855 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7567, 3.8317, 4.2356, 2.9307, 3.7903, 4.1322, 3.9107, 2.4862], device='cuda:5'), covar=tensor([0.0290, 0.0141, 0.0025, 0.0219, 0.0046, 0.0056, 0.0039, 0.0273], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0061, 0.0061, 0.0115, 0.0065, 0.0076, 0.0068, 0.0111], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 16:06:23,280 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5934, 2.6261, 2.0205, 2.3105, 2.9700, 2.7143, 3.4672, 3.1633], device='cuda:5'), covar=tensor([0.0051, 0.0203, 0.0269, 0.0256, 0.0147, 0.0219, 0.0147, 0.0143], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0175, 0.0171, 0.0174, 0.0172, 0.0177, 0.0175, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:06:45,523 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:48,847 INFO [train.py:904] (5/8) Epoch 7, batch 3650, loss[loss=0.2186, simple_loss=0.2772, pruned_loss=0.08003, over 16871.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2768, pruned_loss=0.06034, over 3305136.21 frames. ], batch size: 116, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:06:52,140 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 16:06:56,272 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 16:07:02,262 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 16:07:19,341 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:40,348 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 2.562e+02 3.124e+02 4.018e+02 1.198e+03, threshold=6.249e+02, percent-clipped=5.0 2023-04-28 16:07:43,317 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:44,582 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64589.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:08:02,010 INFO [train.py:904] (5/8) Epoch 7, batch 3700, loss[loss=0.1896, simple_loss=0.2532, pruned_loss=0.06304, over 16784.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2758, pruned_loss=0.06177, over 3292119.17 frames. ], batch size: 102, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:08:27,963 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:10,003 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:12,296 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:14,075 INFO [train.py:904] (5/8) Epoch 7, batch 3750, loss[loss=0.2566, simple_loss=0.3261, pruned_loss=0.09358, over 11543.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2766, pruned_loss=0.06325, over 3286668.51 frames. ], batch size: 246, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:09:40,409 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8391, 3.3139, 2.9926, 1.8746, 2.6674, 2.2307, 3.3595, 3.3896], device='cuda:5'), covar=tensor([0.0232, 0.0585, 0.0603, 0.1611, 0.0736, 0.0885, 0.0506, 0.0723], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0138, 0.0154, 0.0139, 0.0131, 0.0122, 0.0136, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:09:55,038 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:04,724 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.843e+02 3.454e+02 4.284e+02 7.405e+02, threshold=6.908e+02, percent-clipped=2.0 2023-04-28 16:10:24,974 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:25,840 INFO [train.py:904] (5/8) Epoch 7, batch 3800, loss[loss=0.2156, simple_loss=0.2814, pruned_loss=0.07491, over 16796.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2776, pruned_loss=0.06442, over 3287155.72 frames. ], batch size: 102, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:10:27,997 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:38,656 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:48,015 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3937, 3.8646, 3.5253, 1.8928, 2.9062, 2.3153, 3.7586, 3.9037], device='cuda:5'), covar=tensor([0.0205, 0.0437, 0.0541, 0.1695, 0.0762, 0.0892, 0.0477, 0.0688], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0138, 0.0155, 0.0140, 0.0132, 0.0123, 0.0137, 0.0148], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:11:02,600 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4195, 2.4550, 1.6972, 2.1848, 2.9359, 2.6196, 3.2614, 3.1778], device='cuda:5'), covar=tensor([0.0040, 0.0199, 0.0318, 0.0256, 0.0121, 0.0197, 0.0075, 0.0109], device='cuda:5'), in_proj_covar=tensor([0.0109, 0.0173, 0.0171, 0.0172, 0.0170, 0.0174, 0.0171, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:11:05,643 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:18,022 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 16:11:31,459 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:40,025 INFO [train.py:904] (5/8) Epoch 7, batch 3850, loss[loss=0.2198, simple_loss=0.2875, pruned_loss=0.07602, over 16324.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2778, pruned_loss=0.06527, over 3276963.89 frames. ], batch size: 165, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:11:41,077 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:58,648 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:12:33,104 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.797e+02 3.034e+02 3.668e+02 1.110e+03, threshold=6.069e+02, percent-clipped=1.0 2023-04-28 16:12:52,616 INFO [train.py:904] (5/8) Epoch 7, batch 3900, loss[loss=0.1866, simple_loss=0.259, pruned_loss=0.05716, over 16407.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2777, pruned_loss=0.06584, over 3269170.32 frames. ], batch size: 75, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:13:00,397 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:08,269 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:20,547 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:14:02,907 INFO [train.py:904] (5/8) Epoch 7, batch 3950, loss[loss=0.2334, simple_loss=0.2996, pruned_loss=0.0836, over 16291.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2776, pruned_loss=0.06635, over 3267756.84 frames. ], batch size: 165, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:14:31,469 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2386, 3.6207, 3.0998, 2.0297, 2.8279, 2.1561, 3.5718, 3.6195], device='cuda:5'), covar=tensor([0.0231, 0.0515, 0.0611, 0.1499, 0.0705, 0.0977, 0.0465, 0.0629], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0138, 0.0155, 0.0139, 0.0132, 0.0124, 0.0136, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:14:47,543 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0564, 3.9988, 4.0986, 3.1762, 4.0316, 1.5756, 3.8078, 3.6116], device='cuda:5'), covar=tensor([0.0131, 0.0101, 0.0142, 0.0504, 0.0104, 0.2539, 0.0140, 0.0272], device='cuda:5'), in_proj_covar=tensor([0.0112, 0.0099, 0.0150, 0.0147, 0.0116, 0.0159, 0.0134, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:14:53,035 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.882e+02 3.338e+02 4.272e+02 6.667e+02, threshold=6.675e+02, percent-clipped=4.0 2023-04-28 16:14:54,275 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:15:12,237 INFO [train.py:904] (5/8) Epoch 7, batch 4000, loss[loss=0.2014, simple_loss=0.2745, pruned_loss=0.06419, over 16846.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2775, pruned_loss=0.06659, over 3266623.95 frames. ], batch size: 116, lr: 9.84e-03, grad_scale: 8.0 2023-04-28 16:15:33,018 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-28 16:15:45,307 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8109, 2.7475, 2.7332, 1.8309, 2.5047, 2.7201, 2.6854, 1.8423], device='cuda:5'), covar=tensor([0.0321, 0.0040, 0.0032, 0.0240, 0.0051, 0.0053, 0.0039, 0.0270], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0060, 0.0060, 0.0114, 0.0064, 0.0074, 0.0066, 0.0110], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 16:16:02,048 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:16,962 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:24,735 INFO [train.py:904] (5/8) Epoch 7, batch 4050, loss[loss=0.1996, simple_loss=0.2802, pruned_loss=0.05952, over 16470.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2768, pruned_loss=0.06477, over 3252579.01 frames. ], batch size: 75, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:16:25,645 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 16:16:59,766 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:17:06,214 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6143, 4.3895, 4.6323, 4.8456, 4.9590, 4.4828, 4.9092, 4.9627], device='cuda:5'), covar=tensor([0.1070, 0.0888, 0.1134, 0.0452, 0.0387, 0.0651, 0.0441, 0.0412], device='cuda:5'), in_proj_covar=tensor([0.0461, 0.0559, 0.0711, 0.0574, 0.0436, 0.0432, 0.0446, 0.0492], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:17:16,162 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.366e+02 2.141e+02 2.464e+02 3.078e+02 9.943e+02, threshold=4.928e+02, percent-clipped=3.0 2023-04-28 16:17:35,692 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65001.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:17:36,414 INFO [train.py:904] (5/8) Epoch 7, batch 4100, loss[loss=0.2336, simple_loss=0.3159, pruned_loss=0.07568, over 15455.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2773, pruned_loss=0.06333, over 3253832.26 frames. ], batch size: 191, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:17:39,310 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:15,479 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 16:18:20,395 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5875, 4.2819, 4.2642, 2.5932, 3.6516, 4.2225, 3.9394, 2.4689], device='cuda:5'), covar=tensor([0.0296, 0.0013, 0.0017, 0.0249, 0.0042, 0.0045, 0.0027, 0.0264], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0058, 0.0059, 0.0113, 0.0063, 0.0073, 0.0065, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 16:18:21,733 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3195, 1.5331, 1.9258, 2.3945, 2.3293, 2.5804, 1.5991, 2.4959], device='cuda:5'), covar=tensor([0.0111, 0.0257, 0.0151, 0.0141, 0.0129, 0.0080, 0.0237, 0.0049], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0156, 0.0141, 0.0142, 0.0149, 0.0107, 0.0152, 0.0096], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 16:18:29,712 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:18:42,627 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6286, 3.8116, 1.7902, 4.1570, 2.5479, 4.1434, 2.1136, 3.0089], device='cuda:5'), covar=tensor([0.0176, 0.0297, 0.1824, 0.0059, 0.0803, 0.0375, 0.1609, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0163, 0.0181, 0.0099, 0.0161, 0.0203, 0.0192, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:18:47,167 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:51,083 INFO [train.py:904] (5/8) Epoch 7, batch 4150, loss[loss=0.2809, simple_loss=0.3389, pruned_loss=0.1114, over 11662.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2855, pruned_loss=0.06706, over 3217425.58 frames. ], batch size: 246, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:18:56,872 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:19:01,902 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:19:44,056 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.893e+02 2.789e+02 3.341e+02 3.997e+02 8.596e+02, threshold=6.682e+02, percent-clipped=9.0 2023-04-28 16:20:03,757 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:04,514 INFO [train.py:904] (5/8) Epoch 7, batch 4200, loss[loss=0.2166, simple_loss=0.3094, pruned_loss=0.06185, over 16716.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2927, pruned_loss=0.06907, over 3188951.41 frames. ], batch size: 89, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:20:05,030 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9893, 4.9102, 4.7772, 4.1585, 4.8270, 1.7310, 4.6546, 4.6435], device='cuda:5'), covar=tensor([0.0046, 0.0040, 0.0086, 0.0267, 0.0045, 0.1977, 0.0072, 0.0120], device='cuda:5'), in_proj_covar=tensor([0.0109, 0.0095, 0.0144, 0.0143, 0.0112, 0.0154, 0.0129, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:20:13,270 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:24,672 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:32,603 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:21:16,727 INFO [train.py:904] (5/8) Epoch 7, batch 4250, loss[loss=0.196, simple_loss=0.2949, pruned_loss=0.04857, over 16771.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2962, pruned_loss=0.06933, over 3167433.62 frames. ], batch size: 89, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:21:35,434 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 16:21:42,350 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:21:51,330 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 16:22:08,031 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.682e+02 2.671e+02 3.307e+02 3.991e+02 9.211e+02, threshold=6.614e+02, percent-clipped=3.0 2023-04-28 16:22:29,058 INFO [train.py:904] (5/8) Epoch 7, batch 4300, loss[loss=0.2473, simple_loss=0.3223, pruned_loss=0.08613, over 11797.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2973, pruned_loss=0.06825, over 3172840.63 frames. ], batch size: 247, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:23:31,903 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:23:41,840 INFO [train.py:904] (5/8) Epoch 7, batch 4350, loss[loss=0.2228, simple_loss=0.3086, pruned_loss=0.0685, over 16660.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3003, pruned_loss=0.06947, over 3157029.89 frames. ], batch size: 134, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:30,751 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:33,883 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.885e+02 3.415e+02 4.232e+02 9.360e+02, threshold=6.829e+02, percent-clipped=2.0 2023-04-28 16:24:42,767 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:54,960 INFO [train.py:904] (5/8) Epoch 7, batch 4400, loss[loss=0.2471, simple_loss=0.3317, pruned_loss=0.08122, over 16929.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3029, pruned_loss=0.07068, over 3162255.47 frames. ], batch size: 109, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:57,749 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:25:17,220 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7159, 3.6284, 3.8377, 3.6506, 3.7694, 4.1295, 3.9452, 3.5248], device='cuda:5'), covar=tensor([0.1950, 0.1937, 0.1578, 0.2010, 0.2567, 0.1653, 0.1104, 0.2565], device='cuda:5'), in_proj_covar=tensor([0.0307, 0.0439, 0.0432, 0.0365, 0.0493, 0.0461, 0.0348, 0.0499], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 16:25:39,727 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:25:59,818 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65346.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:07,822 INFO [train.py:904] (5/8) Epoch 7, batch 4450, loss[loss=0.2329, simple_loss=0.3152, pruned_loss=0.07529, over 17129.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3061, pruned_loss=0.07147, over 3171511.95 frames. ], batch size: 47, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:26:08,162 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:17,503 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:58,768 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.495e+02 2.995e+02 3.500e+02 6.394e+02, threshold=5.989e+02, percent-clipped=0.0 2023-04-28 16:27:19,298 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:20,146 INFO [train.py:904] (5/8) Epoch 7, batch 4500, loss[loss=0.2266, simple_loss=0.3067, pruned_loss=0.07326, over 16698.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3055, pruned_loss=0.07094, over 3190575.17 frames. ], batch size: 76, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:27:25,603 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8166, 3.3434, 2.7871, 4.9835, 4.1068, 4.6448, 1.5610, 3.4310], device='cuda:5'), covar=tensor([0.1274, 0.0560, 0.1060, 0.0090, 0.0248, 0.0238, 0.1469, 0.0624], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0145, 0.0168, 0.0104, 0.0198, 0.0193, 0.0166, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 16:27:26,558 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:28,813 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:32,346 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:12,387 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 16:28:27,212 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:31,293 INFO [train.py:904] (5/8) Epoch 7, batch 4550, loss[loss=0.2383, simple_loss=0.3053, pruned_loss=0.08568, over 11821.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3054, pruned_loss=0.07103, over 3202468.57 frames. ], batch size: 246, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:28:38,293 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:29:21,414 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.206e+02 2.596e+02 3.051e+02 6.081e+02, threshold=5.193e+02, percent-clipped=1.0 2023-04-28 16:29:26,067 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6340, 3.8233, 2.8450, 2.3006, 2.7986, 2.1523, 3.9139, 3.6321], device='cuda:5'), covar=tensor([0.2258, 0.0665, 0.1487, 0.1726, 0.1911, 0.1701, 0.0485, 0.0682], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0254, 0.0272, 0.0261, 0.0290, 0.0212, 0.0254, 0.0274], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:29:41,305 INFO [train.py:904] (5/8) Epoch 7, batch 4600, loss[loss=0.2169, simple_loss=0.3012, pruned_loss=0.06629, over 16255.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3066, pruned_loss=0.07138, over 3197515.93 frames. ], batch size: 165, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:30:24,279 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:30:50,591 INFO [train.py:904] (5/8) Epoch 7, batch 4650, loss[loss=0.2074, simple_loss=0.2899, pruned_loss=0.06246, over 16870.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3052, pruned_loss=0.07135, over 3206773.24 frames. ], batch size: 116, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:31:36,768 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 16:31:41,970 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 2.212e+02 2.707e+02 3.141e+02 8.468e+02, threshold=5.414e+02, percent-clipped=4.0 2023-04-28 16:31:49,862 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:32:03,161 INFO [train.py:904] (5/8) Epoch 7, batch 4700, loss[loss=0.2104, simple_loss=0.2931, pruned_loss=0.06383, over 15332.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3026, pruned_loss=0.06998, over 3204331.24 frames. ], batch size: 191, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:32:07,088 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2836, 4.4144, 4.4619, 4.4233, 4.3899, 4.9312, 4.5538, 4.1902], device='cuda:5'), covar=tensor([0.1366, 0.1507, 0.1338, 0.1512, 0.2223, 0.1034, 0.0995, 0.2223], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0434, 0.0429, 0.0361, 0.0494, 0.0466, 0.0345, 0.0498], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 16:32:44,003 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2318, 5.5425, 5.2786, 5.4410, 5.0435, 4.8104, 5.0296, 5.6902], device='cuda:5'), covar=tensor([0.0825, 0.0695, 0.0893, 0.0447, 0.0583, 0.0575, 0.0749, 0.0687], device='cuda:5'), in_proj_covar=tensor([0.0426, 0.0552, 0.0468, 0.0360, 0.0348, 0.0369, 0.0458, 0.0407], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:32:45,741 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:32:56,223 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:33:12,645 INFO [train.py:904] (5/8) Epoch 7, batch 4750, loss[loss=0.1736, simple_loss=0.2552, pruned_loss=0.04593, over 17108.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2981, pruned_loss=0.06773, over 3209188.60 frames. ], batch size: 47, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:33:39,695 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-04-28 16:33:53,639 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:34:03,336 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.191e+02 2.638e+02 3.130e+02 6.213e+02, threshold=5.275e+02, percent-clipped=2.0 2023-04-28 16:34:22,794 INFO [train.py:904] (5/8) Epoch 7, batch 4800, loss[loss=0.1911, simple_loss=0.2872, pruned_loss=0.04746, over 16918.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2948, pruned_loss=0.06568, over 3218851.27 frames. ], batch size: 96, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:34:36,389 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:35:11,125 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2775, 4.2696, 4.2909, 3.5797, 4.2738, 1.7183, 4.0231, 4.0495], device='cuda:5'), covar=tensor([0.0073, 0.0065, 0.0075, 0.0341, 0.0056, 0.1825, 0.0091, 0.0124], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0090, 0.0136, 0.0135, 0.0105, 0.0149, 0.0120, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:35:13,576 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-04-28 16:35:36,917 INFO [train.py:904] (5/8) Epoch 7, batch 4850, loss[loss=0.2048, simple_loss=0.2895, pruned_loss=0.06004, over 16575.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2959, pruned_loss=0.06531, over 3200948.18 frames. ], batch size: 75, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:35:45,498 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:36:28,554 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.411e+02 2.781e+02 3.259e+02 6.496e+02, threshold=5.561e+02, percent-clipped=1.0 2023-04-28 16:36:48,888 INFO [train.py:904] (5/8) Epoch 7, batch 4900, loss[loss=0.1904, simple_loss=0.2763, pruned_loss=0.05227, over 16879.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2953, pruned_loss=0.06385, over 3189304.27 frames. ], batch size: 109, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:37:43,996 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:38:02,383 INFO [train.py:904] (5/8) Epoch 7, batch 4950, loss[loss=0.2012, simple_loss=0.2819, pruned_loss=0.0603, over 17136.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2951, pruned_loss=0.06331, over 3200024.39 frames. ], batch size: 48, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:38:53,491 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.669e+02 3.212e+02 3.846e+02 7.226e+02, threshold=6.424e+02, percent-clipped=6.0 2023-04-28 16:38:53,946 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:39:08,353 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:39:12,684 INFO [train.py:904] (5/8) Epoch 7, batch 5000, loss[loss=0.2534, simple_loss=0.3389, pruned_loss=0.08393, over 12560.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2972, pruned_loss=0.06407, over 3195443.51 frames. ], batch size: 248, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:40:09,166 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:40:23,712 INFO [train.py:904] (5/8) Epoch 7, batch 5050, loss[loss=0.2173, simple_loss=0.3048, pruned_loss=0.06496, over 16484.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2973, pruned_loss=0.06375, over 3204781.06 frames. ], batch size: 75, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:40:43,537 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 16:41:01,222 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3267, 3.8665, 3.5494, 2.0946, 3.2109, 2.5485, 3.6209, 3.8117], device='cuda:5'), covar=tensor([0.0217, 0.0497, 0.0504, 0.1550, 0.0614, 0.0786, 0.0605, 0.0822], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0134, 0.0157, 0.0142, 0.0133, 0.0126, 0.0139, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:41:11,225 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1024, 1.9413, 2.1825, 3.5924, 1.9224, 2.4124, 2.1422, 2.0858], device='cuda:5'), covar=tensor([0.0764, 0.2474, 0.1486, 0.0362, 0.3116, 0.1664, 0.2345, 0.2519], device='cuda:5'), in_proj_covar=tensor([0.0332, 0.0350, 0.0294, 0.0317, 0.0386, 0.0382, 0.0314, 0.0414], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:41:14,579 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.609e+02 2.967e+02 3.441e+02 6.265e+02, threshold=5.933e+02, percent-clipped=0.0 2023-04-28 16:41:16,171 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:41:35,940 INFO [train.py:904] (5/8) Epoch 7, batch 5100, loss[loss=0.1943, simple_loss=0.2824, pruned_loss=0.05308, over 16736.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2956, pruned_loss=0.06323, over 3212946.93 frames. ], batch size: 89, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:42:11,009 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3090, 2.3857, 2.0672, 2.2926, 2.8718, 2.5974, 3.2003, 3.1775], device='cuda:5'), covar=tensor([0.0038, 0.0258, 0.0294, 0.0239, 0.0144, 0.0195, 0.0097, 0.0106], device='cuda:5'), in_proj_covar=tensor([0.0102, 0.0174, 0.0174, 0.0172, 0.0169, 0.0175, 0.0163, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:42:39,163 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7322, 1.2826, 1.5481, 1.7532, 1.8528, 1.8834, 1.4034, 1.8531], device='cuda:5'), covar=tensor([0.0121, 0.0215, 0.0127, 0.0164, 0.0136, 0.0082, 0.0211, 0.0052], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0158, 0.0143, 0.0144, 0.0151, 0.0106, 0.0155, 0.0096], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 16:42:46,405 INFO [train.py:904] (5/8) Epoch 7, batch 5150, loss[loss=0.2327, simple_loss=0.3103, pruned_loss=0.07758, over 16602.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2951, pruned_loss=0.0625, over 3199939.85 frames. ], batch size: 57, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:43:37,177 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.591e+02 3.145e+02 3.855e+02 6.325e+02, threshold=6.290e+02, percent-clipped=3.0 2023-04-28 16:43:42,338 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6799, 3.3708, 2.9397, 1.7071, 2.6487, 2.0997, 3.1393, 3.2901], device='cuda:5'), covar=tensor([0.0263, 0.0454, 0.0633, 0.1713, 0.0745, 0.0917, 0.0648, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0131, 0.0154, 0.0139, 0.0131, 0.0123, 0.0136, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:43:53,761 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 16:43:55,948 INFO [train.py:904] (5/8) Epoch 7, batch 5200, loss[loss=0.2023, simple_loss=0.291, pruned_loss=0.05683, over 16707.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.294, pruned_loss=0.06247, over 3194083.27 frames. ], batch size: 134, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:44:56,191 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:45:06,329 INFO [train.py:904] (5/8) Epoch 7, batch 5250, loss[loss=0.179, simple_loss=0.2702, pruned_loss=0.04395, over 16817.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.291, pruned_loss=0.06159, over 3205164.77 frames. ], batch size: 102, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:45:16,817 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7078, 2.7250, 2.2755, 4.5789, 3.0878, 4.0957, 1.4800, 2.9668], device='cuda:5'), covar=tensor([0.1388, 0.0722, 0.1301, 0.0072, 0.0251, 0.0335, 0.1505, 0.0816], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0148, 0.0170, 0.0103, 0.0196, 0.0195, 0.0167, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 16:45:19,090 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:45:59,717 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.502e+02 2.872e+02 3.517e+02 7.118e+02, threshold=5.743e+02, percent-clipped=1.0 2023-04-28 16:46:00,034 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:46:06,944 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:46:18,396 INFO [train.py:904] (5/8) Epoch 7, batch 5300, loss[loss=0.1716, simple_loss=0.2584, pruned_loss=0.04239, over 16873.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2879, pruned_loss=0.06035, over 3210383.27 frames. ], batch size: 102, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:46:23,368 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:46:45,718 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:47:06,138 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:47:23,466 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6018, 4.3949, 4.5958, 4.8305, 4.9853, 4.4590, 4.9937, 4.9350], device='cuda:5'), covar=tensor([0.1065, 0.0854, 0.1225, 0.0449, 0.0382, 0.0691, 0.0309, 0.0407], device='cuda:5'), in_proj_covar=tensor([0.0445, 0.0540, 0.0679, 0.0552, 0.0414, 0.0412, 0.0420, 0.0468], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:47:28,953 INFO [train.py:904] (5/8) Epoch 7, batch 5350, loss[loss=0.2553, simple_loss=0.3175, pruned_loss=0.09652, over 11969.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2868, pruned_loss=0.06012, over 3205627.78 frames. ], batch size: 247, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:47:36,938 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:47:43,312 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 16:48:20,491 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.423e+02 2.781e+02 3.267e+02 5.283e+02, threshold=5.561e+02, percent-clipped=0.0 2023-04-28 16:48:40,528 INFO [train.py:904] (5/8) Epoch 7, batch 5400, loss[loss=0.2912, simple_loss=0.3428, pruned_loss=0.1199, over 11921.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2898, pruned_loss=0.06114, over 3199119.64 frames. ], batch size: 248, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:49:01,676 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4925, 3.4829, 3.4561, 2.8507, 3.3969, 1.9744, 3.2095, 2.9495], device='cuda:5'), covar=tensor([0.0094, 0.0076, 0.0101, 0.0222, 0.0063, 0.1736, 0.0089, 0.0154], device='cuda:5'), in_proj_covar=tensor([0.0106, 0.0093, 0.0142, 0.0141, 0.0109, 0.0155, 0.0125, 0.0134], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:49:05,948 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:49:55,999 INFO [train.py:904] (5/8) Epoch 7, batch 5450, loss[loss=0.2658, simple_loss=0.3308, pruned_loss=0.1004, over 12162.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2935, pruned_loss=0.06364, over 3186787.59 frames. ], batch size: 247, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:50:00,557 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8755, 1.9599, 2.3151, 3.1534, 2.1061, 2.4089, 2.2478, 2.1348], device='cuda:5'), covar=tensor([0.0744, 0.2465, 0.1284, 0.0418, 0.2801, 0.1410, 0.2091, 0.2260], device='cuda:5'), in_proj_covar=tensor([0.0333, 0.0347, 0.0292, 0.0317, 0.0386, 0.0380, 0.0312, 0.0412], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:50:50,983 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 3.043e+02 3.833e+02 5.336e+02 9.922e+02, threshold=7.666e+02, percent-clipped=19.0 2023-04-28 16:51:12,556 INFO [train.py:904] (5/8) Epoch 7, batch 5500, loss[loss=0.3084, simple_loss=0.3573, pruned_loss=0.1297, over 11710.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3019, pruned_loss=0.06944, over 3164257.19 frames. ], batch size: 247, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:52:29,872 INFO [train.py:904] (5/8) Epoch 7, batch 5550, loss[loss=0.2511, simple_loss=0.3282, pruned_loss=0.08704, over 16397.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3094, pruned_loss=0.07423, over 3181190.83 frames. ], batch size: 35, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:53:09,615 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7191, 5.0281, 4.6959, 4.7983, 4.4236, 4.4449, 4.5408, 5.0512], device='cuda:5'), covar=tensor([0.0760, 0.0681, 0.0888, 0.0515, 0.0699, 0.0833, 0.0744, 0.0724], device='cuda:5'), in_proj_covar=tensor([0.0430, 0.0548, 0.0465, 0.0363, 0.0343, 0.0366, 0.0453, 0.0404], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:53:27,056 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.385e+02 3.933e+02 4.760e+02 6.147e+02 1.111e+03, threshold=9.520e+02, percent-clipped=8.0 2023-04-28 16:53:36,404 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:53:47,084 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:53:49,158 INFO [train.py:904] (5/8) Epoch 7, batch 5600, loss[loss=0.261, simple_loss=0.3412, pruned_loss=0.09045, over 16681.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3144, pruned_loss=0.07849, over 3161252.11 frames. ], batch size: 134, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:53:59,843 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5234, 3.6223, 1.7348, 3.8979, 2.5223, 3.9396, 2.0130, 2.6593], device='cuda:5'), covar=tensor([0.0167, 0.0293, 0.1701, 0.0078, 0.0818, 0.0369, 0.1488, 0.0662], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0159, 0.0181, 0.0096, 0.0165, 0.0196, 0.0192, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:54:11,816 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:54:21,778 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:54:54,216 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:55:08,328 INFO [train.py:904] (5/8) Epoch 7, batch 5650, loss[loss=0.2619, simple_loss=0.3301, pruned_loss=0.09681, over 15300.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3219, pruned_loss=0.08552, over 3115311.56 frames. ], batch size: 190, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:55:53,704 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:56:03,093 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.602e+02 4.256e+02 5.191e+02 6.863e+02 2.150e+03, threshold=1.038e+03, percent-clipped=9.0 2023-04-28 16:56:23,906 INFO [train.py:904] (5/8) Epoch 7, batch 5700, loss[loss=0.2273, simple_loss=0.3196, pruned_loss=0.06752, over 16819.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3246, pruned_loss=0.08783, over 3089334.53 frames. ], batch size: 83, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:56:42,490 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:57:39,232 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0576, 2.2735, 2.3656, 2.9178, 2.5181, 3.2312, 1.7624, 2.7917], device='cuda:5'), covar=tensor([0.1031, 0.0488, 0.0904, 0.0127, 0.0218, 0.0404, 0.1211, 0.0586], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0146, 0.0168, 0.0103, 0.0196, 0.0194, 0.0166, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 16:57:41,029 INFO [train.py:904] (5/8) Epoch 7, batch 5750, loss[loss=0.2557, simple_loss=0.3347, pruned_loss=0.0884, over 15308.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3274, pruned_loss=0.08908, over 3076418.43 frames. ], batch size: 190, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:58:10,365 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0712, 3.1114, 1.6012, 3.2271, 2.3283, 3.2677, 1.9092, 2.5303], device='cuda:5'), covar=tensor([0.0229, 0.0349, 0.1687, 0.0108, 0.0807, 0.0486, 0.1547, 0.0712], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0157, 0.0179, 0.0095, 0.0163, 0.0194, 0.0190, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 16:58:35,554 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-28 16:58:39,456 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.661e+02 4.395e+02 5.448e+02 7.543e+02, threshold=8.790e+02, percent-clipped=0.0 2023-04-28 16:58:54,666 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6361, 4.6266, 4.4866, 4.3277, 4.0282, 4.5465, 4.5159, 4.1838], device='cuda:5'), covar=tensor([0.0569, 0.0503, 0.0283, 0.0237, 0.1085, 0.0423, 0.0332, 0.0645], device='cuda:5'), in_proj_covar=tensor([0.0201, 0.0236, 0.0237, 0.0211, 0.0267, 0.0242, 0.0165, 0.0271], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 16:59:02,129 INFO [train.py:904] (5/8) Epoch 7, batch 5800, loss[loss=0.2592, simple_loss=0.335, pruned_loss=0.09174, over 16764.00 frames. ], tot_loss[loss=0.2511, simple_loss=0.3271, pruned_loss=0.08753, over 3076511.21 frames. ], batch size: 124, lr: 9.70e-03, grad_scale: 16.0 2023-04-28 16:59:48,385 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:00:19,936 INFO [train.py:904] (5/8) Epoch 7, batch 5850, loss[loss=0.2636, simple_loss=0.3275, pruned_loss=0.09985, over 11585.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3245, pruned_loss=0.08561, over 3065050.06 frames. ], batch size: 246, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:00:41,264 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6774, 2.0692, 1.5629, 1.9052, 2.4689, 2.2290, 2.6501, 2.6684], device='cuda:5'), covar=tensor([0.0063, 0.0238, 0.0309, 0.0275, 0.0125, 0.0213, 0.0113, 0.0137], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0170, 0.0170, 0.0168, 0.0166, 0.0172, 0.0162, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:01:20,004 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.089e+02 3.451e+02 4.238e+02 5.428e+02 1.064e+03, threshold=8.477e+02, percent-clipped=4.0 2023-04-28 17:01:25,734 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:01:38,974 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:01:41,102 INFO [train.py:904] (5/8) Epoch 7, batch 5900, loss[loss=0.2313, simple_loss=0.3052, pruned_loss=0.07868, over 16272.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3235, pruned_loss=0.08504, over 3079507.32 frames. ], batch size: 165, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:02:08,268 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:02:22,464 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-28 17:02:56,757 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:03:02,877 INFO [train.py:904] (5/8) Epoch 7, batch 5950, loss[loss=0.2226, simple_loss=0.3085, pruned_loss=0.06837, over 16781.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3242, pruned_loss=0.08333, over 3101981.86 frames. ], batch size: 83, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:03:23,047 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:03:41,757 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:04:00,337 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 3.507e+02 4.238e+02 5.487e+02 1.167e+03, threshold=8.477e+02, percent-clipped=3.0 2023-04-28 17:04:14,885 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0242, 2.3459, 2.2497, 3.0317, 2.4191, 3.2435, 1.7117, 2.7139], device='cuda:5'), covar=tensor([0.1045, 0.0441, 0.0970, 0.0115, 0.0175, 0.0389, 0.1177, 0.0605], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0148, 0.0170, 0.0104, 0.0198, 0.0197, 0.0168, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 17:04:22,271 INFO [train.py:904] (5/8) Epoch 7, batch 6000, loss[loss=0.2379, simple_loss=0.3178, pruned_loss=0.079, over 16704.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3234, pruned_loss=0.08295, over 3101606.83 frames. ], batch size: 124, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:04:22,272 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 17:04:32,879 INFO [train.py:938] (5/8) Epoch 7, validation: loss=0.1758, simple_loss=0.2891, pruned_loss=0.03127, over 944034.00 frames. 2023-04-28 17:04:32,880 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 17:04:51,816 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:05:02,389 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8156, 3.6903, 3.8399, 3.9901, 4.0413, 3.6097, 4.0079, 4.0776], device='cuda:5'), covar=tensor([0.1186, 0.0842, 0.1232, 0.0533, 0.0540, 0.1558, 0.0571, 0.0511], device='cuda:5'), in_proj_covar=tensor([0.0444, 0.0541, 0.0675, 0.0543, 0.0417, 0.0416, 0.0423, 0.0470], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:05:53,515 INFO [train.py:904] (5/8) Epoch 7, batch 6050, loss[loss=0.2127, simple_loss=0.3038, pruned_loss=0.06083, over 16872.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.32, pruned_loss=0.08076, over 3133775.24 frames. ], batch size: 116, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:06:02,623 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:09,584 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:50,908 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 3.435e+02 4.153e+02 5.087e+02 8.285e+02, threshold=8.306e+02, percent-clipped=0.0 2023-04-28 17:07:12,274 INFO [train.py:904] (5/8) Epoch 7, batch 6100, loss[loss=0.1994, simple_loss=0.287, pruned_loss=0.05593, over 16718.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3196, pruned_loss=0.08011, over 3137809.19 frames. ], batch size: 76, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:07:39,964 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:08:32,448 INFO [train.py:904] (5/8) Epoch 7, batch 6150, loss[loss=0.2026, simple_loss=0.2876, pruned_loss=0.05882, over 17037.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3178, pruned_loss=0.07979, over 3123867.65 frames. ], batch size: 53, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:09:23,233 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:09:29,284 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:09:31,360 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 3.456e+02 4.175e+02 5.279e+02 9.590e+02, threshold=8.351e+02, percent-clipped=3.0 2023-04-28 17:09:51,777 INFO [train.py:904] (5/8) Epoch 7, batch 6200, loss[loss=0.2665, simple_loss=0.3445, pruned_loss=0.09427, over 16977.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3158, pruned_loss=0.0791, over 3149739.64 frames. ], batch size: 41, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:10:22,175 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 17:11:02,518 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:11:14,123 INFO [train.py:904] (5/8) Epoch 7, batch 6250, loss[loss=0.2278, simple_loss=0.3135, pruned_loss=0.07103, over 16235.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3145, pruned_loss=0.07825, over 3146481.50 frames. ], batch size: 165, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:11:29,052 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 17:11:52,539 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:12:11,782 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.341e+02 3.226e+02 4.153e+02 4.983e+02 8.969e+02, threshold=8.306e+02, percent-clipped=3.0 2023-04-28 17:12:28,938 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5851, 1.4556, 2.1642, 2.4907, 2.5334, 2.8445, 1.4799, 2.7889], device='cuda:5'), covar=tensor([0.0094, 0.0310, 0.0157, 0.0128, 0.0133, 0.0073, 0.0366, 0.0048], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0154, 0.0139, 0.0138, 0.0147, 0.0103, 0.0155, 0.0094], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 17:12:30,910 INFO [train.py:904] (5/8) Epoch 7, batch 6300, loss[loss=0.2188, simple_loss=0.307, pruned_loss=0.06531, over 16868.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3149, pruned_loss=0.07814, over 3138051.99 frames. ], batch size: 96, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:13:08,747 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:13:49,871 INFO [train.py:904] (5/8) Epoch 7, batch 6350, loss[loss=0.3085, simple_loss=0.3487, pruned_loss=0.1342, over 11391.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3161, pruned_loss=0.07994, over 3118057.14 frames. ], batch size: 250, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:14:47,872 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.015e+02 3.646e+02 4.526e+02 5.415e+02 1.244e+03, threshold=9.052e+02, percent-clipped=4.0 2023-04-28 17:15:05,415 INFO [train.py:904] (5/8) Epoch 7, batch 6400, loss[loss=0.3181, simple_loss=0.3695, pruned_loss=0.1334, over 11170.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3161, pruned_loss=0.08078, over 3119922.12 frames. ], batch size: 247, lr: 9.66e-03, grad_scale: 8.0 2023-04-28 17:15:13,270 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7097, 2.1020, 1.5683, 1.8731, 2.5427, 2.2271, 2.7248, 2.7320], device='cuda:5'), covar=tensor([0.0064, 0.0240, 0.0330, 0.0311, 0.0133, 0.0224, 0.0122, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0174, 0.0173, 0.0171, 0.0168, 0.0175, 0.0166, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:15:22,353 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:15:41,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4991, 4.2490, 4.4887, 4.7472, 4.8828, 4.3648, 4.8355, 4.8155], device='cuda:5'), covar=tensor([0.1330, 0.0909, 0.1491, 0.0530, 0.0485, 0.0733, 0.0491, 0.0478], device='cuda:5'), in_proj_covar=tensor([0.0438, 0.0538, 0.0671, 0.0541, 0.0414, 0.0408, 0.0420, 0.0470], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:16:16,911 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 17:16:20,298 INFO [train.py:904] (5/8) Epoch 7, batch 6450, loss[loss=0.2733, simple_loss=0.3277, pruned_loss=0.1095, over 11812.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3154, pruned_loss=0.07969, over 3119342.98 frames. ], batch size: 250, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:17:10,105 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6590, 5.0462, 5.1547, 5.0941, 5.1304, 5.6377, 5.2793, 4.9913], device='cuda:5'), covar=tensor([0.0916, 0.1469, 0.1466, 0.1383, 0.1856, 0.0917, 0.1115, 0.1867], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0423, 0.0433, 0.0364, 0.0489, 0.0465, 0.0345, 0.0499], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 17:17:16,541 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:17:22,054 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.155e+02 3.316e+02 4.009e+02 5.317e+02 1.482e+03, threshold=8.018e+02, percent-clipped=5.0 2023-04-28 17:17:24,044 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6362, 3.8541, 1.8796, 4.1694, 2.5431, 4.1237, 2.0039, 2.7797], device='cuda:5'), covar=tensor([0.0180, 0.0261, 0.1700, 0.0072, 0.0797, 0.0470, 0.1638, 0.0636], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0155, 0.0179, 0.0094, 0.0163, 0.0192, 0.0187, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 17:17:38,436 INFO [train.py:904] (5/8) Epoch 7, batch 6500, loss[loss=0.252, simple_loss=0.3348, pruned_loss=0.08466, over 16739.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3137, pruned_loss=0.07898, over 3116131.91 frames. ], batch size: 83, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:18:29,456 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:18:36,524 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:18:57,739 INFO [train.py:904] (5/8) Epoch 7, batch 6550, loss[loss=0.2255, simple_loss=0.3183, pruned_loss=0.06637, over 16609.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.317, pruned_loss=0.08071, over 3095667.92 frames. ], batch size: 62, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:19:42,946 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 17:19:56,001 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.411e+02 3.629e+02 4.533e+02 5.660e+02 1.346e+03, threshold=9.066e+02, percent-clipped=11.0 2023-04-28 17:19:59,138 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7393, 3.0576, 2.7921, 4.9306, 4.0713, 4.4848, 1.6676, 3.0834], device='cuda:5'), covar=tensor([0.1385, 0.0614, 0.1070, 0.0117, 0.0329, 0.0316, 0.1417, 0.0776], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0147, 0.0169, 0.0104, 0.0197, 0.0196, 0.0167, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 17:20:13,107 INFO [train.py:904] (5/8) Epoch 7, batch 6600, loss[loss=0.2295, simple_loss=0.3064, pruned_loss=0.07629, over 16150.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3188, pruned_loss=0.08106, over 3115827.71 frames. ], batch size: 35, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:20:30,803 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:20:50,164 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6400, 4.6483, 4.5015, 3.6110, 4.5368, 1.4942, 4.2783, 4.3278], device='cuda:5'), covar=tensor([0.0084, 0.0062, 0.0113, 0.0397, 0.0062, 0.2253, 0.0104, 0.0173], device='cuda:5'), in_proj_covar=tensor([0.0104, 0.0091, 0.0139, 0.0134, 0.0105, 0.0154, 0.0123, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:21:29,934 INFO [train.py:904] (5/8) Epoch 7, batch 6650, loss[loss=0.2009, simple_loss=0.2813, pruned_loss=0.06025, over 17061.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3198, pruned_loss=0.0828, over 3089167.80 frames. ], batch size: 53, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:22:02,947 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:22:27,318 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.256e+02 3.378e+02 3.968e+02 4.972e+02 1.148e+03, threshold=7.935e+02, percent-clipped=1.0 2023-04-28 17:22:30,295 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-28 17:22:43,252 INFO [train.py:904] (5/8) Epoch 7, batch 6700, loss[loss=0.2456, simple_loss=0.3194, pruned_loss=0.0859, over 15300.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3177, pruned_loss=0.08192, over 3096788.09 frames. ], batch size: 191, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:22:46,106 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3869, 3.3888, 3.3600, 2.6499, 3.3146, 2.0626, 2.9802, 2.6299], device='cuda:5'), covar=tensor([0.0133, 0.0103, 0.0147, 0.0361, 0.0089, 0.2160, 0.0123, 0.0231], device='cuda:5'), in_proj_covar=tensor([0.0103, 0.0091, 0.0138, 0.0134, 0.0105, 0.0154, 0.0123, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:23:00,273 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:23:21,766 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1462, 4.1397, 4.2811, 4.1853, 4.2459, 4.7558, 4.3728, 4.1290], device='cuda:5'), covar=tensor([0.1555, 0.1832, 0.1728, 0.1661, 0.2357, 0.0992, 0.1384, 0.2328], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0432, 0.0445, 0.0367, 0.0499, 0.0474, 0.0355, 0.0508], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 17:23:41,155 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3075, 3.2656, 2.5665, 2.0896, 2.3662, 1.9812, 3.3575, 3.1782], device='cuda:5'), covar=tensor([0.2447, 0.0755, 0.1456, 0.1791, 0.1934, 0.1724, 0.0483, 0.0857], device='cuda:5'), in_proj_covar=tensor([0.0292, 0.0252, 0.0276, 0.0261, 0.0283, 0.0209, 0.0254, 0.0272], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:23:49,542 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 17:23:56,587 INFO [train.py:904] (5/8) Epoch 7, batch 6750, loss[loss=0.2168, simple_loss=0.3049, pruned_loss=0.06435, over 16834.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3152, pruned_loss=0.0806, over 3123220.92 frames. ], batch size: 102, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:24:11,056 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:37,884 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:54,245 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.424e+02 3.570e+02 4.418e+02 5.557e+02 1.177e+03, threshold=8.835e+02, percent-clipped=2.0 2023-04-28 17:24:55,994 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67692.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:25:05,600 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5670, 2.0523, 1.5429, 1.8346, 2.4505, 2.1767, 2.5668, 2.6071], device='cuda:5'), covar=tensor([0.0066, 0.0230, 0.0317, 0.0301, 0.0129, 0.0224, 0.0123, 0.0138], device='cuda:5'), in_proj_covar=tensor([0.0100, 0.0174, 0.0175, 0.0172, 0.0168, 0.0176, 0.0165, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:25:10,437 INFO [train.py:904] (5/8) Epoch 7, batch 6800, loss[loss=0.2332, simple_loss=0.3137, pruned_loss=0.07633, over 16232.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3161, pruned_loss=0.08081, over 3118217.22 frames. ], batch size: 165, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:26:06,776 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:26:10,700 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67741.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:26:23,541 INFO [train.py:904] (5/8) Epoch 7, batch 6850, loss[loss=0.2196, simple_loss=0.3188, pruned_loss=0.06024, over 16580.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3181, pruned_loss=0.08172, over 3101685.59 frames. ], batch size: 68, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:26:25,729 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:26:52,060 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3404, 1.8326, 2.0196, 4.1408, 1.8092, 2.4405, 2.0084, 1.9673], device='cuda:5'), covar=tensor([0.0859, 0.3203, 0.1837, 0.0305, 0.3867, 0.1870, 0.2916, 0.2899], device='cuda:5'), in_proj_covar=tensor([0.0335, 0.0353, 0.0295, 0.0319, 0.0393, 0.0383, 0.0316, 0.0415], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:27:10,053 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 17:27:14,416 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:27:20,979 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.331e+02 3.323e+02 3.993e+02 4.819e+02 7.017e+02, threshold=7.986e+02, percent-clipped=0.0 2023-04-28 17:27:35,615 INFO [train.py:904] (5/8) Epoch 7, batch 6900, loss[loss=0.2762, simple_loss=0.3656, pruned_loss=0.09339, over 16730.00 frames. ], tot_loss[loss=0.2426, simple_loss=0.3211, pruned_loss=0.08207, over 3092868.82 frames. ], batch size: 83, lr: 9.63e-03, grad_scale: 2.0 2023-04-28 17:28:09,318 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 17:28:09,322 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-28 17:28:46,548 INFO [train.py:904] (5/8) Epoch 7, batch 6950, loss[loss=0.2175, simple_loss=0.3032, pruned_loss=0.06594, over 16823.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3235, pruned_loss=0.08444, over 3082591.08 frames. ], batch size: 102, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:29:01,175 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 17:29:11,600 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 17:29:13,183 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:29:46,383 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.351e+02 3.618e+02 4.486e+02 5.724e+02 1.200e+03, threshold=8.972e+02, percent-clipped=9.0 2023-04-28 17:29:59,772 INFO [train.py:904] (5/8) Epoch 7, batch 7000, loss[loss=0.2291, simple_loss=0.3213, pruned_loss=0.06847, over 16753.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3231, pruned_loss=0.08316, over 3085552.31 frames. ], batch size: 102, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:30:03,898 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 17:31:13,269 INFO [train.py:904] (5/8) Epoch 7, batch 7050, loss[loss=0.2371, simple_loss=0.3185, pruned_loss=0.07788, over 16733.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3235, pruned_loss=0.08293, over 3073723.97 frames. ], batch size: 124, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:31:49,129 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 17:31:57,375 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67981.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:32:14,207 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 3.870e+02 4.768e+02 5.699e+02 1.240e+03, threshold=9.537e+02, percent-clipped=2.0 2023-04-28 17:32:29,898 INFO [train.py:904] (5/8) Epoch 7, batch 7100, loss[loss=0.233, simple_loss=0.3115, pruned_loss=0.07728, over 16898.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3203, pruned_loss=0.08163, over 3090806.19 frames. ], batch size: 109, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:33:09,773 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 17:33:21,413 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:30,251 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:39,219 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:33:44,541 INFO [train.py:904] (5/8) Epoch 7, batch 7150, loss[loss=0.2592, simple_loss=0.3294, pruned_loss=0.09454, over 16512.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3191, pruned_loss=0.08187, over 3076119.65 frames. ], batch size: 68, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:34:37,084 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:34:44,536 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.291e+02 3.518e+02 4.375e+02 5.869e+02 1.278e+03, threshold=8.749e+02, percent-clipped=4.0 2023-04-28 17:34:58,212 INFO [train.py:904] (5/8) Epoch 7, batch 7200, loss[loss=0.2223, simple_loss=0.3176, pruned_loss=0.06352, over 15327.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3164, pruned_loss=0.07941, over 3070552.55 frames. ], batch size: 190, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:35:28,064 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5844, 3.9042, 1.9661, 4.2505, 2.6054, 4.1655, 2.3155, 2.7731], device='cuda:5'), covar=tensor([0.0181, 0.0222, 0.1686, 0.0056, 0.0810, 0.0353, 0.1389, 0.0704], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0155, 0.0179, 0.0094, 0.0162, 0.0192, 0.0189, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 17:35:29,352 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4745, 3.2451, 3.1214, 1.9582, 2.7241, 2.1933, 2.9719, 3.2452], device='cuda:5'), covar=tensor([0.0265, 0.0480, 0.0470, 0.1553, 0.0713, 0.0886, 0.0615, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0129, 0.0153, 0.0139, 0.0133, 0.0124, 0.0135, 0.0140], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 17:36:11,155 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:36:16,026 INFO [train.py:904] (5/8) Epoch 7, batch 7250, loss[loss=0.2045, simple_loss=0.2801, pruned_loss=0.06447, over 16250.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3142, pruned_loss=0.07849, over 3062648.67 frames. ], batch size: 165, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:36:41,052 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:37:16,279 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 3.214e+02 3.991e+02 5.016e+02 1.051e+03, threshold=7.982e+02, percent-clipped=2.0 2023-04-28 17:37:29,946 INFO [train.py:904] (5/8) Epoch 7, batch 7300, loss[loss=0.224, simple_loss=0.3078, pruned_loss=0.07008, over 16416.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3137, pruned_loss=0.07842, over 3070752.56 frames. ], batch size: 75, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:37:52,949 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:38:44,625 INFO [train.py:904] (5/8) Epoch 7, batch 7350, loss[loss=0.2224, simple_loss=0.2992, pruned_loss=0.07287, over 16723.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3141, pruned_loss=0.0791, over 3060706.31 frames. ], batch size: 62, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:39:09,037 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 17:39:48,567 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 3.085e+02 3.772e+02 4.630e+02 1.239e+03, threshold=7.544e+02, percent-clipped=2.0 2023-04-28 17:40:02,136 INFO [train.py:904] (5/8) Epoch 7, batch 7400, loss[loss=0.216, simple_loss=0.3076, pruned_loss=0.0622, over 16879.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3151, pruned_loss=0.07976, over 3053525.08 frames. ], batch size: 83, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:40:23,518 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 17:40:53,736 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:40:54,927 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:40:55,557 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 17:41:14,134 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:18,419 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9343, 1.4523, 2.2660, 2.7888, 2.5508, 3.1393, 1.9265, 2.9583], device='cuda:5'), covar=tensor([0.0089, 0.0304, 0.0172, 0.0144, 0.0153, 0.0080, 0.0261, 0.0067], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0152, 0.0133, 0.0133, 0.0144, 0.0101, 0.0151, 0.0091], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 17:41:19,088 INFO [train.py:904] (5/8) Epoch 7, batch 7450, loss[loss=0.2546, simple_loss=0.3346, pruned_loss=0.0873, over 16453.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3155, pruned_loss=0.0795, over 3085652.15 frames. ], batch size: 68, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:42:11,321 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68384.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:26,328 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.512e+02 3.665e+02 4.367e+02 5.646e+02 8.920e+02, threshold=8.734e+02, percent-clipped=5.0 2023-04-28 17:42:30,795 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:35,795 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:39,848 INFO [train.py:904] (5/8) Epoch 7, batch 7500, loss[loss=0.2251, simple_loss=0.2996, pruned_loss=0.07526, over 16516.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3167, pruned_loss=0.07952, over 3071952.21 frames. ], batch size: 68, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:43:44,405 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:43:57,921 INFO [train.py:904] (5/8) Epoch 7, batch 7550, loss[loss=0.2897, simple_loss=0.343, pruned_loss=0.1182, over 11423.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3158, pruned_loss=0.07957, over 3069907.14 frames. ], batch size: 250, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:44:10,859 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:44:34,377 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 17:44:59,748 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.309e+02 3.638e+02 4.521e+02 5.512e+02 1.190e+03, threshold=9.043e+02, percent-clipped=2.0 2023-04-28 17:45:12,772 INFO [train.py:904] (5/8) Epoch 7, batch 7600, loss[loss=0.2379, simple_loss=0.3153, pruned_loss=0.08024, over 16532.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.315, pruned_loss=0.07991, over 3070073.83 frames. ], batch size: 75, lr: 9.58e-03, grad_scale: 8.0 2023-04-28 17:45:22,634 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8266, 1.7228, 1.5147, 1.5297, 1.8293, 1.6595, 1.7339, 1.9535], device='cuda:5'), covar=tensor([0.0071, 0.0155, 0.0242, 0.0223, 0.0115, 0.0168, 0.0122, 0.0120], device='cuda:5'), in_proj_covar=tensor([0.0101, 0.0175, 0.0174, 0.0173, 0.0170, 0.0177, 0.0165, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:45:43,063 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8129, 3.0712, 3.2035, 1.8753, 2.9524, 3.0982, 2.9541, 1.7769], device='cuda:5'), covar=tensor([0.0386, 0.0031, 0.0030, 0.0300, 0.0058, 0.0076, 0.0046, 0.0331], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0056, 0.0060, 0.0116, 0.0066, 0.0077, 0.0066, 0.0111], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 17:46:28,266 INFO [train.py:904] (5/8) Epoch 7, batch 7650, loss[loss=0.2298, simple_loss=0.3178, pruned_loss=0.07085, over 16847.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3161, pruned_loss=0.08067, over 3076803.30 frames. ], batch size: 90, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:46:41,050 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 17:47:08,522 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:47:29,673 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.494e+02 3.559e+02 4.201e+02 5.687e+02 1.184e+03, threshold=8.403e+02, percent-clipped=4.0 2023-04-28 17:47:42,699 INFO [train.py:904] (5/8) Epoch 7, batch 7700, loss[loss=0.2595, simple_loss=0.3292, pruned_loss=0.09488, over 16877.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3161, pruned_loss=0.08138, over 3066963.18 frames. ], batch size: 116, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:48:35,294 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:38,398 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:43,241 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7962, 4.7074, 4.6426, 4.4362, 4.2069, 4.6020, 4.5931, 4.3706], device='cuda:5'), covar=tensor([0.0459, 0.0412, 0.0218, 0.0213, 0.0880, 0.0410, 0.0269, 0.0514], device='cuda:5'), in_proj_covar=tensor([0.0203, 0.0236, 0.0235, 0.0209, 0.0265, 0.0238, 0.0165, 0.0272], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:48:57,122 INFO [train.py:904] (5/8) Epoch 7, batch 7750, loss[loss=0.2132, simple_loss=0.2991, pruned_loss=0.0637, over 16845.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3168, pruned_loss=0.08153, over 3072195.76 frames. ], batch size: 96, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:49:44,494 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:49:44,684 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:49:57,425 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.246e+02 3.570e+02 4.272e+02 5.150e+02 9.085e+02, threshold=8.545e+02, percent-clipped=2.0 2023-04-28 17:50:11,059 INFO [train.py:904] (5/8) Epoch 7, batch 7800, loss[loss=0.3193, simple_loss=0.3658, pruned_loss=0.1364, over 11286.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3179, pruned_loss=0.08217, over 3082252.75 frames. ], batch size: 247, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:13,409 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:51:17,580 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:51:25,161 INFO [train.py:904] (5/8) Epoch 7, batch 7850, loss[loss=0.2476, simple_loss=0.3139, pruned_loss=0.09066, over 11911.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3188, pruned_loss=0.08197, over 3097876.16 frames. ], batch size: 247, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:30,896 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:52:24,298 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:52:26,860 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 3.517e+02 4.266e+02 5.466e+02 1.247e+03, threshold=8.532e+02, percent-clipped=3.0 2023-04-28 17:52:40,546 INFO [train.py:904] (5/8) Epoch 7, batch 7900, loss[loss=0.2549, simple_loss=0.3291, pruned_loss=0.09039, over 16753.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3179, pruned_loss=0.08132, over 3109389.34 frames. ], batch size: 124, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:53:58,005 INFO [train.py:904] (5/8) Epoch 7, batch 7950, loss[loss=0.2318, simple_loss=0.309, pruned_loss=0.07735, over 16127.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3175, pruned_loss=0.08115, over 3117229.76 frames. ], batch size: 165, lr: 9.55e-03, grad_scale: 2.0 2023-04-28 17:54:05,141 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9080, 5.2300, 4.9474, 4.9769, 4.6471, 4.5439, 4.6876, 5.2844], device='cuda:5'), covar=tensor([0.0738, 0.0626, 0.0859, 0.0506, 0.0654, 0.0737, 0.0794, 0.0775], device='cuda:5'), in_proj_covar=tensor([0.0447, 0.0562, 0.0481, 0.0372, 0.0353, 0.0383, 0.0468, 0.0418], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 17:54:27,422 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 17:55:03,364 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.354e+02 3.365e+02 3.868e+02 4.958e+02 9.415e+02, threshold=7.735e+02, percent-clipped=1.0 2023-04-28 17:55:12,758 INFO [train.py:904] (5/8) Epoch 7, batch 8000, loss[loss=0.2393, simple_loss=0.3158, pruned_loss=0.0814, over 16833.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3184, pruned_loss=0.08224, over 3097713.88 frames. ], batch size: 116, lr: 9.55e-03, grad_scale: 4.0 2023-04-28 17:55:28,993 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 17:55:37,516 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5966, 3.7550, 4.0167, 1.6303, 4.1904, 4.2605, 2.8352, 3.0433], device='cuda:5'), covar=tensor([0.0681, 0.0155, 0.0141, 0.1281, 0.0045, 0.0062, 0.0384, 0.0391], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0093, 0.0081, 0.0137, 0.0068, 0.0084, 0.0117, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 17:56:00,095 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:56:25,017 INFO [train.py:904] (5/8) Epoch 7, batch 8050, loss[loss=0.281, simple_loss=0.3347, pruned_loss=0.1137, over 11413.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3177, pruned_loss=0.08183, over 3095249.67 frames. ], batch size: 247, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:57:30,024 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.364e+02 3.694e+02 4.392e+02 5.257e+02 9.021e+02, threshold=8.783e+02, percent-clipped=4.0 2023-04-28 17:57:41,499 INFO [train.py:904] (5/8) Epoch 7, batch 8100, loss[loss=0.196, simple_loss=0.2783, pruned_loss=0.05683, over 17056.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.317, pruned_loss=0.08109, over 3091089.73 frames. ], batch size: 50, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:58:06,332 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-04-28 17:58:40,375 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:58:43,485 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69043.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:58:55,989 INFO [train.py:904] (5/8) Epoch 7, batch 8150, loss[loss=0.1949, simple_loss=0.2698, pruned_loss=0.05997, over 16959.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3144, pruned_loss=0.07999, over 3100097.63 frames. ], batch size: 55, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:59:00,702 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:59:59,700 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.706e+02 3.772e+02 4.723e+02 6.532e+02 2.181e+03, threshold=9.446e+02, percent-clipped=10.0 2023-04-28 18:00:10,161 INFO [train.py:904] (5/8) Epoch 7, batch 8200, loss[loss=0.246, simple_loss=0.3126, pruned_loss=0.08968, over 11567.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3117, pruned_loss=0.07857, over 3103421.18 frames. ], batch size: 248, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:00:12,049 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:00:13,590 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:21,780 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3851, 3.2720, 3.4010, 3.5175, 3.5468, 3.2033, 3.4961, 3.5928], device='cuda:5'), covar=tensor([0.0932, 0.0756, 0.1040, 0.0518, 0.0564, 0.2133, 0.0786, 0.0577], device='cuda:5'), in_proj_covar=tensor([0.0445, 0.0546, 0.0675, 0.0548, 0.0413, 0.0419, 0.0439, 0.0473], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 18:01:30,736 INFO [train.py:904] (5/8) Epoch 7, batch 8250, loss[loss=0.2005, simple_loss=0.2823, pruned_loss=0.05933, over 12012.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3101, pruned_loss=0.07626, over 3070869.00 frames. ], batch size: 247, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:01:47,247 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:58,188 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:02:40,829 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.107e+02 2.942e+02 3.716e+02 4.471e+02 8.485e+02, threshold=7.433e+02, percent-clipped=0.0 2023-04-28 18:02:52,449 INFO [train.py:904] (5/8) Epoch 7, batch 8300, loss[loss=0.1951, simple_loss=0.2894, pruned_loss=0.05046, over 16724.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.307, pruned_loss=0.07267, over 3063039.30 frames. ], batch size: 124, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:03:26,051 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:29,133 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:35,536 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:42,645 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:04:00,597 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:04:11,881 INFO [train.py:904] (5/8) Epoch 7, batch 8350, loss[loss=0.2308, simple_loss=0.301, pruned_loss=0.08025, over 11658.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.305, pruned_loss=0.06943, over 3058302.58 frames. ], batch size: 248, lr: 9.52e-03, grad_scale: 4.0 2023-04-28 18:04:46,417 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:00,052 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:07,198 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:20,096 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8488, 1.2093, 1.6474, 1.7549, 1.8464, 1.8414, 1.4403, 1.7824], device='cuda:5'), covar=tensor([0.0143, 0.0245, 0.0121, 0.0145, 0.0147, 0.0124, 0.0294, 0.0060], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0154, 0.0134, 0.0135, 0.0144, 0.0100, 0.0152, 0.0092], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 18:05:20,697 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.640e+02 3.244e+02 4.107e+02 9.415e+02, threshold=6.488e+02, percent-clipped=2.0 2023-04-28 18:05:32,118 INFO [train.py:904] (5/8) Epoch 7, batch 8400, loss[loss=0.2036, simple_loss=0.2864, pruned_loss=0.06038, over 16248.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3018, pruned_loss=0.06684, over 3063632.10 frames. ], batch size: 165, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:05:39,328 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:06:02,570 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:23,822 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:06:35,220 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:43,568 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-28 18:06:52,099 INFO [train.py:904] (5/8) Epoch 7, batch 8450, loss[loss=0.1921, simple_loss=0.2779, pruned_loss=0.05314, over 16709.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2998, pruned_loss=0.06477, over 3066403.23 frames. ], batch size: 57, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:07:18,717 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:38,857 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:50,678 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:58,565 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:59,250 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.708e+02 3.167e+02 3.913e+02 6.077e+02, threshold=6.334e+02, percent-clipped=0.0 2023-04-28 18:08:06,559 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:08:10,405 INFO [train.py:904] (5/8) Epoch 7, batch 8500, loss[loss=0.1974, simple_loss=0.2793, pruned_loss=0.05775, over 15213.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2957, pruned_loss=0.06212, over 3051643.29 frames. ], batch size: 191, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:08:55,068 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:09:33,176 INFO [train.py:904] (5/8) Epoch 7, batch 8550, loss[loss=0.2102, simple_loss=0.2848, pruned_loss=0.06773, over 11967.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2931, pruned_loss=0.06068, over 3044403.42 frames. ], batch size: 248, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:09:41,798 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:10:58,766 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 3.006e+02 3.588e+02 4.597e+02 9.217e+02, threshold=7.177e+02, percent-clipped=6.0 2023-04-28 18:11:12,033 INFO [train.py:904] (5/8) Epoch 7, batch 8600, loss[loss=0.2099, simple_loss=0.2959, pruned_loss=0.06193, over 16778.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2936, pruned_loss=0.05969, over 3058231.72 frames. ], batch size: 124, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:11:46,305 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:11:56,251 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:12:47,740 INFO [train.py:904] (5/8) Epoch 7, batch 8650, loss[loss=0.1877, simple_loss=0.2876, pruned_loss=0.04387, over 16787.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2918, pruned_loss=0.05809, over 3057608.60 frames. ], batch size: 76, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:13:19,539 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5138, 4.2248, 4.5196, 4.6885, 4.8492, 4.3331, 4.7913, 4.7813], device='cuda:5'), covar=tensor([0.1130, 0.0946, 0.1475, 0.0625, 0.0397, 0.0841, 0.0426, 0.0432], device='cuda:5'), in_proj_covar=tensor([0.0420, 0.0521, 0.0646, 0.0528, 0.0394, 0.0401, 0.0415, 0.0454], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 18:13:54,358 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:14:19,152 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.551e+02 2.997e+02 3.527e+02 6.166e+02, threshold=5.994e+02, percent-clipped=0.0 2023-04-28 18:14:31,909 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:14:32,613 INFO [train.py:904] (5/8) Epoch 7, batch 8700, loss[loss=0.1945, simple_loss=0.279, pruned_loss=0.05493, over 16264.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2886, pruned_loss=0.05662, over 3049686.37 frames. ], batch size: 165, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:14:36,527 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-28 18:15:22,788 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:16:08,633 INFO [train.py:904] (5/8) Epoch 7, batch 8750, loss[loss=0.2047, simple_loss=0.3032, pruned_loss=0.0531, over 16160.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2883, pruned_loss=0.05603, over 3057420.77 frames. ], batch size: 165, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:17:04,389 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:17:07,512 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:17:45,989 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.541e+02 2.592e+02 3.192e+02 4.101e+02 7.533e+02, threshold=6.383e+02, percent-clipped=6.0 2023-04-28 18:17:55,785 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:18:01,180 INFO [train.py:904] (5/8) Epoch 7, batch 8800, loss[loss=0.2354, simple_loss=0.3111, pruned_loss=0.07987, over 12759.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2868, pruned_loss=0.05487, over 3073348.89 frames. ], batch size: 248, lr: 9.49e-03, grad_scale: 8.0 2023-04-28 18:18:02,204 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5507, 3.5860, 3.3379, 3.1829, 3.2047, 3.4594, 3.2822, 3.2758], device='cuda:5'), covar=tensor([0.0467, 0.0357, 0.0206, 0.0189, 0.0593, 0.0317, 0.1043, 0.0466], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0228, 0.0229, 0.0204, 0.0252, 0.0229, 0.0161, 0.0265], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 18:18:47,925 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:09,550 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5610, 4.9000, 5.0231, 4.9682, 4.8933, 5.4322, 5.0435, 4.8632], device='cuda:5'), covar=tensor([0.0864, 0.1475, 0.1090, 0.1401, 0.2283, 0.0914, 0.1149, 0.2156], device='cuda:5'), in_proj_covar=tensor([0.0289, 0.0404, 0.0417, 0.0355, 0.0465, 0.0454, 0.0337, 0.0472], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 18:19:11,516 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:19:36,495 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:43,882 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:46,920 INFO [train.py:904] (5/8) Epoch 7, batch 8850, loss[loss=0.2019, simple_loss=0.3029, pruned_loss=0.05048, over 16158.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2886, pruned_loss=0.05418, over 3060340.84 frames. ], batch size: 165, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:21:03,139 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 18:21:21,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.840e+02 3.473e+02 4.436e+02 7.105e+02, threshold=6.946e+02, percent-clipped=4.0 2023-04-28 18:21:34,931 INFO [train.py:904] (5/8) Epoch 7, batch 8900, loss[loss=0.1758, simple_loss=0.2712, pruned_loss=0.04026, over 16834.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2871, pruned_loss=0.05278, over 3047138.11 frames. ], batch size: 96, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:21:49,818 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 18:22:06,702 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:22:21,960 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:23:38,927 INFO [train.py:904] (5/8) Epoch 7, batch 8950, loss[loss=0.1942, simple_loss=0.2791, pruned_loss=0.05461, over 12980.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2867, pruned_loss=0.05305, over 3077098.35 frames. ], batch size: 250, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:24:09,594 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:22,086 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:42,997 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:25:14,734 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.573e+02 2.911e+02 3.451e+02 7.486e+02, threshold=5.822e+02, percent-clipped=1.0 2023-04-28 18:25:27,349 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:25:27,974 INFO [train.py:904] (5/8) Epoch 7, batch 9000, loss[loss=0.185, simple_loss=0.2784, pruned_loss=0.04577, over 15447.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2838, pruned_loss=0.05189, over 3067705.75 frames. ], batch size: 191, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:25:27,974 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 18:25:37,201 INFO [train.py:938] (5/8) Epoch 7, validation: loss=0.1647, simple_loss=0.2682, pruned_loss=0.03062, over 944034.00 frames. 2023-04-28 18:25:37,202 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 18:26:33,167 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:26:33,320 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:27:14,337 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:27:19,720 INFO [train.py:904] (5/8) Epoch 7, batch 9050, loss[loss=0.1713, simple_loss=0.2602, pruned_loss=0.04118, over 16929.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2849, pruned_loss=0.05251, over 3077506.20 frames. ], batch size: 102, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:27:47,077 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 18:28:08,902 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:28:10,107 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:28:50,068 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.800e+02 3.566e+02 4.608e+02 8.238e+02, threshold=7.132e+02, percent-clipped=9.0 2023-04-28 18:29:06,540 INFO [train.py:904] (5/8) Epoch 7, batch 9100, loss[loss=0.2181, simple_loss=0.3073, pruned_loss=0.06444, over 15339.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2853, pruned_loss=0.05365, over 3081239.19 frames. ], batch size: 191, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:29:58,315 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:29:58,373 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:30:05,349 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 18:30:13,172 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:31:01,111 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:31:03,198 INFO [train.py:904] (5/8) Epoch 7, batch 9150, loss[loss=0.1613, simple_loss=0.2598, pruned_loss=0.03146, over 16844.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2855, pruned_loss=0.05294, over 3076403.19 frames. ], batch size: 90, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:31:31,603 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9207, 2.6018, 2.7205, 1.9068, 2.5117, 2.7099, 2.6085, 1.7653], device='cuda:5'), covar=tensor([0.0275, 0.0030, 0.0037, 0.0231, 0.0074, 0.0050, 0.0044, 0.0310], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0055, 0.0058, 0.0114, 0.0063, 0.0072, 0.0064, 0.0110], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 18:31:47,584 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:03,980 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4607, 4.5476, 4.6953, 4.5663, 4.5326, 5.1319, 4.7068, 4.3778], device='cuda:5'), covar=tensor([0.0956, 0.1505, 0.1337, 0.1598, 0.2214, 0.0809, 0.1128, 0.2183], device='cuda:5'), in_proj_covar=tensor([0.0283, 0.0396, 0.0414, 0.0350, 0.0457, 0.0439, 0.0332, 0.0468], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 18:32:35,335 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.681e+02 3.453e+02 4.314e+02 7.152e+02, threshold=6.905e+02, percent-clipped=1.0 2023-04-28 18:32:39,002 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:45,144 INFO [train.py:904] (5/8) Epoch 7, batch 9200, loss[loss=0.1916, simple_loss=0.2758, pruned_loss=0.05372, over 17087.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2816, pruned_loss=0.05201, over 3075412.53 frames. ], batch size: 53, lr: 9.47e-03, grad_scale: 8.0 2023-04-28 18:33:01,894 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:33:24,427 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 18:34:22,203 INFO [train.py:904] (5/8) Epoch 7, batch 9250, loss[loss=0.166, simple_loss=0.2488, pruned_loss=0.04158, over 12451.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2812, pruned_loss=0.052, over 3071171.66 frames. ], batch size: 248, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:35:02,143 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:35:16,845 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0074, 4.9690, 5.4661, 5.4793, 5.4186, 5.1038, 5.0683, 4.8158], device='cuda:5'), covar=tensor([0.0213, 0.0371, 0.0287, 0.0326, 0.0397, 0.0262, 0.0692, 0.0312], device='cuda:5'), in_proj_covar=tensor([0.0259, 0.0264, 0.0263, 0.0253, 0.0303, 0.0284, 0.0370, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 18:35:46,723 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3915, 3.5031, 1.8426, 3.7157, 2.4137, 3.6570, 1.8816, 2.6875], device='cuda:5'), covar=tensor([0.0177, 0.0280, 0.1506, 0.0098, 0.0762, 0.0432, 0.1541, 0.0601], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0147, 0.0174, 0.0090, 0.0155, 0.0178, 0.0183, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 18:36:01,191 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.738e+02 2.717e+02 3.344e+02 3.929e+02 8.889e+02, threshold=6.688e+02, percent-clipped=4.0 2023-04-28 18:36:02,292 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8172, 1.2651, 1.6061, 1.7893, 1.7808, 1.8704, 1.4249, 1.8448], device='cuda:5'), covar=tensor([0.0129, 0.0221, 0.0121, 0.0155, 0.0160, 0.0113, 0.0249, 0.0065], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0151, 0.0133, 0.0135, 0.0143, 0.0097, 0.0151, 0.0089], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 18:36:13,557 INFO [train.py:904] (5/8) Epoch 7, batch 9300, loss[loss=0.1859, simple_loss=0.2748, pruned_loss=0.04844, over 15370.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.28, pruned_loss=0.05149, over 3076592.11 frames. ], batch size: 192, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:37:58,359 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8872, 1.7686, 1.5148, 1.4749, 1.8561, 1.6386, 1.7604, 1.9246], device='cuda:5'), covar=tensor([0.0053, 0.0176, 0.0226, 0.0228, 0.0109, 0.0156, 0.0101, 0.0104], device='cuda:5'), in_proj_covar=tensor([0.0096, 0.0173, 0.0168, 0.0168, 0.0165, 0.0171, 0.0154, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 18:37:59,028 INFO [train.py:904] (5/8) Epoch 7, batch 9350, loss[loss=0.1934, simple_loss=0.2715, pruned_loss=0.05767, over 12095.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2797, pruned_loss=0.05139, over 3075407.66 frames. ], batch size: 247, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:38:01,339 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-28 18:38:18,325 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:38:59,982 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-28 18:39:01,529 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0180, 4.0485, 4.4495, 4.3981, 4.4149, 4.0842, 4.1656, 4.1014], device='cuda:5'), covar=tensor([0.0268, 0.0412, 0.0333, 0.0416, 0.0442, 0.0310, 0.0662, 0.0358], device='cuda:5'), in_proj_covar=tensor([0.0252, 0.0254, 0.0256, 0.0245, 0.0288, 0.0275, 0.0355, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 18:39:31,295 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.755e+02 2.689e+02 3.082e+02 3.847e+02 8.568e+02, threshold=6.165e+02, percent-clipped=3.0 2023-04-28 18:39:40,131 INFO [train.py:904] (5/8) Epoch 7, batch 9400, loss[loss=0.2223, simple_loss=0.3134, pruned_loss=0.06558, over 16841.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2802, pruned_loss=0.05125, over 3062029.46 frames. ], batch size: 116, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:40:18,704 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:40:36,621 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:41:20,568 INFO [train.py:904] (5/8) Epoch 7, batch 9450, loss[loss=0.1827, simple_loss=0.2681, pruned_loss=0.04859, over 12775.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2816, pruned_loss=0.05175, over 3054186.38 frames. ], batch size: 250, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:42:14,166 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:42:33,294 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5184, 3.1306, 2.9795, 1.8009, 2.6486, 2.1800, 3.0623, 3.1147], device='cuda:5'), covar=tensor([0.0259, 0.0589, 0.0527, 0.1585, 0.0729, 0.0866, 0.0625, 0.0777], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0123, 0.0152, 0.0139, 0.0132, 0.0123, 0.0132, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 18:42:50,918 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 2.717e+02 3.594e+02 4.518e+02 8.204e+02, threshold=7.189e+02, percent-clipped=4.0 2023-04-28 18:43:02,102 INFO [train.py:904] (5/8) Epoch 7, batch 9500, loss[loss=0.1726, simple_loss=0.2799, pruned_loss=0.0326, over 16890.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2808, pruned_loss=0.05124, over 3050859.95 frames. ], batch size: 102, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:43:35,909 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0808, 2.8476, 2.7705, 1.9510, 2.6344, 2.0808, 2.7892, 2.9098], device='cuda:5'), covar=tensor([0.0252, 0.0599, 0.0441, 0.1395, 0.0614, 0.0826, 0.0509, 0.0583], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0122, 0.0151, 0.0137, 0.0130, 0.0122, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 18:43:45,436 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 18:44:04,277 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 18:44:45,656 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2446, 3.3898, 3.4461, 1.5916, 3.7033, 3.6685, 2.8198, 2.8823], device='cuda:5'), covar=tensor([0.0813, 0.0172, 0.0166, 0.1330, 0.0053, 0.0103, 0.0395, 0.0388], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0093, 0.0079, 0.0138, 0.0065, 0.0082, 0.0116, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 18:44:47,824 INFO [train.py:904] (5/8) Epoch 7, batch 9550, loss[loss=0.228, simple_loss=0.3106, pruned_loss=0.07269, over 16442.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2804, pruned_loss=0.05123, over 3057824.95 frames. ], batch size: 146, lr: 9.44e-03, grad_scale: 4.0 2023-04-28 18:45:18,578 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:46:04,336 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5370, 3.6705, 1.5144, 3.8915, 2.4400, 3.7785, 1.7093, 2.7662], device='cuda:5'), covar=tensor([0.0149, 0.0192, 0.1764, 0.0078, 0.0784, 0.0374, 0.1678, 0.0560], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0147, 0.0175, 0.0090, 0.0156, 0.0178, 0.0183, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 18:46:18,737 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.839e+02 3.500e+02 4.170e+02 7.519e+02, threshold=6.999e+02, percent-clipped=3.0 2023-04-28 18:46:26,840 INFO [train.py:904] (5/8) Epoch 7, batch 9600, loss[loss=0.1961, simple_loss=0.2779, pruned_loss=0.05717, over 12550.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2826, pruned_loss=0.05234, over 3048457.08 frames. ], batch size: 247, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:48:15,882 INFO [train.py:904] (5/8) Epoch 7, batch 9650, loss[loss=0.1839, simple_loss=0.2866, pruned_loss=0.04063, over 16874.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2843, pruned_loss=0.05248, over 3048027.93 frames. ], batch size: 102, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:49:42,979 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 18:49:55,423 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.673e+02 3.296e+02 3.931e+02 1.187e+03, threshold=6.593e+02, percent-clipped=2.0 2023-04-28 18:50:06,089 INFO [train.py:904] (5/8) Epoch 7, batch 9700, loss[loss=0.1966, simple_loss=0.2837, pruned_loss=0.05474, over 16682.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2831, pruned_loss=0.05207, over 3050857.83 frames. ], batch size: 134, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:50:33,469 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:51:47,798 INFO [train.py:904] (5/8) Epoch 7, batch 9750, loss[loss=0.1837, simple_loss=0.2627, pruned_loss=0.0524, over 12280.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2819, pruned_loss=0.05197, over 3057817.47 frames. ], batch size: 250, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:53:01,604 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-28 18:53:18,484 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.655e+02 3.090e+02 3.865e+02 1.050e+03, threshold=6.180e+02, percent-clipped=3.0 2023-04-28 18:53:27,257 INFO [train.py:904] (5/8) Epoch 7, batch 9800, loss[loss=0.1986, simple_loss=0.2911, pruned_loss=0.05303, over 16924.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2816, pruned_loss=0.05075, over 3062663.85 frames. ], batch size: 116, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:55:12,749 INFO [train.py:904] (5/8) Epoch 7, batch 9850, loss[loss=0.1889, simple_loss=0.2872, pruned_loss=0.04526, over 15198.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2831, pruned_loss=0.05086, over 3068898.85 frames. ], batch size: 191, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:55:43,262 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:56:38,790 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5836, 2.4769, 2.3619, 3.9169, 2.7171, 3.9016, 1.2593, 2.8458], device='cuda:5'), covar=tensor([0.1397, 0.0652, 0.1085, 0.0095, 0.0143, 0.0329, 0.1576, 0.0716], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0145, 0.0167, 0.0102, 0.0170, 0.0192, 0.0169, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 18:56:54,663 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.569e+02 3.072e+02 3.785e+02 9.202e+02, threshold=6.144e+02, percent-clipped=3.0 2023-04-28 18:57:04,349 INFO [train.py:904] (5/8) Epoch 7, batch 9900, loss[loss=0.1939, simple_loss=0.2735, pruned_loss=0.05711, over 12147.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2834, pruned_loss=0.0509, over 3056633.22 frames. ], batch size: 248, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:57:33,475 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:57:35,123 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0082, 3.1351, 1.7549, 3.2822, 2.2504, 3.2241, 1.8462, 2.5382], device='cuda:5'), covar=tensor([0.0202, 0.0358, 0.1601, 0.0116, 0.0794, 0.0523, 0.1565, 0.0683], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0149, 0.0177, 0.0091, 0.0160, 0.0180, 0.0188, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 18:58:35,190 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 18:59:01,234 INFO [train.py:904] (5/8) Epoch 7, batch 9950, loss[loss=0.1844, simple_loss=0.2828, pruned_loss=0.04296, over 16918.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2859, pruned_loss=0.05111, over 3075274.74 frames. ], batch size: 125, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 19:00:47,807 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.726e+02 3.249e+02 3.973e+02 1.098e+03, threshold=6.498e+02, percent-clipped=6.0 2023-04-28 19:01:00,574 INFO [train.py:904] (5/8) Epoch 7, batch 10000, loss[loss=0.1623, simple_loss=0.2654, pruned_loss=0.02965, over 16886.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.284, pruned_loss=0.05062, over 3091151.37 frames. ], batch size: 96, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:01:29,753 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:02:39,088 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 19:02:40,828 INFO [train.py:904] (5/8) Epoch 7, batch 10050, loss[loss=0.1864, simple_loss=0.2852, pruned_loss=0.04385, over 16343.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2836, pruned_loss=0.05036, over 3072382.71 frames. ], batch size: 146, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:03:04,296 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:03:04,465 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4237, 3.6997, 3.8439, 2.7796, 3.4834, 3.6479, 3.5826, 1.8797], device='cuda:5'), covar=tensor([0.0349, 0.0020, 0.0024, 0.0229, 0.0058, 0.0051, 0.0039, 0.0395], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0057, 0.0060, 0.0117, 0.0066, 0.0073, 0.0066, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 19:03:17,371 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8858, 1.9975, 2.2989, 3.2672, 2.0577, 2.3018, 2.2789, 2.0075], device='cuda:5'), covar=tensor([0.0647, 0.2547, 0.1377, 0.0392, 0.3099, 0.1667, 0.2143, 0.2771], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0334, 0.0289, 0.0302, 0.0378, 0.0360, 0.0304, 0.0392], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:03:56,306 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70992.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:04:04,313 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.613e+02 3.026e+02 3.875e+02 7.173e+02, threshold=6.053e+02, percent-clipped=4.0 2023-04-28 19:04:12,152 INFO [train.py:904] (5/8) Epoch 7, batch 10100, loss[loss=0.2079, simple_loss=0.2999, pruned_loss=0.05798, over 16739.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2844, pruned_loss=0.05091, over 3089691.76 frames. ], batch size: 134, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:05:07,064 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:05:54,945 INFO [train.py:904] (5/8) Epoch 8, batch 0, loss[loss=0.3182, simple_loss=0.3604, pruned_loss=0.138, over 16227.00 frames. ], tot_loss[loss=0.3182, simple_loss=0.3604, pruned_loss=0.138, over 16227.00 frames. ], batch size: 165, lr: 8.86e-03, grad_scale: 8.0 2023-04-28 19:05:54,945 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 19:06:02,579 INFO [train.py:938] (5/8) Epoch 8, validation: loss=0.1627, simple_loss=0.2663, pruned_loss=0.02958, over 944034.00 frames. 2023-04-28 19:06:02,580 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 19:06:04,012 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:06:28,441 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 19:06:55,611 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71091.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:07:07,692 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 3.480e+02 4.237e+02 5.161e+02 1.332e+03, threshold=8.475e+02, percent-clipped=15.0 2023-04-28 19:07:09,992 INFO [train.py:904] (5/8) Epoch 8, batch 50, loss[loss=0.2065, simple_loss=0.2943, pruned_loss=0.0594, over 17077.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3063, pruned_loss=0.07794, over 740380.81 frames. ], batch size: 50, lr: 8.86e-03, grad_scale: 1.0 2023-04-28 19:07:28,549 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1096, 4.4653, 3.4692, 2.5839, 3.0852, 2.4330, 4.7214, 4.1286], device='cuda:5'), covar=tensor([0.1998, 0.0519, 0.1194, 0.1725, 0.2035, 0.1524, 0.0297, 0.0730], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0239, 0.0268, 0.0251, 0.0243, 0.0204, 0.0242, 0.0257], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:08:17,892 INFO [train.py:904] (5/8) Epoch 8, batch 100, loss[loss=0.1856, simple_loss=0.2724, pruned_loss=0.0494, over 17234.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.298, pruned_loss=0.07186, over 1309935.82 frames. ], batch size: 43, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:08:34,859 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 19:09:20,755 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6893, 4.5827, 5.2172, 5.2273, 5.2560, 4.8248, 4.7947, 4.5445], device='cuda:5'), covar=tensor([0.0287, 0.0433, 0.0448, 0.0439, 0.0473, 0.0322, 0.0860, 0.0428], device='cuda:5'), in_proj_covar=tensor([0.0265, 0.0274, 0.0273, 0.0258, 0.0308, 0.0292, 0.0380, 0.0237], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 19:09:23,833 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 2.858e+02 3.485e+02 4.000e+02 7.217e+02, threshold=6.971e+02, percent-clipped=0.0 2023-04-28 19:09:26,729 INFO [train.py:904] (5/8) Epoch 8, batch 150, loss[loss=0.2132, simple_loss=0.3, pruned_loss=0.06321, over 17247.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.295, pruned_loss=0.06975, over 1756619.52 frames. ], batch size: 45, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:10:06,101 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 19:10:31,468 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9273, 4.8808, 5.4345, 5.4503, 5.4773, 5.0310, 4.9110, 4.7348], device='cuda:5'), covar=tensor([0.0223, 0.0390, 0.0389, 0.0367, 0.0352, 0.0260, 0.0797, 0.0399], device='cuda:5'), in_proj_covar=tensor([0.0265, 0.0275, 0.0275, 0.0259, 0.0309, 0.0293, 0.0384, 0.0238], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 19:10:33,459 INFO [train.py:904] (5/8) Epoch 8, batch 200, loss[loss=0.1873, simple_loss=0.2744, pruned_loss=0.05008, over 17242.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.293, pruned_loss=0.06829, over 2107337.69 frames. ], batch size: 52, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:11:04,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9653, 4.0957, 3.8674, 3.8187, 3.2319, 4.0660, 3.9789, 3.6736], device='cuda:5'), covar=tensor([0.0794, 0.0489, 0.0450, 0.0343, 0.1539, 0.0436, 0.0763, 0.0681], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0245, 0.0244, 0.0216, 0.0272, 0.0247, 0.0166, 0.0284], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:11:39,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1347, 5.1284, 4.8489, 4.3056, 4.8733, 1.6597, 4.6954, 4.9184], device='cuda:5'), covar=tensor([0.0059, 0.0049, 0.0121, 0.0290, 0.0066, 0.2142, 0.0090, 0.0136], device='cuda:5'), in_proj_covar=tensor([0.0107, 0.0094, 0.0144, 0.0132, 0.0109, 0.0163, 0.0127, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:11:40,019 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.597e+02 3.114e+02 3.652e+02 1.080e+03, threshold=6.228e+02, percent-clipped=1.0 2023-04-28 19:11:42,955 INFO [train.py:904] (5/8) Epoch 8, batch 250, loss[loss=0.1829, simple_loss=0.2597, pruned_loss=0.05303, over 16779.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2885, pruned_loss=0.06672, over 2377667.86 frames. ], batch size: 39, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:11:43,366 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3413, 3.1579, 3.2826, 3.4286, 3.5064, 3.2504, 3.4315, 3.5675], device='cuda:5'), covar=tensor([0.0857, 0.0797, 0.1042, 0.0567, 0.0562, 0.2027, 0.0943, 0.0581], device='cuda:5'), in_proj_covar=tensor([0.0452, 0.0568, 0.0693, 0.0565, 0.0427, 0.0429, 0.0441, 0.0491], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:11:56,949 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8631, 4.0446, 3.0895, 2.3701, 2.7953, 2.2929, 4.3121, 3.7924], device='cuda:5'), covar=tensor([0.2345, 0.0634, 0.1341, 0.1843, 0.2247, 0.1640, 0.0377, 0.0878], device='cuda:5'), in_proj_covar=tensor([0.0286, 0.0245, 0.0275, 0.0257, 0.0258, 0.0209, 0.0250, 0.0269], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:12:02,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1245, 4.0863, 4.5126, 3.3586, 4.0397, 4.4205, 4.1486, 2.6914], device='cuda:5'), covar=tensor([0.0287, 0.0036, 0.0022, 0.0203, 0.0047, 0.0043, 0.0034, 0.0288], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0063, 0.0062, 0.0119, 0.0066, 0.0075, 0.0067, 0.0115], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 19:12:47,183 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:12:52,131 INFO [train.py:904] (5/8) Epoch 8, batch 300, loss[loss=0.2036, simple_loss=0.2746, pruned_loss=0.06634, over 16187.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.284, pruned_loss=0.06307, over 2587320.86 frames. ], batch size: 165, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:13:39,945 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71386.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:13:43,975 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 19:13:58,501 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.511e+02 3.030e+02 3.934e+02 1.129e+03, threshold=6.060e+02, percent-clipped=6.0 2023-04-28 19:14:01,161 INFO [train.py:904] (5/8) Epoch 8, batch 350, loss[loss=0.1981, simple_loss=0.2843, pruned_loss=0.05591, over 17095.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2813, pruned_loss=0.06153, over 2741797.42 frames. ], batch size: 55, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:14:02,931 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0321, 2.2216, 2.2714, 4.8212, 2.0724, 2.8336, 2.3734, 2.4350], device='cuda:5'), covar=tensor([0.0595, 0.2824, 0.1788, 0.0250, 0.3468, 0.1799, 0.2355, 0.2980], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0351, 0.0298, 0.0317, 0.0389, 0.0384, 0.0317, 0.0416], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:14:24,019 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 19:15:10,458 INFO [train.py:904] (5/8) Epoch 8, batch 400, loss[loss=0.1658, simple_loss=0.2509, pruned_loss=0.04036, over 17256.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2791, pruned_loss=0.06154, over 2873862.78 frames. ], batch size: 44, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:15:36,000 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8852, 3.2885, 2.7309, 4.6450, 4.0422, 4.4594, 1.6740, 3.1493], device='cuda:5'), covar=tensor([0.1313, 0.0524, 0.1077, 0.0131, 0.0336, 0.0378, 0.1384, 0.0774], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0149, 0.0171, 0.0109, 0.0184, 0.0200, 0.0171, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 19:16:06,027 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1023, 3.0848, 3.1421, 1.6322, 3.3375, 3.3748, 2.6801, 2.5644], device='cuda:5'), covar=tensor([0.0834, 0.0164, 0.0189, 0.1217, 0.0078, 0.0139, 0.0396, 0.0488], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0092, 0.0081, 0.0138, 0.0067, 0.0088, 0.0117, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 19:16:17,823 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.026e+02 2.889e+02 3.429e+02 4.074e+02 1.363e+03, threshold=6.859e+02, percent-clipped=4.0 2023-04-28 19:16:20,159 INFO [train.py:904] (5/8) Epoch 8, batch 450, loss[loss=0.2029, simple_loss=0.2668, pruned_loss=0.06955, over 16856.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2777, pruned_loss=0.06022, over 2973601.68 frames. ], batch size: 116, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:17:28,373 INFO [train.py:904] (5/8) Epoch 8, batch 500, loss[loss=0.1935, simple_loss=0.2895, pruned_loss=0.04876, over 17112.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2761, pruned_loss=0.05935, over 3058098.38 frames. ], batch size: 48, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:18:33,695 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.499e+02 3.026e+02 3.865e+02 9.130e+02, threshold=6.053e+02, percent-clipped=1.0 2023-04-28 19:18:37,356 INFO [train.py:904] (5/8) Epoch 8, batch 550, loss[loss=0.2079, simple_loss=0.277, pruned_loss=0.06943, over 16823.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2758, pruned_loss=0.05876, over 3124683.40 frames. ], batch size: 102, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:18:38,115 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 19:18:56,791 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:31,947 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 19:19:41,725 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:45,783 INFO [train.py:904] (5/8) Epoch 8, batch 600, loss[loss=0.1837, simple_loss=0.2746, pruned_loss=0.04641, over 17204.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2748, pruned_loss=0.05903, over 3152463.37 frames. ], batch size: 44, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:20:11,455 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 19:20:19,096 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:34,069 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:45,947 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=71696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:50,943 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.476e+02 3.033e+02 3.542e+02 7.789e+02, threshold=6.066e+02, percent-clipped=2.0 2023-04-28 19:20:53,327 INFO [train.py:904] (5/8) Epoch 8, batch 650, loss[loss=0.183, simple_loss=0.2688, pruned_loss=0.04863, over 16297.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2735, pruned_loss=0.05813, over 3196156.97 frames. ], batch size: 36, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:21:33,459 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 19:21:37,504 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:22:01,898 INFO [train.py:904] (5/8) Epoch 8, batch 700, loss[loss=0.2093, simple_loss=0.2821, pruned_loss=0.0683, over 16653.00 frames. ], tot_loss[loss=0.195, simple_loss=0.274, pruned_loss=0.05795, over 3232047.40 frames. ], batch size: 68, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:22:07,175 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1665, 1.6976, 2.3407, 3.1229, 2.8250, 3.2649, 2.1467, 3.3321], device='cuda:5'), covar=tensor([0.0101, 0.0302, 0.0218, 0.0133, 0.0157, 0.0127, 0.0276, 0.0076], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0157, 0.0142, 0.0140, 0.0149, 0.0103, 0.0156, 0.0093], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 19:22:15,890 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4708, 5.8109, 5.4719, 5.6549, 5.1842, 5.0370, 5.2656, 5.9222], device='cuda:5'), covar=tensor([0.1022, 0.0853, 0.1067, 0.0585, 0.0757, 0.0665, 0.0831, 0.0823], device='cuda:5'), in_proj_covar=tensor([0.0479, 0.0612, 0.0509, 0.0412, 0.0390, 0.0405, 0.0510, 0.0452], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:22:43,834 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.8028, 6.1382, 5.8012, 5.9490, 5.4943, 5.1071, 5.5786, 6.2247], device='cuda:5'), covar=tensor([0.0833, 0.0755, 0.1049, 0.0573, 0.0687, 0.0718, 0.0845, 0.0659], device='cuda:5'), in_proj_covar=tensor([0.0476, 0.0609, 0.0507, 0.0409, 0.0389, 0.0404, 0.0509, 0.0449], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:23:05,956 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.350e+02 2.585e+02 3.182e+02 3.748e+02 9.363e+02, threshold=6.364e+02, percent-clipped=4.0 2023-04-28 19:23:08,643 INFO [train.py:904] (5/8) Epoch 8, batch 750, loss[loss=0.2253, simple_loss=0.2916, pruned_loss=0.07947, over 16510.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2741, pruned_loss=0.05811, over 3251726.87 frames. ], batch size: 146, lr: 8.81e-03, grad_scale: 2.0 2023-04-28 19:24:17,953 INFO [train.py:904] (5/8) Epoch 8, batch 800, loss[loss=0.1595, simple_loss=0.24, pruned_loss=0.03953, over 17021.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2733, pruned_loss=0.05735, over 3271921.85 frames. ], batch size: 41, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:25:07,640 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 19:25:23,720 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.663e+02 3.116e+02 3.838e+02 7.032e+02, threshold=6.231e+02, percent-clipped=1.0 2023-04-28 19:25:25,982 INFO [train.py:904] (5/8) Epoch 8, batch 850, loss[loss=0.2174, simple_loss=0.3044, pruned_loss=0.06524, over 16711.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2732, pruned_loss=0.05736, over 3283485.35 frames. ], batch size: 57, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:25:49,756 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:26:15,637 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 19:26:32,721 INFO [train.py:904] (5/8) Epoch 8, batch 900, loss[loss=0.2259, simple_loss=0.2866, pruned_loss=0.0826, over 16902.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2724, pruned_loss=0.05714, over 3297751.81 frames. ], batch size: 109, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:26:39,967 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6454, 4.1034, 4.2187, 1.8911, 4.5133, 4.5617, 3.2552, 3.3743], device='cuda:5'), covar=tensor([0.0776, 0.0147, 0.0165, 0.1153, 0.0049, 0.0084, 0.0363, 0.0345], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0093, 0.0082, 0.0139, 0.0067, 0.0089, 0.0118, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 19:26:48,557 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 19:27:00,573 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:13,375 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:38,438 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.573e+02 3.042e+02 3.857e+02 8.032e+02, threshold=6.083e+02, percent-clipped=5.0 2023-04-28 19:27:44,223 INFO [train.py:904] (5/8) Epoch 8, batch 950, loss[loss=0.1972, simple_loss=0.2637, pruned_loss=0.06539, over 16779.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2729, pruned_loss=0.05718, over 3290404.37 frames. ], batch size: 102, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:28:52,162 INFO [train.py:904] (5/8) Epoch 8, batch 1000, loss[loss=0.2091, simple_loss=0.2751, pruned_loss=0.07155, over 16766.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2714, pruned_loss=0.05685, over 3299157.13 frames. ], batch size: 124, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:29:09,324 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6810, 3.1725, 2.8580, 1.8336, 2.5683, 2.2213, 3.2186, 3.1365], device='cuda:5'), covar=tensor([0.0316, 0.0656, 0.0689, 0.1735, 0.0854, 0.0902, 0.0699, 0.0801], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0133, 0.0153, 0.0138, 0.0132, 0.0123, 0.0134, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 19:29:58,312 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.648e+02 3.149e+02 3.724e+02 6.885e+02, threshold=6.299e+02, percent-clipped=3.0 2023-04-28 19:30:01,630 INFO [train.py:904] (5/8) Epoch 8, batch 1050, loss[loss=0.1954, simple_loss=0.2808, pruned_loss=0.05498, over 16757.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2714, pruned_loss=0.0573, over 3310937.22 frames. ], batch size: 57, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:30:27,773 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9803, 1.6045, 2.3171, 2.8428, 2.7967, 3.2979, 1.7825, 3.2436], device='cuda:5'), covar=tensor([0.0117, 0.0342, 0.0205, 0.0173, 0.0155, 0.0106, 0.0303, 0.0092], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0159, 0.0143, 0.0143, 0.0150, 0.0105, 0.0156, 0.0096], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 19:30:31,229 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4163, 5.7776, 5.5488, 5.5953, 5.1942, 5.0607, 5.2525, 5.8961], device='cuda:5'), covar=tensor([0.0989, 0.0932, 0.1037, 0.0592, 0.0813, 0.0704, 0.0856, 0.0866], device='cuda:5'), in_proj_covar=tensor([0.0481, 0.0619, 0.0512, 0.0414, 0.0392, 0.0406, 0.0515, 0.0457], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:30:40,185 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:31:10,580 INFO [train.py:904] (5/8) Epoch 8, batch 1100, loss[loss=0.1913, simple_loss=0.2841, pruned_loss=0.04919, over 16738.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2708, pruned_loss=0.05663, over 3312606.03 frames. ], batch size: 57, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:32:03,602 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:32:15,691 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.462e+02 2.908e+02 3.717e+02 7.729e+02, threshold=5.817e+02, percent-clipped=1.0 2023-04-28 19:32:18,240 INFO [train.py:904] (5/8) Epoch 8, batch 1150, loss[loss=0.2015, simple_loss=0.2908, pruned_loss=0.0561, over 17143.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2703, pruned_loss=0.05612, over 3312626.87 frames. ], batch size: 49, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:32:39,443 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 19:32:41,375 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4629, 2.3343, 1.6972, 1.9428, 2.7951, 2.5283, 3.4890, 3.0380], device='cuda:5'), covar=tensor([0.0055, 0.0297, 0.0410, 0.0378, 0.0174, 0.0297, 0.0134, 0.0166], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0187, 0.0183, 0.0182, 0.0181, 0.0189, 0.0184, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:33:26,642 INFO [train.py:904] (5/8) Epoch 8, batch 1200, loss[loss=0.192, simple_loss=0.2811, pruned_loss=0.05144, over 16603.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2691, pruned_loss=0.05601, over 3308954.92 frames. ], batch size: 62, lr: 8.79e-03, grad_scale: 8.0 2023-04-28 19:33:33,431 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5188, 4.4408, 4.3952, 3.8971, 4.4188, 1.7942, 4.1539, 4.1125], device='cuda:5'), covar=tensor([0.0073, 0.0068, 0.0121, 0.0260, 0.0063, 0.1946, 0.0112, 0.0158], device='cuda:5'), in_proj_covar=tensor([0.0113, 0.0101, 0.0152, 0.0145, 0.0117, 0.0165, 0.0138, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:33:50,407 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3272, 3.5250, 3.2459, 1.9883, 2.8886, 2.5322, 3.7621, 3.6078], device='cuda:5'), covar=tensor([0.0225, 0.0657, 0.0618, 0.1614, 0.0725, 0.0864, 0.0458, 0.0707], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0136, 0.0155, 0.0140, 0.0135, 0.0124, 0.0136, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 19:33:53,190 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:33:56,994 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72275.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:34:30,640 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.822e+02 3.117e+02 3.864e+02 9.153e+02, threshold=6.234e+02, percent-clipped=4.0 2023-04-28 19:34:32,976 INFO [train.py:904] (5/8) Epoch 8, batch 1250, loss[loss=0.1944, simple_loss=0.2707, pruned_loss=0.05908, over 16465.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.27, pruned_loss=0.05695, over 3314214.59 frames. ], batch size: 68, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:34:42,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1624, 2.0827, 1.6248, 1.9125, 2.4219, 2.2653, 2.4484, 2.5542], device='cuda:5'), covar=tensor([0.0122, 0.0240, 0.0324, 0.0279, 0.0125, 0.0193, 0.0167, 0.0134], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0184, 0.0180, 0.0180, 0.0179, 0.0187, 0.0183, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:34:58,484 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:06,157 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:41,555 INFO [train.py:904] (5/8) Epoch 8, batch 1300, loss[loss=0.2085, simple_loss=0.2846, pruned_loss=0.06615, over 15569.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2699, pruned_loss=0.05676, over 3318949.42 frames. ], batch size: 191, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:36:30,419 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72387.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:36:49,365 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.738e+02 2.612e+02 3.072e+02 3.829e+02 7.494e+02, threshold=6.144e+02, percent-clipped=3.0 2023-04-28 19:36:52,128 INFO [train.py:904] (5/8) Epoch 8, batch 1350, loss[loss=0.1988, simple_loss=0.2652, pruned_loss=0.06622, over 16453.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2698, pruned_loss=0.05614, over 3321340.72 frames. ], batch size: 146, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:36:52,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4920, 4.0160, 4.1232, 2.1785, 3.4091, 2.6359, 3.9313, 3.7827], device='cuda:5'), covar=tensor([0.0282, 0.0555, 0.0435, 0.1546, 0.0627, 0.0839, 0.0548, 0.0925], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0135, 0.0154, 0.0139, 0.0133, 0.0123, 0.0135, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 19:38:01,935 INFO [train.py:904] (5/8) Epoch 8, batch 1400, loss[loss=0.1909, simple_loss=0.2704, pruned_loss=0.05572, over 16797.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2693, pruned_loss=0.05607, over 3308317.60 frames. ], batch size: 83, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:38:10,055 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-28 19:38:47,784 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72485.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:38:48,078 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8149, 2.0096, 2.1367, 4.5740, 1.9520, 2.7321, 2.2264, 2.3043], device='cuda:5'), covar=tensor([0.0711, 0.3302, 0.2011, 0.0322, 0.3694, 0.2002, 0.2709, 0.3332], device='cuda:5'), in_proj_covar=tensor([0.0350, 0.0366, 0.0308, 0.0328, 0.0399, 0.0407, 0.0330, 0.0435], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:39:07,475 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.583e+02 2.954e+02 3.513e+02 6.005e+02, threshold=5.909e+02, percent-clipped=0.0 2023-04-28 19:39:11,106 INFO [train.py:904] (5/8) Epoch 8, batch 1450, loss[loss=0.1776, simple_loss=0.2627, pruned_loss=0.0462, over 17116.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2687, pruned_loss=0.05538, over 3320761.36 frames. ], batch size: 47, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:39:55,114 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3293, 3.6155, 3.2797, 2.0151, 2.9067, 2.5343, 3.6830, 3.7248], device='cuda:5'), covar=tensor([0.0202, 0.0554, 0.0616, 0.1481, 0.0729, 0.0837, 0.0486, 0.0638], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0137, 0.0154, 0.0140, 0.0134, 0.0124, 0.0135, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 19:40:20,581 INFO [train.py:904] (5/8) Epoch 8, batch 1500, loss[loss=0.2065, simple_loss=0.2672, pruned_loss=0.07288, over 16892.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2682, pruned_loss=0.05512, over 3325518.51 frames. ], batch size: 109, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:40:40,482 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8038, 3.2273, 2.4992, 4.4534, 3.9055, 4.3116, 1.5479, 3.0174], device='cuda:5'), covar=tensor([0.1227, 0.0443, 0.1036, 0.0105, 0.0202, 0.0354, 0.1310, 0.0690], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0152, 0.0172, 0.0115, 0.0194, 0.0204, 0.0170, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 19:40:52,037 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:41:25,450 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.563e+02 3.021e+02 3.528e+02 6.404e+02, threshold=6.041e+02, percent-clipped=2.0 2023-04-28 19:41:28,331 INFO [train.py:904] (5/8) Epoch 8, batch 1550, loss[loss=0.2161, simple_loss=0.2766, pruned_loss=0.07775, over 16891.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2697, pruned_loss=0.05635, over 3331744.05 frames. ], batch size: 96, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:41:50,961 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7913, 3.8679, 3.0325, 2.2756, 2.7280, 2.2243, 3.9241, 3.6541], device='cuda:5'), covar=tensor([0.1986, 0.0519, 0.1262, 0.2008, 0.2043, 0.1593, 0.0414, 0.1034], device='cuda:5'), in_proj_covar=tensor([0.0286, 0.0254, 0.0275, 0.0259, 0.0274, 0.0211, 0.0254, 0.0281], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:41:58,137 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:42:39,336 INFO [train.py:904] (5/8) Epoch 8, batch 1600, loss[loss=0.227, simple_loss=0.3004, pruned_loss=0.07684, over 16524.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2721, pruned_loss=0.05754, over 3328300.57 frames. ], batch size: 76, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:43:20,434 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:43:44,410 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.814e+02 3.483e+02 4.056e+02 7.586e+02, threshold=6.967e+02, percent-clipped=5.0 2023-04-28 19:43:47,307 INFO [train.py:904] (5/8) Epoch 8, batch 1650, loss[loss=0.179, simple_loss=0.2671, pruned_loss=0.04548, over 17123.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2736, pruned_loss=0.05801, over 3315107.15 frames. ], batch size: 49, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:44:58,254 INFO [train.py:904] (5/8) Epoch 8, batch 1700, loss[loss=0.1901, simple_loss=0.2677, pruned_loss=0.05626, over 16835.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2757, pruned_loss=0.05903, over 3314152.43 frames. ], batch size: 102, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:45:44,103 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:46:05,113 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.648e+02 3.116e+02 3.582e+02 6.867e+02, threshold=6.231e+02, percent-clipped=0.0 2023-04-28 19:46:07,550 INFO [train.py:904] (5/8) Epoch 8, batch 1750, loss[loss=0.1878, simple_loss=0.2735, pruned_loss=0.05105, over 17220.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2758, pruned_loss=0.05883, over 3322060.15 frames. ], batch size: 44, lr: 8.75e-03, grad_scale: 8.0 2023-04-28 19:46:12,298 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5223, 4.3476, 4.5837, 4.7986, 4.8936, 4.4560, 4.7404, 4.8605], device='cuda:5'), covar=tensor([0.1263, 0.1004, 0.1378, 0.0598, 0.0517, 0.0880, 0.1104, 0.0613], device='cuda:5'), in_proj_covar=tensor([0.0490, 0.0612, 0.0760, 0.0616, 0.0463, 0.0466, 0.0476, 0.0523], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:46:21,544 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8969, 3.8783, 3.8896, 3.0354, 3.8932, 1.7303, 3.6454, 3.3835], device='cuda:5'), covar=tensor([0.0131, 0.0105, 0.0170, 0.0473, 0.0097, 0.2586, 0.0150, 0.0298], device='cuda:5'), in_proj_covar=tensor([0.0113, 0.0102, 0.0153, 0.0147, 0.0118, 0.0164, 0.0139, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:46:50,159 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:47:03,675 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7046, 4.6894, 5.2078, 5.1874, 5.1733, 4.8198, 4.7643, 4.4873], device='cuda:5'), covar=tensor([0.0284, 0.0431, 0.0312, 0.0397, 0.0407, 0.0312, 0.0810, 0.0477], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0311, 0.0310, 0.0295, 0.0351, 0.0329, 0.0431, 0.0266], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 19:47:16,121 INFO [train.py:904] (5/8) Epoch 8, batch 1800, loss[loss=0.2247, simple_loss=0.3095, pruned_loss=0.06991, over 16521.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2774, pruned_loss=0.05882, over 3307825.16 frames. ], batch size: 68, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:47:27,959 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4848, 3.3799, 3.4668, 2.8946, 3.4037, 1.9762, 3.1745, 2.8524], device='cuda:5'), covar=tensor([0.0106, 0.0101, 0.0146, 0.0257, 0.0081, 0.1968, 0.0122, 0.0207], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0102, 0.0154, 0.0148, 0.0119, 0.0166, 0.0140, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:48:11,274 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9655, 2.2007, 2.3070, 4.5945, 2.0772, 2.9592, 2.4173, 2.4734], device='cuda:5'), covar=tensor([0.0621, 0.2945, 0.1782, 0.0302, 0.3440, 0.1821, 0.2380, 0.2954], device='cuda:5'), in_proj_covar=tensor([0.0346, 0.0362, 0.0305, 0.0324, 0.0392, 0.0403, 0.0325, 0.0429], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:48:15,255 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:48:24,991 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 2.735e+02 3.300e+02 4.209e+02 8.740e+02, threshold=6.600e+02, percent-clipped=4.0 2023-04-28 19:48:26,871 INFO [train.py:904] (5/8) Epoch 8, batch 1850, loss[loss=0.216, simple_loss=0.2989, pruned_loss=0.0665, over 16735.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2786, pruned_loss=0.05877, over 3310110.40 frames. ], batch size: 62, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:49:35,234 INFO [train.py:904] (5/8) Epoch 8, batch 1900, loss[loss=0.1708, simple_loss=0.2533, pruned_loss=0.04415, over 16805.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2777, pruned_loss=0.05832, over 3320352.83 frames. ], batch size: 39, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:49:39,759 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:49:52,051 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 19:50:12,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3422, 5.2520, 5.1154, 4.5376, 5.0991, 1.9004, 4.9784, 5.1354], device='cuda:5'), covar=tensor([0.0054, 0.0050, 0.0110, 0.0294, 0.0062, 0.1962, 0.0089, 0.0113], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0104, 0.0157, 0.0150, 0.0121, 0.0169, 0.0142, 0.0148], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:50:16,620 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:50:41,099 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.549e+02 2.878e+02 3.342e+02 6.835e+02, threshold=5.756e+02, percent-clipped=2.0 2023-04-28 19:50:42,532 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-28 19:50:43,030 INFO [train.py:904] (5/8) Epoch 8, batch 1950, loss[loss=0.1634, simple_loss=0.2514, pruned_loss=0.03767, over 16837.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2777, pruned_loss=0.05812, over 3315168.59 frames. ], batch size: 42, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:51:14,283 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 19:51:21,297 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:51:50,197 INFO [train.py:904] (5/8) Epoch 8, batch 2000, loss[loss=0.2219, simple_loss=0.3074, pruned_loss=0.06818, over 17023.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2772, pruned_loss=0.05769, over 3317074.08 frames. ], batch size: 55, lr: 8.74e-03, grad_scale: 8.0 2023-04-28 19:52:11,915 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5031, 2.4955, 2.1356, 2.4438, 2.8421, 2.6006, 3.4686, 3.1767], device='cuda:5'), covar=tensor([0.0054, 0.0236, 0.0290, 0.0238, 0.0161, 0.0241, 0.0127, 0.0138], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0184, 0.0181, 0.0181, 0.0183, 0.0187, 0.0186, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:52:58,784 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.718e+02 3.255e+02 3.787e+02 1.366e+03, threshold=6.510e+02, percent-clipped=4.0 2023-04-28 19:53:00,042 INFO [train.py:904] (5/8) Epoch 8, batch 2050, loss[loss=0.203, simple_loss=0.273, pruned_loss=0.06649, over 16271.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2775, pruned_loss=0.05852, over 3310468.60 frames. ], batch size: 165, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:53:22,934 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5661, 4.3156, 4.3106, 4.8046, 4.9181, 4.4950, 4.8713, 4.8530], device='cuda:5'), covar=tensor([0.1197, 0.1100, 0.2018, 0.0778, 0.0705, 0.0917, 0.0841, 0.0699], device='cuda:5'), in_proj_covar=tensor([0.0492, 0.0614, 0.0766, 0.0621, 0.0467, 0.0470, 0.0480, 0.0525], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:53:30,533 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5639, 5.9686, 5.7121, 5.8335, 5.2317, 5.1785, 5.4558, 6.0893], device='cuda:5'), covar=tensor([0.1024, 0.0784, 0.0978, 0.0569, 0.0799, 0.0631, 0.0728, 0.0816], device='cuda:5'), in_proj_covar=tensor([0.0482, 0.0615, 0.0510, 0.0408, 0.0385, 0.0403, 0.0510, 0.0454], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 19:54:09,340 INFO [train.py:904] (5/8) Epoch 8, batch 2100, loss[loss=0.2098, simple_loss=0.2798, pruned_loss=0.0699, over 16788.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2786, pruned_loss=0.05947, over 3319752.97 frames. ], batch size: 102, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:54:56,514 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 19:55:14,762 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.680e+02 3.272e+02 3.798e+02 7.634e+02, threshold=6.545e+02, percent-clipped=2.0 2023-04-28 19:55:16,490 INFO [train.py:904] (5/8) Epoch 8, batch 2150, loss[loss=0.2237, simple_loss=0.2902, pruned_loss=0.07861, over 16877.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2792, pruned_loss=0.06002, over 3315235.24 frames. ], batch size: 116, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:56:21,743 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:56:23,762 INFO [train.py:904] (5/8) Epoch 8, batch 2200, loss[loss=0.1991, simple_loss=0.283, pruned_loss=0.05766, over 16708.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.28, pruned_loss=0.06072, over 3316034.48 frames. ], batch size: 62, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:57:04,280 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5922, 3.8012, 1.9717, 3.9697, 2.6041, 3.8902, 2.0908, 2.8728], device='cuda:5'), covar=tensor([0.0175, 0.0237, 0.1331, 0.0128, 0.0741, 0.0434, 0.1227, 0.0590], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0164, 0.0182, 0.0108, 0.0166, 0.0202, 0.0192, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 19:57:20,291 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:57:29,785 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.707e+02 2.689e+02 3.130e+02 3.780e+02 7.244e+02, threshold=6.259e+02, percent-clipped=1.0 2023-04-28 19:57:31,977 INFO [train.py:904] (5/8) Epoch 8, batch 2250, loss[loss=0.1664, simple_loss=0.246, pruned_loss=0.04339, over 16972.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2806, pruned_loss=0.06088, over 3313959.25 frames. ], batch size: 41, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:57:56,963 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 19:58:40,767 INFO [train.py:904] (5/8) Epoch 8, batch 2300, loss[loss=0.189, simple_loss=0.2776, pruned_loss=0.05018, over 17281.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2809, pruned_loss=0.06045, over 3321631.33 frames. ], batch size: 52, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:58:41,152 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5728, 4.5259, 4.4696, 4.2541, 4.1409, 4.5074, 4.3783, 4.2121], device='cuda:5'), covar=tensor([0.0526, 0.0451, 0.0233, 0.0222, 0.0795, 0.0397, 0.0366, 0.0599], device='cuda:5'), in_proj_covar=tensor([0.0236, 0.0278, 0.0277, 0.0250, 0.0310, 0.0279, 0.0189, 0.0319], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 19:58:44,969 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:59:43,194 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 19:59:48,864 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.910e+02 3.420e+02 4.127e+02 1.105e+03, threshold=6.841e+02, percent-clipped=4.0 2023-04-28 19:59:49,986 INFO [train.py:904] (5/8) Epoch 8, batch 2350, loss[loss=0.219, simple_loss=0.2829, pruned_loss=0.07754, over 16850.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2802, pruned_loss=0.06011, over 3330202.46 frames. ], batch size: 116, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:59:54,639 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-04-28 20:00:21,232 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 20:00:35,285 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-28 20:00:58,128 INFO [train.py:904] (5/8) Epoch 8, batch 2400, loss[loss=0.2142, simple_loss=0.309, pruned_loss=0.05969, over 17128.00 frames. ], tot_loss[loss=0.201, simple_loss=0.281, pruned_loss=0.06049, over 3328368.83 frames. ], batch size: 48, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:01:43,800 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2921, 3.8547, 3.7813, 1.9009, 3.9961, 3.9439, 3.0630, 2.7403], device='cuda:5'), covar=tensor([0.0967, 0.0103, 0.0148, 0.1188, 0.0070, 0.0128, 0.0379, 0.0549], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0095, 0.0084, 0.0138, 0.0069, 0.0093, 0.0120, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 20:02:03,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3556, 1.4544, 1.9488, 2.1894, 2.3320, 2.2635, 1.6259, 2.3199], device='cuda:5'), covar=tensor([0.0117, 0.0270, 0.0173, 0.0178, 0.0148, 0.0134, 0.0270, 0.0078], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0159, 0.0145, 0.0148, 0.0153, 0.0107, 0.0159, 0.0099], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 20:02:04,601 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.637e+02 3.145e+02 3.743e+02 9.455e+02, threshold=6.290e+02, percent-clipped=3.0 2023-04-28 20:02:05,792 INFO [train.py:904] (5/8) Epoch 8, batch 2450, loss[loss=0.192, simple_loss=0.2793, pruned_loss=0.05234, over 17122.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2811, pruned_loss=0.06009, over 3327893.31 frames. ], batch size: 48, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:02:34,301 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-28 20:03:12,569 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:03:14,452 INFO [train.py:904] (5/8) Epoch 8, batch 2500, loss[loss=0.1876, simple_loss=0.277, pruned_loss=0.04912, over 17130.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2808, pruned_loss=0.05974, over 3331753.22 frames. ], batch size: 47, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:04:07,541 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:18,856 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:22,388 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.774e+02 3.514e+02 4.177e+02 6.912e+02, threshold=7.028e+02, percent-clipped=4.0 2023-04-28 20:04:23,444 INFO [train.py:904] (5/8) Epoch 8, batch 2550, loss[loss=0.2318, simple_loss=0.3144, pruned_loss=0.07454, over 15645.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2809, pruned_loss=0.05956, over 3327416.57 frames. ], batch size: 191, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:04:49,485 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 20:05:29,134 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,312 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,987 INFO [train.py:904] (5/8) Epoch 8, batch 2600, loss[loss=0.16, simple_loss=0.2518, pruned_loss=0.03411, over 17174.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2802, pruned_loss=0.05843, over 3326530.45 frames. ], batch size: 46, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:05:55,320 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:06:22,606 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0714, 4.1244, 4.3819, 1.8886, 4.7316, 4.7143, 3.3974, 3.3904], device='cuda:5'), covar=tensor([0.0575, 0.0115, 0.0145, 0.1023, 0.0038, 0.0071, 0.0309, 0.0379], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0095, 0.0084, 0.0138, 0.0070, 0.0093, 0.0120, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 20:06:43,000 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.639e+02 3.290e+02 4.275e+02 7.096e+02, threshold=6.579e+02, percent-clipped=1.0 2023-04-28 20:06:43,015 INFO [train.py:904] (5/8) Epoch 8, batch 2650, loss[loss=0.2208, simple_loss=0.3011, pruned_loss=0.07025, over 16824.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2812, pruned_loss=0.05868, over 3315350.72 frames. ], batch size: 102, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:07:23,290 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:07:50,963 INFO [train.py:904] (5/8) Epoch 8, batch 2700, loss[loss=0.184, simple_loss=0.2734, pruned_loss=0.04728, over 17222.00 frames. ], tot_loss[loss=0.199, simple_loss=0.281, pruned_loss=0.05853, over 3312401.23 frames. ], batch size: 45, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:08:19,261 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0827, 5.5037, 5.7087, 5.4614, 5.4619, 6.0112, 5.5922, 5.2641], device='cuda:5'), covar=tensor([0.0763, 0.1513, 0.1519, 0.1601, 0.2509, 0.0868, 0.1153, 0.2091], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0465, 0.0482, 0.0408, 0.0539, 0.0515, 0.0388, 0.0547], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 20:08:46,260 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:08:57,113 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.547e+02 3.040e+02 3.898e+02 5.528e+02, threshold=6.080e+02, percent-clipped=0.0 2023-04-28 20:08:57,129 INFO [train.py:904] (5/8) Epoch 8, batch 2750, loss[loss=0.2164, simple_loss=0.2966, pruned_loss=0.06806, over 16500.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2813, pruned_loss=0.05808, over 3319068.05 frames. ], batch size: 75, lr: 8.69e-03, grad_scale: 4.0 2023-04-28 20:09:22,652 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7843, 3.6860, 3.8484, 4.0062, 4.0517, 3.6111, 3.8879, 4.0354], device='cuda:5'), covar=tensor([0.1124, 0.0812, 0.1041, 0.0466, 0.0523, 0.1853, 0.1187, 0.0550], device='cuda:5'), in_proj_covar=tensor([0.0507, 0.0626, 0.0776, 0.0625, 0.0475, 0.0481, 0.0492, 0.0541], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:09:23,088 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-28 20:10:05,137 INFO [train.py:904] (5/8) Epoch 8, batch 2800, loss[loss=0.211, simple_loss=0.2835, pruned_loss=0.06927, over 16871.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2813, pruned_loss=0.0581, over 3324772.48 frames. ], batch size: 90, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:10:14,563 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2364, 1.6321, 2.4962, 2.9895, 2.7477, 3.3941, 1.8512, 3.4127], device='cuda:5'), covar=tensor([0.0101, 0.0337, 0.0184, 0.0160, 0.0168, 0.0088, 0.0344, 0.0068], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0162, 0.0147, 0.0151, 0.0158, 0.0111, 0.0161, 0.0100], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 20:11:14,611 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.765e+02 3.378e+02 4.166e+02 1.030e+03, threshold=6.755e+02, percent-clipped=2.0 2023-04-28 20:11:14,632 INFO [train.py:904] (5/8) Epoch 8, batch 2850, loss[loss=0.1966, simple_loss=0.2653, pruned_loss=0.06397, over 16801.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2801, pruned_loss=0.05783, over 3321593.67 frames. ], batch size: 83, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:11:35,172 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3568, 3.6932, 3.3772, 2.1904, 2.8670, 2.4952, 3.7611, 3.6513], device='cuda:5'), covar=tensor([0.0180, 0.0525, 0.0537, 0.1373, 0.0709, 0.0839, 0.0477, 0.0778], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0141, 0.0156, 0.0140, 0.0136, 0.0124, 0.0137, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 20:11:36,092 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5569, 4.3490, 4.5384, 4.7255, 4.8606, 4.3473, 4.6841, 4.8060], device='cuda:5'), covar=tensor([0.1068, 0.0857, 0.1294, 0.0612, 0.0499, 0.0970, 0.1100, 0.0507], device='cuda:5'), in_proj_covar=tensor([0.0504, 0.0623, 0.0772, 0.0624, 0.0473, 0.0480, 0.0490, 0.0541], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:12:16,879 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:20,774 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:23,990 INFO [train.py:904] (5/8) Epoch 8, batch 2900, loss[loss=0.212, simple_loss=0.2774, pruned_loss=0.07328, over 16214.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.28, pruned_loss=0.05891, over 3313318.89 frames. ], batch size: 165, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:12:25,145 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6759, 2.4597, 2.1599, 2.3338, 2.9809, 2.7895, 3.5518, 3.2771], device='cuda:5'), covar=tensor([0.0052, 0.0251, 0.0290, 0.0268, 0.0149, 0.0224, 0.0125, 0.0126], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0185, 0.0181, 0.0182, 0.0182, 0.0184, 0.0187, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:13:28,886 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:13:38,232 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.787e+02 3.423e+02 3.955e+02 9.086e+02, threshold=6.847e+02, percent-clipped=3.0 2023-04-28 20:13:38,254 INFO [train.py:904] (5/8) Epoch 8, batch 2950, loss[loss=0.2083, simple_loss=0.2779, pruned_loss=0.06931, over 16783.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2793, pruned_loss=0.05935, over 3312868.32 frames. ], batch size: 124, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:13:43,598 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 20:13:49,490 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0903, 4.0349, 4.4685, 4.4247, 4.5120, 4.1819, 4.1983, 4.0667], device='cuda:5'), covar=tensor([0.0346, 0.0467, 0.0365, 0.0460, 0.0397, 0.0353, 0.0737, 0.0519], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0308, 0.0310, 0.0294, 0.0347, 0.0324, 0.0432, 0.0264], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 20:13:54,255 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 20:14:35,338 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4796, 2.0539, 2.2675, 4.1993, 2.0907, 2.6992, 2.1644, 2.3475], device='cuda:5'), covar=tensor([0.0762, 0.2851, 0.1643, 0.0322, 0.2995, 0.1698, 0.2654, 0.2394], device='cuda:5'), in_proj_covar=tensor([0.0352, 0.0367, 0.0308, 0.0327, 0.0395, 0.0411, 0.0329, 0.0436], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:14:46,261 INFO [train.py:904] (5/8) Epoch 8, batch 3000, loss[loss=0.2307, simple_loss=0.3181, pruned_loss=0.07167, over 16794.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2795, pruned_loss=0.05989, over 3310780.96 frames. ], batch size: 57, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:14:46,261 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 20:14:55,854 INFO [train.py:938] (5/8) Epoch 8, validation: loss=0.1462, simple_loss=0.2525, pruned_loss=0.01995, over 944034.00 frames. 2023-04-28 20:14:55,855 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 20:15:13,498 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:15:17,024 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6072, 4.5836, 4.5166, 3.7990, 4.5392, 1.7470, 4.3009, 4.3239], device='cuda:5'), covar=tensor([0.0102, 0.0076, 0.0141, 0.0414, 0.0086, 0.2230, 0.0144, 0.0188], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0106, 0.0158, 0.0153, 0.0123, 0.0167, 0.0144, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:15:45,918 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:16:06,767 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.715e+02 3.274e+02 3.889e+02 8.736e+02, threshold=6.548e+02, percent-clipped=1.0 2023-04-28 20:16:06,789 INFO [train.py:904] (5/8) Epoch 8, batch 3050, loss[loss=0.2274, simple_loss=0.2992, pruned_loss=0.07781, over 16908.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2786, pruned_loss=0.05944, over 3315856.18 frames. ], batch size: 109, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:16:38,625 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:17:13,374 INFO [train.py:904] (5/8) Epoch 8, batch 3100, loss[loss=0.2012, simple_loss=0.2674, pruned_loss=0.06745, over 16692.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2787, pruned_loss=0.05974, over 3317422.36 frames. ], batch size: 134, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:18:16,359 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5185, 4.5514, 5.0434, 5.0013, 5.0175, 4.6584, 4.6248, 4.4504], device='cuda:5'), covar=tensor([0.0277, 0.0398, 0.0289, 0.0390, 0.0420, 0.0312, 0.0768, 0.0431], device='cuda:5'), in_proj_covar=tensor([0.0306, 0.0310, 0.0311, 0.0299, 0.0350, 0.0328, 0.0438, 0.0267], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 20:18:21,292 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 2.546e+02 2.917e+02 3.424e+02 8.256e+02, threshold=5.834e+02, percent-clipped=1.0 2023-04-28 20:18:21,308 INFO [train.py:904] (5/8) Epoch 8, batch 3150, loss[loss=0.208, simple_loss=0.2905, pruned_loss=0.06278, over 17082.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2773, pruned_loss=0.05908, over 3317732.30 frames. ], batch size: 53, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:18:45,882 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-04-28 20:19:15,039 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:23,360 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:28,120 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-28 20:19:32,380 INFO [train.py:904] (5/8) Epoch 8, batch 3200, loss[loss=0.1835, simple_loss=0.2798, pruned_loss=0.04365, over 17244.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2767, pruned_loss=0.0588, over 3313838.53 frames. ], batch size: 52, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:20:16,531 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0562, 3.5212, 3.0784, 1.9481, 2.6544, 2.4393, 3.5334, 3.4776], device='cuda:5'), covar=tensor([0.0250, 0.0645, 0.0705, 0.1635, 0.0859, 0.0896, 0.0542, 0.0735], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0143, 0.0156, 0.0140, 0.0136, 0.0124, 0.0138, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 20:20:29,797 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3352, 3.9624, 3.9989, 2.1023, 3.0762, 2.6933, 3.8015, 3.8056], device='cuda:5'), covar=tensor([0.0253, 0.0559, 0.0409, 0.1506, 0.0718, 0.0764, 0.0532, 0.0878], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0142, 0.0154, 0.0139, 0.0135, 0.0123, 0.0137, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 20:20:30,748 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:40,613 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:41,303 INFO [train.py:904] (5/8) Epoch 8, batch 3250, loss[loss=0.2011, simple_loss=0.277, pruned_loss=0.06262, over 16819.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2764, pruned_loss=0.05803, over 3318823.82 frames. ], batch size: 102, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:20:42,359 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.591e+02 3.147e+02 3.802e+02 9.622e+02, threshold=6.293e+02, percent-clipped=5.0 2023-04-28 20:21:14,083 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3683, 2.2250, 1.6438, 1.9831, 2.6437, 2.4109, 2.7137, 2.7538], device='cuda:5'), covar=tensor([0.0103, 0.0210, 0.0337, 0.0278, 0.0107, 0.0200, 0.0129, 0.0122], device='cuda:5'), in_proj_covar=tensor([0.0121, 0.0183, 0.0180, 0.0181, 0.0181, 0.0185, 0.0188, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:21:30,192 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.15 vs. limit=5.0 2023-04-28 20:21:36,427 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6235, 2.3381, 1.7082, 2.0669, 2.8021, 2.6053, 2.8979, 2.8946], device='cuda:5'), covar=tensor([0.0103, 0.0216, 0.0330, 0.0280, 0.0109, 0.0178, 0.0137, 0.0122], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0185, 0.0181, 0.0183, 0.0183, 0.0187, 0.0189, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:21:41,155 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1290, 5.0714, 4.8356, 4.3042, 4.9410, 1.8353, 4.6419, 4.8358], device='cuda:5'), covar=tensor([0.0057, 0.0049, 0.0117, 0.0298, 0.0068, 0.1935, 0.0097, 0.0128], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0108, 0.0160, 0.0154, 0.0124, 0.0169, 0.0145, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:21:52,440 INFO [train.py:904] (5/8) Epoch 8, batch 3300, loss[loss=0.1648, simple_loss=0.2453, pruned_loss=0.04214, over 16814.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2776, pruned_loss=0.05821, over 3320427.76 frames. ], batch size: 39, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:22:24,906 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-28 20:22:35,055 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2304, 5.1679, 4.9937, 4.3384, 5.0387, 1.7651, 4.7098, 5.0091], device='cuda:5'), covar=tensor([0.0068, 0.0058, 0.0125, 0.0368, 0.0075, 0.2077, 0.0111, 0.0135], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0108, 0.0160, 0.0155, 0.0125, 0.0169, 0.0145, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:22:41,315 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:23:01,508 INFO [train.py:904] (5/8) Epoch 8, batch 3350, loss[loss=0.224, simple_loss=0.2897, pruned_loss=0.07915, over 16863.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2779, pruned_loss=0.05797, over 3321101.87 frames. ], batch size: 116, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:23:02,761 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.621e+02 3.247e+02 4.157e+02 8.305e+02, threshold=6.494e+02, percent-clipped=1.0 2023-04-28 20:23:27,946 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:23:49,804 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:24:11,181 INFO [train.py:904] (5/8) Epoch 8, batch 3400, loss[loss=0.2365, simple_loss=0.3121, pruned_loss=0.08045, over 16793.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2776, pruned_loss=0.05746, over 3325584.84 frames. ], batch size: 102, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:24:30,010 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5451, 5.9403, 5.6222, 5.7093, 5.2814, 5.0386, 5.3447, 5.9947], device='cuda:5'), covar=tensor([0.0982, 0.0719, 0.1038, 0.0536, 0.0693, 0.0628, 0.0788, 0.0770], device='cuda:5'), in_proj_covar=tensor([0.0489, 0.0623, 0.0514, 0.0421, 0.0391, 0.0407, 0.0517, 0.0462], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:25:21,725 INFO [train.py:904] (5/8) Epoch 8, batch 3450, loss[loss=0.2465, simple_loss=0.3122, pruned_loss=0.09036, over 11702.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2758, pruned_loss=0.05719, over 3317662.93 frames. ], batch size: 246, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:25:22,842 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.611e+02 3.016e+02 3.656e+02 6.729e+02, threshold=6.032e+02, percent-clipped=3.0 2023-04-28 20:26:30,767 INFO [train.py:904] (5/8) Epoch 8, batch 3500, loss[loss=0.1875, simple_loss=0.2745, pruned_loss=0.05024, over 16601.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2745, pruned_loss=0.05631, over 3318139.77 frames. ], batch size: 68, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:26:53,227 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6908, 6.0930, 5.7277, 5.8926, 5.3085, 5.1518, 5.5385, 6.1147], device='cuda:5'), covar=tensor([0.0940, 0.0762, 0.1048, 0.0564, 0.0780, 0.0648, 0.0734, 0.0837], device='cuda:5'), in_proj_covar=tensor([0.0495, 0.0630, 0.0523, 0.0426, 0.0396, 0.0412, 0.0522, 0.0466], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:27:32,546 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:34,459 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:35,784 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0141, 4.7232, 4.9371, 5.1947, 5.3911, 4.6120, 5.3042, 5.3100], device='cuda:5'), covar=tensor([0.1299, 0.0983, 0.1641, 0.0559, 0.0429, 0.0835, 0.0425, 0.0451], device='cuda:5'), in_proj_covar=tensor([0.0508, 0.0629, 0.0786, 0.0639, 0.0482, 0.0485, 0.0496, 0.0550], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:27:42,875 INFO [train.py:904] (5/8) Epoch 8, batch 3550, loss[loss=0.2016, simple_loss=0.2836, pruned_loss=0.05983, over 16778.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2735, pruned_loss=0.05613, over 3318631.69 frames. ], batch size: 57, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:27:43,961 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.810e+02 2.461e+02 3.024e+02 3.861e+02 7.667e+02, threshold=6.049e+02, percent-clipped=4.0 2023-04-28 20:27:48,983 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0626, 5.4071, 5.5478, 5.4055, 5.3524, 5.9793, 5.6286, 5.2889], device='cuda:5'), covar=tensor([0.0829, 0.1616, 0.1711, 0.1927, 0.2964, 0.0969, 0.1148, 0.2483], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0466, 0.0480, 0.0408, 0.0540, 0.0511, 0.0387, 0.0545], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 20:28:10,883 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9983, 4.3027, 4.5187, 2.1339, 4.7906, 4.7085, 3.2938, 3.7668], device='cuda:5'), covar=tensor([0.0628, 0.0102, 0.0148, 0.0982, 0.0037, 0.0089, 0.0314, 0.0305], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0096, 0.0086, 0.0137, 0.0070, 0.0094, 0.0120, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 20:28:51,894 INFO [train.py:904] (5/8) Epoch 8, batch 3600, loss[loss=0.1815, simple_loss=0.274, pruned_loss=0.04454, over 17057.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2725, pruned_loss=0.05589, over 3327201.70 frames. ], batch size: 50, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:28:54,630 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8707, 4.8035, 4.7307, 4.5563, 4.3467, 4.8111, 4.5936, 4.5254], device='cuda:5'), covar=tensor([0.0501, 0.0457, 0.0235, 0.0243, 0.0829, 0.0343, 0.0453, 0.0528], device='cuda:5'), in_proj_covar=tensor([0.0240, 0.0284, 0.0281, 0.0252, 0.0314, 0.0287, 0.0193, 0.0321], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 20:28:57,027 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74656.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:29:42,551 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8595, 1.7479, 2.3427, 2.6902, 2.8122, 2.5618, 1.7815, 2.7158], device='cuda:5'), covar=tensor([0.0093, 0.0263, 0.0204, 0.0178, 0.0128, 0.0158, 0.0292, 0.0103], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0161, 0.0147, 0.0150, 0.0156, 0.0112, 0.0160, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 20:30:00,904 INFO [train.py:904] (5/8) Epoch 8, batch 3650, loss[loss=0.1889, simple_loss=0.2552, pruned_loss=0.06129, over 16489.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2721, pruned_loss=0.05671, over 3323464.95 frames. ], batch size: 146, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:30:02,112 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.393e+02 2.913e+02 3.939e+02 7.282e+02, threshold=5.826e+02, percent-clipped=2.0 2023-04-28 20:30:14,323 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-28 20:30:27,919 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:30:40,529 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1090, 1.9551, 1.4721, 1.7057, 2.2233, 2.0034, 2.2974, 2.3580], device='cuda:5'), covar=tensor([0.0100, 0.0197, 0.0288, 0.0287, 0.0124, 0.0220, 0.0122, 0.0147], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0185, 0.0180, 0.0182, 0.0182, 0.0186, 0.0189, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:31:13,894 INFO [train.py:904] (5/8) Epoch 8, batch 3700, loss[loss=0.1952, simple_loss=0.2654, pruned_loss=0.06251, over 16723.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.271, pruned_loss=0.05838, over 3308192.05 frames. ], batch size: 134, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:31:18,188 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1825, 1.3543, 1.9035, 1.9642, 2.1839, 2.1325, 1.5329, 2.1048], device='cuda:5'), covar=tensor([0.0117, 0.0266, 0.0150, 0.0159, 0.0142, 0.0134, 0.0246, 0.0067], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0160, 0.0145, 0.0149, 0.0155, 0.0111, 0.0159, 0.0101], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 20:31:38,770 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:31:58,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0939, 1.9724, 2.1771, 3.5899, 2.0162, 2.3527, 2.1230, 2.1289], device='cuda:5'), covar=tensor([0.0871, 0.2820, 0.1692, 0.0493, 0.3138, 0.1968, 0.2650, 0.2663], device='cuda:5'), in_proj_covar=tensor([0.0352, 0.0367, 0.0308, 0.0328, 0.0396, 0.0414, 0.0331, 0.0437], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:32:07,220 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6656, 1.5659, 2.1151, 2.4520, 2.5695, 2.3518, 1.6622, 2.5789], device='cuda:5'), covar=tensor([0.0085, 0.0263, 0.0181, 0.0148, 0.0133, 0.0146, 0.0269, 0.0078], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0160, 0.0145, 0.0149, 0.0155, 0.0111, 0.0159, 0.0101], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 20:32:29,736 INFO [train.py:904] (5/8) Epoch 8, batch 3750, loss[loss=0.2197, simple_loss=0.2759, pruned_loss=0.0818, over 16920.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2719, pruned_loss=0.06039, over 3284293.32 frames. ], batch size: 116, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:32:30,692 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.639e+02 3.130e+02 3.703e+02 1.006e+03, threshold=6.260e+02, percent-clipped=3.0 2023-04-28 20:33:04,103 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5842, 5.5098, 5.4357, 5.1919, 4.9458, 5.4446, 5.3434, 5.2068], device='cuda:5'), covar=tensor([0.0448, 0.0198, 0.0189, 0.0176, 0.0928, 0.0241, 0.0201, 0.0462], device='cuda:5'), in_proj_covar=tensor([0.0236, 0.0277, 0.0274, 0.0244, 0.0306, 0.0278, 0.0187, 0.0312], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 20:33:10,693 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 20:33:41,198 INFO [train.py:904] (5/8) Epoch 8, batch 3800, loss[loss=0.2353, simple_loss=0.3035, pruned_loss=0.08358, over 16608.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2738, pruned_loss=0.06195, over 3285632.77 frames. ], batch size: 146, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:34:45,342 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:34:51,949 INFO [train.py:904] (5/8) Epoch 8, batch 3850, loss[loss=0.1935, simple_loss=0.263, pruned_loss=0.062, over 16775.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2737, pruned_loss=0.06243, over 3283350.80 frames. ], batch size: 102, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:34:53,140 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.503e+02 3.025e+02 3.649e+02 5.657e+02, threshold=6.049e+02, percent-clipped=0.0 2023-04-28 20:35:03,479 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:35:47,568 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 20:35:52,793 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:01,464 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:02,290 INFO [train.py:904] (5/8) Epoch 8, batch 3900, loss[loss=0.2179, simple_loss=0.2856, pruned_loss=0.07507, over 16914.00 frames. ], tot_loss[loss=0.199, simple_loss=0.273, pruned_loss=0.06247, over 3282920.09 frames. ], batch size: 116, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:36:20,395 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:28,100 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:37:12,362 INFO [train.py:904] (5/8) Epoch 8, batch 3950, loss[loss=0.2255, simple_loss=0.291, pruned_loss=0.08, over 15579.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2731, pruned_loss=0.06331, over 3286888.42 frames. ], batch size: 191, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:37:14,094 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.648e+02 3.131e+02 3.735e+02 8.073e+02, threshold=6.262e+02, percent-clipped=3.0 2023-04-28 20:37:46,169 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75025.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:38:23,636 INFO [train.py:904] (5/8) Epoch 8, batch 4000, loss[loss=0.1915, simple_loss=0.269, pruned_loss=0.05702, over 16900.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2736, pruned_loss=0.06414, over 3289930.82 frames. ], batch size: 96, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:39:36,997 INFO [train.py:904] (5/8) Epoch 8, batch 4050, loss[loss=0.1881, simple_loss=0.2709, pruned_loss=0.05262, over 16449.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2735, pruned_loss=0.06242, over 3283964.61 frames. ], batch size: 146, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:39:38,165 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.447e+02 2.780e+02 3.424e+02 6.417e+02, threshold=5.561e+02, percent-clipped=2.0 2023-04-28 20:40:49,067 INFO [train.py:904] (5/8) Epoch 8, batch 4100, loss[loss=0.2184, simple_loss=0.2914, pruned_loss=0.07275, over 16608.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2742, pruned_loss=0.06133, over 3274094.81 frames. ], batch size: 62, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:41:57,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7413, 1.6582, 1.4009, 1.4801, 1.8302, 1.5043, 1.7208, 1.8996], device='cuda:5'), covar=tensor([0.0093, 0.0188, 0.0258, 0.0231, 0.0127, 0.0203, 0.0105, 0.0132], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0187, 0.0182, 0.0184, 0.0184, 0.0188, 0.0187, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:42:02,428 INFO [train.py:904] (5/8) Epoch 8, batch 4150, loss[loss=0.2285, simple_loss=0.312, pruned_loss=0.07247, over 16317.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2808, pruned_loss=0.06394, over 3241378.44 frames. ], batch size: 146, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:42:04,252 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 2.372e+02 3.014e+02 3.829e+02 9.608e+02, threshold=6.029e+02, percent-clipped=8.0 2023-04-28 20:42:24,672 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5372, 2.6591, 2.2246, 4.2021, 2.9553, 3.9212, 1.4522, 2.9033], device='cuda:5'), covar=tensor([0.1505, 0.0802, 0.1364, 0.0127, 0.0346, 0.0417, 0.1671, 0.0878], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0154, 0.0173, 0.0121, 0.0206, 0.0208, 0.0173, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 20:42:43,273 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:19,486 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:20,233 INFO [train.py:904] (5/8) Epoch 8, batch 4200, loss[loss=0.2262, simple_loss=0.3052, pruned_loss=0.0736, over 16530.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2879, pruned_loss=0.06559, over 3225575.19 frames. ], batch size: 62, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:43:41,465 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:48,156 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:52,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8902, 5.1382, 4.9597, 4.9162, 4.6232, 4.5241, 4.5667, 5.2523], device='cuda:5'), covar=tensor([0.0703, 0.0656, 0.0809, 0.0562, 0.0760, 0.0730, 0.0845, 0.0681], device='cuda:5'), in_proj_covar=tensor([0.0472, 0.0599, 0.0503, 0.0407, 0.0381, 0.0397, 0.0500, 0.0446], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:44:03,500 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7812, 4.6939, 4.4416, 3.8540, 4.6161, 1.7322, 4.3460, 4.4825], device='cuda:5'), covar=tensor([0.0058, 0.0047, 0.0127, 0.0309, 0.0064, 0.1964, 0.0090, 0.0129], device='cuda:5'), in_proj_covar=tensor([0.0115, 0.0103, 0.0153, 0.0148, 0.0120, 0.0164, 0.0139, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:44:16,479 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:18,340 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5205, 3.3884, 3.4086, 2.7999, 3.3808, 2.0601, 3.1969, 2.9736], device='cuda:5'), covar=tensor([0.0123, 0.0103, 0.0155, 0.0248, 0.0099, 0.1768, 0.0125, 0.0203], device='cuda:5'), in_proj_covar=tensor([0.0115, 0.0103, 0.0153, 0.0148, 0.0119, 0.0164, 0.0139, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:44:30,801 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:32,166 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2669, 2.4168, 1.9628, 2.1830, 2.8536, 2.5094, 3.2062, 3.1754], device='cuda:5'), covar=tensor([0.0055, 0.0280, 0.0333, 0.0315, 0.0139, 0.0214, 0.0096, 0.0115], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0183, 0.0180, 0.0181, 0.0182, 0.0184, 0.0181, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:44:34,945 INFO [train.py:904] (5/8) Epoch 8, batch 4250, loss[loss=0.1687, simple_loss=0.2588, pruned_loss=0.03925, over 17254.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2909, pruned_loss=0.06558, over 3206161.42 frames. ], batch size: 45, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:44:36,199 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 2.663e+02 3.246e+02 3.785e+02 9.237e+02, threshold=6.492e+02, percent-clipped=4.0 2023-04-28 20:45:01,377 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:45:03,488 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 20:45:18,389 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:45:48,480 INFO [train.py:904] (5/8) Epoch 8, batch 4300, loss[loss=0.211, simple_loss=0.3028, pruned_loss=0.05957, over 16657.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2922, pruned_loss=0.06478, over 3199293.27 frames. ], batch size: 62, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:46:10,661 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 20:47:02,668 INFO [train.py:904] (5/8) Epoch 8, batch 4350, loss[loss=0.188, simple_loss=0.2743, pruned_loss=0.05079, over 17011.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2958, pruned_loss=0.06598, over 3196489.11 frames. ], batch size: 50, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:47:03,096 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0425, 3.9840, 3.9042, 3.3678, 3.9852, 1.7648, 3.7616, 3.5911], device='cuda:5'), covar=tensor([0.0062, 0.0056, 0.0107, 0.0228, 0.0051, 0.1988, 0.0089, 0.0143], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0102, 0.0151, 0.0146, 0.0118, 0.0162, 0.0137, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:47:03,853 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.615e+02 3.065e+02 3.856e+02 8.729e+02, threshold=6.129e+02, percent-clipped=2.0 2023-04-28 20:47:16,474 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 20:47:33,552 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-28 20:48:17,344 INFO [train.py:904] (5/8) Epoch 8, batch 4400, loss[loss=0.2266, simple_loss=0.311, pruned_loss=0.07105, over 16949.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.298, pruned_loss=0.0673, over 3177634.51 frames. ], batch size: 116, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:49:26,870 INFO [train.py:904] (5/8) Epoch 8, batch 4450, loss[loss=0.2168, simple_loss=0.3, pruned_loss=0.06682, over 16601.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3006, pruned_loss=0.06758, over 3188186.00 frames. ], batch size: 62, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:49:28,914 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.422e+02 2.913e+02 3.508e+02 6.103e+02, threshold=5.826e+02, percent-clipped=0.0 2023-04-28 20:50:27,102 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8471, 2.1627, 1.7083, 2.0015, 2.6382, 2.2407, 2.9095, 2.8811], device='cuda:5'), covar=tensor([0.0065, 0.0236, 0.0337, 0.0300, 0.0131, 0.0244, 0.0122, 0.0127], device='cuda:5'), in_proj_covar=tensor([0.0114, 0.0181, 0.0179, 0.0181, 0.0179, 0.0184, 0.0180, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 20:50:38,190 INFO [train.py:904] (5/8) Epoch 8, batch 4500, loss[loss=0.2102, simple_loss=0.2951, pruned_loss=0.06271, over 17005.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3006, pruned_loss=0.06774, over 3195328.55 frames. ], batch size: 41, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:50:57,725 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:24,900 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:51,108 INFO [train.py:904] (5/8) Epoch 8, batch 4550, loss[loss=0.2352, simple_loss=0.3129, pruned_loss=0.0787, over 16483.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3014, pruned_loss=0.06811, over 3217103.05 frames. ], batch size: 68, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:51:51,642 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9228, 2.4056, 2.3476, 2.7302, 2.3758, 3.2931, 1.6928, 2.7498], device='cuda:5'), covar=tensor([0.1063, 0.0528, 0.0924, 0.0112, 0.0177, 0.0341, 0.1280, 0.0600], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0153, 0.0174, 0.0119, 0.0204, 0.0204, 0.0172, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 20:51:52,275 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.189e+02 2.672e+02 3.111e+02 5.807e+02, threshold=5.345e+02, percent-clipped=0.0 2023-04-28 20:52:06,458 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:17,144 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:25,173 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:53:02,644 INFO [train.py:904] (5/8) Epoch 8, batch 4600, loss[loss=0.1998, simple_loss=0.2859, pruned_loss=0.05686, over 16784.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3024, pruned_loss=0.06798, over 3217680.19 frames. ], batch size: 39, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:53:25,888 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:53:46,652 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5501, 4.2012, 4.3124, 2.8956, 3.7887, 4.2206, 3.9417, 2.4087], device='cuda:5'), covar=tensor([0.0338, 0.0019, 0.0012, 0.0236, 0.0045, 0.0045, 0.0031, 0.0271], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0063, 0.0062, 0.0119, 0.0068, 0.0077, 0.0069, 0.0112], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 20:54:12,075 INFO [train.py:904] (5/8) Epoch 8, batch 4650, loss[loss=0.2581, simple_loss=0.3219, pruned_loss=0.09714, over 11524.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3013, pruned_loss=0.0676, over 3220420.61 frames. ], batch size: 246, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:54:13,263 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.294e+02 2.810e+02 3.370e+02 9.861e+02, threshold=5.619e+02, percent-clipped=3.0 2023-04-28 20:54:26,143 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8097, 3.8968, 2.0334, 4.4914, 2.9157, 4.3704, 2.2968, 2.9189], device='cuda:5'), covar=tensor([0.0199, 0.0265, 0.1703, 0.0051, 0.0756, 0.0289, 0.1353, 0.0715], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0158, 0.0181, 0.0101, 0.0162, 0.0198, 0.0188, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 20:55:23,515 INFO [train.py:904] (5/8) Epoch 8, batch 4700, loss[loss=0.2031, simple_loss=0.28, pruned_loss=0.06313, over 16573.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2981, pruned_loss=0.06643, over 3217234.48 frames. ], batch size: 57, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:55:46,666 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0080, 4.2456, 4.6355, 2.0520, 5.0588, 5.0436, 3.2691, 3.7339], device='cuda:5'), covar=tensor([0.0666, 0.0123, 0.0123, 0.1076, 0.0023, 0.0029, 0.0318, 0.0313], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0096, 0.0083, 0.0138, 0.0068, 0.0090, 0.0118, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 20:55:46,862 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 20:55:53,506 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7670, 4.0318, 4.3271, 1.7414, 4.6652, 4.6572, 3.0999, 3.4001], device='cuda:5'), covar=tensor([0.0681, 0.0122, 0.0113, 0.1211, 0.0027, 0.0034, 0.0339, 0.0345], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0096, 0.0083, 0.0138, 0.0068, 0.0090, 0.0118, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 20:56:31,932 INFO [train.py:904] (5/8) Epoch 8, batch 4750, loss[loss=0.1896, simple_loss=0.2721, pruned_loss=0.05354, over 16722.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2941, pruned_loss=0.06443, over 3221260.49 frames. ], batch size: 124, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:56:33,072 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 2.114e+02 2.531e+02 3.127e+02 7.196e+02, threshold=5.061e+02, percent-clipped=1.0 2023-04-28 20:57:44,154 INFO [train.py:904] (5/8) Epoch 8, batch 4800, loss[loss=0.2131, simple_loss=0.288, pruned_loss=0.06915, over 11915.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2904, pruned_loss=0.06273, over 3205863.30 frames. ], batch size: 247, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:58:27,034 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 20:58:32,135 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:58:58,564 INFO [train.py:904] (5/8) Epoch 8, batch 4850, loss[loss=0.2124, simple_loss=0.306, pruned_loss=0.05942, over 16448.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2913, pruned_loss=0.06215, over 3192853.67 frames. ], batch size: 146, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 20:59:01,506 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.327e+02 2.698e+02 3.138e+02 6.949e+02, threshold=5.395e+02, percent-clipped=1.0 2023-04-28 20:59:36,902 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:59:46,583 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:00:17,705 INFO [train.py:904] (5/8) Epoch 8, batch 4900, loss[loss=0.2098, simple_loss=0.291, pruned_loss=0.06434, over 12426.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2913, pruned_loss=0.06145, over 3180562.85 frames. ], batch size: 246, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:00:36,508 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7678, 1.5383, 2.3052, 2.8274, 2.6537, 3.0967, 1.8637, 3.0988], device='cuda:5'), covar=tensor([0.0118, 0.0319, 0.0183, 0.0136, 0.0151, 0.0077, 0.0289, 0.0052], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0158, 0.0142, 0.0143, 0.0151, 0.0106, 0.0159, 0.0098], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 21:00:41,826 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5527, 4.5266, 4.3637, 4.1642, 3.9464, 4.3949, 4.2266, 4.1410], device='cuda:5'), covar=tensor([0.0468, 0.0377, 0.0224, 0.0203, 0.0924, 0.0369, 0.0386, 0.0502], device='cuda:5'), in_proj_covar=tensor([0.0213, 0.0249, 0.0249, 0.0222, 0.0278, 0.0250, 0.0170, 0.0281], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:00:49,946 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:01:34,541 INFO [train.py:904] (5/8) Epoch 8, batch 4950, loss[loss=0.2357, simple_loss=0.3098, pruned_loss=0.08086, over 12044.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2912, pruned_loss=0.06105, over 3181941.48 frames. ], batch size: 246, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:01:36,820 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.340e+02 2.828e+02 3.481e+02 8.052e+02, threshold=5.656e+02, percent-clipped=2.0 2023-04-28 21:02:02,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3198, 3.1139, 2.5258, 2.0874, 2.2074, 2.0732, 3.2392, 3.0948], device='cuda:5'), covar=tensor([0.2270, 0.0764, 0.1422, 0.1884, 0.1820, 0.1576, 0.0496, 0.0841], device='cuda:5'), in_proj_covar=tensor([0.0288, 0.0250, 0.0273, 0.0261, 0.0279, 0.0209, 0.0254, 0.0276], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:02:45,346 INFO [train.py:904] (5/8) Epoch 8, batch 5000, loss[loss=0.2022, simple_loss=0.2887, pruned_loss=0.05787, over 16438.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2928, pruned_loss=0.06124, over 3182382.05 frames. ], batch size: 62, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:37,967 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 21:03:55,766 INFO [train.py:904] (5/8) Epoch 8, batch 5050, loss[loss=0.21, simple_loss=0.2969, pruned_loss=0.06155, over 16894.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.293, pruned_loss=0.06084, over 3196584.21 frames. ], batch size: 109, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:57,928 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.576e+02 2.969e+02 3.606e+02 8.836e+02, threshold=5.938e+02, percent-clipped=5.0 2023-04-28 21:04:29,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6721, 2.5635, 2.2026, 3.2355, 2.5576, 3.6385, 1.4134, 2.7991], device='cuda:5'), covar=tensor([0.1333, 0.0617, 0.1165, 0.0145, 0.0174, 0.0340, 0.1509, 0.0724], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0153, 0.0176, 0.0120, 0.0204, 0.0205, 0.0174, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:05:07,315 INFO [train.py:904] (5/8) Epoch 8, batch 5100, loss[loss=0.1786, simple_loss=0.2624, pruned_loss=0.04735, over 16632.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2918, pruned_loss=0.06039, over 3191793.60 frames. ], batch size: 62, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:05:38,063 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-28 21:06:20,905 INFO [train.py:904] (5/8) Epoch 8, batch 5150, loss[loss=0.2033, simple_loss=0.3, pruned_loss=0.05332, over 16847.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2918, pruned_loss=0.05928, over 3195511.77 frames. ], batch size: 102, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:06:24,103 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.361e+02 2.714e+02 3.177e+02 5.443e+02, threshold=5.429e+02, percent-clipped=0.0 2023-04-28 21:07:33,717 INFO [train.py:904] (5/8) Epoch 8, batch 5200, loss[loss=0.1925, simple_loss=0.2683, pruned_loss=0.05832, over 16594.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2901, pruned_loss=0.05883, over 3206691.79 frames. ], batch size: 57, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:07:59,804 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:08:36,440 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:08:48,740 INFO [train.py:904] (5/8) Epoch 8, batch 5250, loss[loss=0.2392, simple_loss=0.3318, pruned_loss=0.07332, over 15490.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2879, pruned_loss=0.05837, over 3216758.48 frames. ], batch size: 191, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:08:51,148 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.354e+02 2.757e+02 3.436e+02 6.643e+02, threshold=5.515e+02, percent-clipped=1.0 2023-04-28 21:09:19,355 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:09:30,928 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:09:58,735 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2016, 3.9993, 4.2093, 4.3841, 4.5212, 4.0924, 4.4990, 4.5197], device='cuda:5'), covar=tensor([0.1126, 0.0931, 0.1268, 0.0525, 0.0394, 0.1079, 0.0456, 0.0415], device='cuda:5'), in_proj_covar=tensor([0.0476, 0.0577, 0.0724, 0.0588, 0.0445, 0.0444, 0.0455, 0.0508], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:09:58,809 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2202, 3.4144, 3.6302, 1.5814, 3.8710, 3.8722, 2.8173, 2.8346], device='cuda:5'), covar=tensor([0.0837, 0.0187, 0.0143, 0.1271, 0.0040, 0.0068, 0.0439, 0.0445], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0097, 0.0083, 0.0139, 0.0068, 0.0090, 0.0119, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 21:10:01,835 INFO [train.py:904] (5/8) Epoch 8, batch 5300, loss[loss=0.1714, simple_loss=0.2562, pruned_loss=0.04335, over 16398.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.284, pruned_loss=0.05675, over 3220493.63 frames. ], batch size: 75, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:10:06,525 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:10:32,415 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6761, 2.1320, 2.2549, 4.2535, 2.1039, 2.6953, 2.2890, 2.3550], device='cuda:5'), covar=tensor([0.0698, 0.2781, 0.1808, 0.0304, 0.3262, 0.1771, 0.2569, 0.2453], device='cuda:5'), in_proj_covar=tensor([0.0344, 0.0363, 0.0305, 0.0318, 0.0394, 0.0406, 0.0326, 0.0430], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:10:48,019 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:10:51,925 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6326, 6.0082, 5.6163, 5.8193, 5.2515, 5.1395, 5.5082, 6.0837], device='cuda:5'), covar=tensor([0.0835, 0.0651, 0.0865, 0.0522, 0.0730, 0.0531, 0.0615, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0473, 0.0596, 0.0506, 0.0409, 0.0383, 0.0391, 0.0498, 0.0439], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:10:55,127 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5629, 2.8851, 2.4846, 4.0703, 3.1213, 4.0472, 1.3667, 3.1176], device='cuda:5'), covar=tensor([0.1407, 0.0600, 0.1098, 0.0091, 0.0205, 0.0306, 0.1572, 0.0672], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0154, 0.0177, 0.0119, 0.0204, 0.0206, 0.0174, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:11:13,485 INFO [train.py:904] (5/8) Epoch 8, batch 5350, loss[loss=0.2022, simple_loss=0.2983, pruned_loss=0.05301, over 16780.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2824, pruned_loss=0.05582, over 3223917.63 frames. ], batch size: 102, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:11:15,943 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.303e+02 2.727e+02 3.252e+02 5.747e+02, threshold=5.455e+02, percent-clipped=1.0 2023-04-28 21:11:28,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6313, 3.7116, 4.0267, 4.0343, 3.9963, 3.7590, 3.7648, 3.7379], device='cuda:5'), covar=tensor([0.0305, 0.0600, 0.0387, 0.0400, 0.0453, 0.0339, 0.0784, 0.0459], device='cuda:5'), in_proj_covar=tensor([0.0293, 0.0294, 0.0294, 0.0286, 0.0337, 0.0313, 0.0418, 0.0258], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 21:12:26,520 INFO [train.py:904] (5/8) Epoch 8, batch 5400, loss[loss=0.2152, simple_loss=0.3085, pruned_loss=0.06093, over 16434.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2861, pruned_loss=0.05728, over 3220566.69 frames. ], batch size: 146, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:12:36,937 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1141, 4.1102, 4.0613, 3.4106, 4.0430, 1.6824, 3.7936, 3.7698], device='cuda:5'), covar=tensor([0.0083, 0.0066, 0.0099, 0.0299, 0.0080, 0.2013, 0.0117, 0.0146], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0098, 0.0146, 0.0144, 0.0114, 0.0160, 0.0131, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:13:01,336 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 21:13:43,644 INFO [train.py:904] (5/8) Epoch 8, batch 5450, loss[loss=0.2303, simple_loss=0.3093, pruned_loss=0.07566, over 16551.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2894, pruned_loss=0.05937, over 3211132.10 frames. ], batch size: 68, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:46,712 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.643e+02 3.277e+02 3.879e+02 8.643e+02, threshold=6.553e+02, percent-clipped=9.0 2023-04-28 21:15:01,754 INFO [train.py:904] (5/8) Epoch 8, batch 5500, loss[loss=0.2367, simple_loss=0.3198, pruned_loss=0.07675, over 17099.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2976, pruned_loss=0.06491, over 3193415.10 frames. ], batch size: 49, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:15:04,431 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7750, 1.2096, 1.6868, 1.6811, 1.7597, 1.8939, 1.4377, 1.7712], device='cuda:5'), covar=tensor([0.0129, 0.0208, 0.0110, 0.0165, 0.0145, 0.0076, 0.0216, 0.0054], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0159, 0.0144, 0.0145, 0.0153, 0.0106, 0.0161, 0.0100], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 21:16:22,246 INFO [train.py:904] (5/8) Epoch 8, batch 5550, loss[loss=0.3556, simple_loss=0.3935, pruned_loss=0.1589, over 10908.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3058, pruned_loss=0.07145, over 3155858.35 frames. ], batch size: 247, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:16:26,052 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.250e+02 3.650e+02 4.259e+02 5.295e+02 1.196e+03, threshold=8.517e+02, percent-clipped=11.0 2023-04-28 21:16:27,967 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 21:17:00,708 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 21:17:40,987 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:17:43,809 INFO [train.py:904] (5/8) Epoch 8, batch 5600, loss[loss=0.2211, simple_loss=0.307, pruned_loss=0.0676, over 16646.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3111, pruned_loss=0.07643, over 3114368.12 frames. ], batch size: 76, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:18:28,921 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:19:06,831 INFO [train.py:904] (5/8) Epoch 8, batch 5650, loss[loss=0.2347, simple_loss=0.3155, pruned_loss=0.07694, over 16840.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3164, pruned_loss=0.08137, over 3093289.32 frames. ], batch size: 39, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:19:10,210 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.391e+02 3.706e+02 4.487e+02 5.532e+02 1.181e+03, threshold=8.975e+02, percent-clipped=2.0 2023-04-28 21:20:03,344 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9162, 3.5038, 3.3324, 1.7006, 2.6440, 2.1764, 3.3357, 3.5228], device='cuda:5'), covar=tensor([0.0266, 0.0590, 0.0576, 0.1881, 0.0852, 0.0954, 0.0716, 0.0882], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0132, 0.0153, 0.0136, 0.0132, 0.0120, 0.0132, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:20:27,964 INFO [train.py:904] (5/8) Epoch 8, batch 5700, loss[loss=0.2585, simple_loss=0.3363, pruned_loss=0.09037, over 15446.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3187, pruned_loss=0.08331, over 3098202.43 frames. ], batch size: 191, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:21:49,369 INFO [train.py:904] (5/8) Epoch 8, batch 5750, loss[loss=0.2169, simple_loss=0.306, pruned_loss=0.06395, over 16540.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3214, pruned_loss=0.08468, over 3081205.53 frames. ], batch size: 75, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:21:54,076 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.644e+02 3.742e+02 4.943e+02 6.173e+02 1.099e+03, threshold=9.887e+02, percent-clipped=1.0 2023-04-28 21:22:52,058 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:23:12,549 INFO [train.py:904] (5/8) Epoch 8, batch 5800, loss[loss=0.2387, simple_loss=0.3157, pruned_loss=0.08088, over 17005.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3217, pruned_loss=0.08438, over 3046969.11 frames. ], batch size: 41, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:23:15,429 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0528, 2.8845, 2.7322, 2.0217, 2.5714, 2.0934, 2.7708, 2.9030], device='cuda:5'), covar=tensor([0.0352, 0.0571, 0.0497, 0.1541, 0.0708, 0.0963, 0.0583, 0.0663], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0132, 0.0154, 0.0138, 0.0133, 0.0122, 0.0134, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:24:30,497 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:24:33,006 INFO [train.py:904] (5/8) Epoch 8, batch 5850, loss[loss=0.2207, simple_loss=0.3119, pruned_loss=0.06478, over 16545.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3187, pruned_loss=0.08194, over 3059562.29 frames. ], batch size: 75, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:24:37,995 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 3.434e+02 4.196e+02 5.124e+02 1.086e+03, threshold=8.393e+02, percent-clipped=1.0 2023-04-28 21:25:09,481 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:25:53,471 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:25:56,470 INFO [train.py:904] (5/8) Epoch 8, batch 5900, loss[loss=0.2365, simple_loss=0.3162, pruned_loss=0.07842, over 16790.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3172, pruned_loss=0.08042, over 3098321.91 frames. ], batch size: 124, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:25:58,935 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0345, 2.6647, 2.6306, 1.8713, 2.7940, 2.7848, 2.4263, 2.3514], device='cuda:5'), covar=tensor([0.0674, 0.0174, 0.0157, 0.0882, 0.0085, 0.0156, 0.0378, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0098, 0.0084, 0.0142, 0.0071, 0.0092, 0.0120, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 21:26:33,737 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:26:42,728 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:27:11,067 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:27:17,812 INFO [train.py:904] (5/8) Epoch 8, batch 5950, loss[loss=0.2234, simple_loss=0.3121, pruned_loss=0.06738, over 16804.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3182, pruned_loss=0.07909, over 3110276.20 frames. ], batch size: 102, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:27:21,564 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.299e+02 3.417e+02 4.127e+02 4.793e+02 1.224e+03, threshold=8.253e+02, percent-clipped=3.0 2023-04-28 21:27:28,478 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6544, 2.5627, 2.1511, 3.4930, 2.6594, 3.6408, 1.4348, 2.6467], device='cuda:5'), covar=tensor([0.1286, 0.0595, 0.1174, 0.0146, 0.0217, 0.0390, 0.1506, 0.0811], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0152, 0.0174, 0.0117, 0.0202, 0.0205, 0.0173, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 21:27:42,565 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 21:27:55,399 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:27:55,605 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6308, 2.9128, 2.5406, 4.5412, 3.6747, 4.1504, 1.5561, 3.1269], device='cuda:5'), covar=tensor([0.1362, 0.0660, 0.1176, 0.0123, 0.0353, 0.0357, 0.1585, 0.0818], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0152, 0.0175, 0.0118, 0.0203, 0.0206, 0.0174, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:28:08,513 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-28 21:28:33,571 INFO [train.py:904] (5/8) Epoch 8, batch 6000, loss[loss=0.2238, simple_loss=0.3058, pruned_loss=0.07086, over 16539.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3168, pruned_loss=0.0782, over 3129714.97 frames. ], batch size: 68, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:28:33,572 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 21:28:44,115 INFO [train.py:938] (5/8) Epoch 8, validation: loss=0.1707, simple_loss=0.284, pruned_loss=0.02871, over 944034.00 frames. 2023-04-28 21:28:44,116 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 21:29:39,972 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7652, 1.3827, 1.6813, 1.7217, 1.7934, 1.8845, 1.5122, 1.7838], device='cuda:5'), covar=tensor([0.0142, 0.0184, 0.0116, 0.0153, 0.0129, 0.0070, 0.0200, 0.0056], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0158, 0.0144, 0.0142, 0.0152, 0.0107, 0.0160, 0.0099], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 21:29:48,612 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:30:00,325 INFO [train.py:904] (5/8) Epoch 8, batch 6050, loss[loss=0.236, simple_loss=0.3318, pruned_loss=0.07009, over 16900.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3149, pruned_loss=0.07713, over 3130598.94 frames. ], batch size: 96, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:30:04,228 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.264e+02 3.621e+02 4.460e+02 5.601e+02 1.783e+03, threshold=8.919e+02, percent-clipped=7.0 2023-04-28 21:31:19,254 INFO [train.py:904] (5/8) Epoch 8, batch 6100, loss[loss=0.2113, simple_loss=0.2936, pruned_loss=0.06447, over 16851.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3153, pruned_loss=0.0769, over 3130364.01 frames. ], batch size: 109, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:31:26,100 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:32:26,662 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:32:37,679 INFO [train.py:904] (5/8) Epoch 8, batch 6150, loss[loss=0.2423, simple_loss=0.3169, pruned_loss=0.08381, over 11738.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3132, pruned_loss=0.07655, over 3113672.75 frames. ], batch size: 247, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:32:42,677 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.089e+02 3.253e+02 3.802e+02 4.768e+02 1.072e+03, threshold=7.605e+02, percent-clipped=4.0 2023-04-28 21:32:52,939 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:33:24,654 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4854, 4.0854, 3.6238, 1.7296, 2.8557, 2.4888, 3.7738, 4.0791], device='cuda:5'), covar=tensor([0.0202, 0.0408, 0.0539, 0.2052, 0.0865, 0.0913, 0.0535, 0.0820], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0134, 0.0154, 0.0139, 0.0134, 0.0123, 0.0134, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:33:59,292 INFO [train.py:904] (5/8) Epoch 8, batch 6200, loss[loss=0.1998, simple_loss=0.2917, pruned_loss=0.05395, over 16803.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3109, pruned_loss=0.07548, over 3122228.64 frames. ], batch size: 83, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:34:31,416 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:33,186 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:39,217 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77277.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:40,896 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 21:34:43,875 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1047, 3.7137, 3.7310, 2.4106, 3.3930, 3.6978, 3.4712, 2.1897], device='cuda:5'), covar=tensor([0.0382, 0.0025, 0.0029, 0.0294, 0.0051, 0.0057, 0.0043, 0.0285], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0063, 0.0064, 0.0122, 0.0069, 0.0079, 0.0071, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 21:34:56,869 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6288, 2.7749, 2.4728, 4.3148, 3.1689, 4.0990, 1.4484, 3.0221], device='cuda:5'), covar=tensor([0.1434, 0.0709, 0.1176, 0.0124, 0.0301, 0.0371, 0.1690, 0.0784], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0152, 0.0175, 0.0118, 0.0204, 0.0206, 0.0174, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:35:16,107 INFO [train.py:904] (5/8) Epoch 8, batch 6250, loss[loss=0.2438, simple_loss=0.3085, pruned_loss=0.08954, over 11564.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3106, pruned_loss=0.07465, over 3143007.63 frames. ], batch size: 248, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:35:22,797 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 3.064e+02 3.768e+02 4.806e+02 9.942e+02, threshold=7.536e+02, percent-clipped=2.0 2023-04-28 21:35:51,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5278, 4.5351, 5.0041, 4.9778, 4.9612, 4.6398, 4.5985, 4.3770], device='cuda:5'), covar=tensor([0.0279, 0.0486, 0.0366, 0.0418, 0.0444, 0.0329, 0.0799, 0.0422], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0293, 0.0294, 0.0284, 0.0333, 0.0313, 0.0416, 0.0255], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 21:36:07,302 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:36:13,073 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77338.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:36:27,494 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9947, 4.9440, 4.8520, 4.5984, 4.3998, 4.9064, 4.8023, 4.5605], device='cuda:5'), covar=tensor([0.0515, 0.0293, 0.0220, 0.0206, 0.0846, 0.0330, 0.0250, 0.0533], device='cuda:5'), in_proj_covar=tensor([0.0216, 0.0255, 0.0250, 0.0220, 0.0278, 0.0257, 0.0172, 0.0290], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:36:34,002 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3110, 2.0738, 1.6984, 1.9212, 2.3299, 2.1125, 2.4435, 2.6073], device='cuda:5'), covar=tensor([0.0090, 0.0248, 0.0317, 0.0290, 0.0138, 0.0217, 0.0116, 0.0132], device='cuda:5'), in_proj_covar=tensor([0.0109, 0.0181, 0.0179, 0.0180, 0.0176, 0.0180, 0.0177, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:36:35,480 INFO [train.py:904] (5/8) Epoch 8, batch 6300, loss[loss=0.2289, simple_loss=0.3121, pruned_loss=0.07289, over 16384.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3098, pruned_loss=0.07407, over 3129301.38 frames. ], batch size: 146, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:37:21,180 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 21:37:54,081 INFO [train.py:904] (5/8) Epoch 8, batch 6350, loss[loss=0.2487, simple_loss=0.319, pruned_loss=0.08921, over 16850.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3117, pruned_loss=0.07625, over 3116165.23 frames. ], batch size: 58, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:38:00,462 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 3.385e+02 4.021e+02 5.215e+02 8.857e+02, threshold=8.042e+02, percent-clipped=2.0 2023-04-28 21:38:07,239 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7358, 3.9625, 1.9268, 4.3396, 2.7828, 4.3524, 2.2392, 2.9137], device='cuda:5'), covar=tensor([0.0182, 0.0264, 0.1676, 0.0080, 0.0727, 0.0377, 0.1455, 0.0636], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0161, 0.0184, 0.0103, 0.0166, 0.0200, 0.0194, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:38:56,037 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7147, 4.9595, 5.1299, 5.0655, 5.0246, 5.5857, 5.0604, 4.9178], device='cuda:5'), covar=tensor([0.0950, 0.1735, 0.1434, 0.1637, 0.2332, 0.0873, 0.1286, 0.2051], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0435, 0.0460, 0.0387, 0.0510, 0.0484, 0.0375, 0.0523], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 21:39:10,468 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:39:11,224 INFO [train.py:904] (5/8) Epoch 8, batch 6400, loss[loss=0.2333, simple_loss=0.3095, pruned_loss=0.07856, over 16659.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3114, pruned_loss=0.07698, over 3112334.72 frames. ], batch size: 134, lr: 8.49e-03, grad_scale: 8.0 2023-04-28 21:39:20,907 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 21:40:17,276 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:40:26,245 INFO [train.py:904] (5/8) Epoch 8, batch 6450, loss[loss=0.2019, simple_loss=0.2893, pruned_loss=0.05725, over 16779.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3112, pruned_loss=0.07619, over 3101725.41 frames. ], batch size: 83, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:40:33,084 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 3.422e+02 4.688e+02 6.041e+02 9.477e+02, threshold=9.377e+02, percent-clipped=7.0 2023-04-28 21:41:31,679 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:41:45,640 INFO [train.py:904] (5/8) Epoch 8, batch 6500, loss[loss=0.2491, simple_loss=0.3334, pruned_loss=0.08244, over 16344.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3088, pruned_loss=0.07515, over 3100835.34 frames. ], batch size: 146, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:42:09,108 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:05,192 INFO [train.py:904] (5/8) Epoch 8, batch 6550, loss[loss=0.2411, simple_loss=0.3295, pruned_loss=0.07633, over 17031.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3118, pruned_loss=0.07578, over 3110479.88 frames. ], batch size: 53, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:43:11,245 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 3.133e+02 3.759e+02 4.463e+02 1.371e+03, threshold=7.519e+02, percent-clipped=1.0 2023-04-28 21:43:28,219 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:47,932 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:54,048 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77633.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:44:22,377 INFO [train.py:904] (5/8) Epoch 8, batch 6600, loss[loss=0.2063, simple_loss=0.3046, pruned_loss=0.05403, over 16886.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3142, pruned_loss=0.07676, over 3085672.12 frames. ], batch size: 102, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:45:00,790 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:00,823 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:38,851 INFO [train.py:904] (5/8) Epoch 8, batch 6650, loss[loss=0.3436, simple_loss=0.3854, pruned_loss=0.1509, over 11406.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.316, pruned_loss=0.07883, over 3072912.90 frames. ], batch size: 248, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:45:45,539 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.195e+02 3.528e+02 4.149e+02 5.029e+02 1.289e+03, threshold=8.299e+02, percent-clipped=5.0 2023-04-28 21:45:57,648 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-28 21:46:35,367 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:46:53,704 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:46:54,555 INFO [train.py:904] (5/8) Epoch 8, batch 6700, loss[loss=0.2365, simple_loss=0.3155, pruned_loss=0.07877, over 16922.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3144, pruned_loss=0.07858, over 3082809.50 frames. ], batch size: 109, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:47:33,955 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7604, 4.2771, 4.1626, 2.0829, 3.3121, 2.9465, 4.0738, 4.3709], device='cuda:5'), covar=tensor([0.0214, 0.0452, 0.0469, 0.1656, 0.0676, 0.0742, 0.0531, 0.0622], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0132, 0.0156, 0.0139, 0.0133, 0.0123, 0.0135, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 21:47:50,741 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3640, 5.7282, 5.3811, 5.4437, 4.9933, 5.0051, 5.1948, 5.7783], device='cuda:5'), covar=tensor([0.0837, 0.0650, 0.0929, 0.0563, 0.0801, 0.0578, 0.0816, 0.0703], device='cuda:5'), in_proj_covar=tensor([0.0473, 0.0587, 0.0497, 0.0400, 0.0371, 0.0386, 0.0488, 0.0430], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:48:06,140 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:48:10,962 INFO [train.py:904] (5/8) Epoch 8, batch 6750, loss[loss=0.2101, simple_loss=0.2951, pruned_loss=0.06253, over 16854.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3126, pruned_loss=0.07791, over 3103109.70 frames. ], batch size: 102, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:48:18,411 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.344e+02 4.001e+02 5.088e+02 1.053e+03, threshold=8.003e+02, percent-clipped=2.0 2023-04-28 21:49:09,428 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-04-28 21:49:25,321 INFO [train.py:904] (5/8) Epoch 8, batch 6800, loss[loss=0.2433, simple_loss=0.3272, pruned_loss=0.07974, over 15350.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.313, pruned_loss=0.07829, over 3093873.93 frames. ], batch size: 190, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:49:49,089 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:50:44,162 INFO [train.py:904] (5/8) Epoch 8, batch 6850, loss[loss=0.269, simple_loss=0.3252, pruned_loss=0.1064, over 11437.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3137, pruned_loss=0.07818, over 3107254.39 frames. ], batch size: 248, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:50:53,214 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.176e+02 3.270e+02 3.895e+02 4.577e+02 9.421e+02, threshold=7.790e+02, percent-clipped=1.0 2023-04-28 21:51:03,730 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:09,845 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2757, 3.9664, 3.9694, 2.6367, 3.6848, 4.0520, 3.7932, 2.4190], device='cuda:5'), covar=tensor([0.0351, 0.0031, 0.0032, 0.0267, 0.0051, 0.0050, 0.0040, 0.0272], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0061, 0.0064, 0.0120, 0.0068, 0.0079, 0.0070, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 21:51:24,311 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:30,467 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:59,512 INFO [train.py:904] (5/8) Epoch 8, batch 6900, loss[loss=0.2933, simple_loss=0.3514, pruned_loss=0.1176, over 11528.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3171, pruned_loss=0.07831, over 3114126.35 frames. ], batch size: 247, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:52:31,663 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:39,050 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:45,341 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77981.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:53:20,824 INFO [train.py:904] (5/8) Epoch 8, batch 6950, loss[loss=0.3349, simple_loss=0.3755, pruned_loss=0.1472, over 11314.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3191, pruned_loss=0.08012, over 3105541.45 frames. ], batch size: 248, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:53:22,992 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1392, 5.1803, 5.6915, 5.6083, 5.6831, 5.2077, 5.1366, 4.7932], device='cuda:5'), covar=tensor([0.0249, 0.0392, 0.0268, 0.0388, 0.0371, 0.0256, 0.0902, 0.0402], device='cuda:5'), in_proj_covar=tensor([0.0292, 0.0300, 0.0300, 0.0289, 0.0339, 0.0317, 0.0420, 0.0259], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 21:53:29,771 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 3.295e+02 4.352e+02 5.795e+02 9.816e+02, threshold=8.703e+02, percent-clipped=9.0 2023-04-28 21:53:55,738 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1057, 1.8560, 2.1400, 3.4991, 1.9187, 2.3523, 2.0723, 2.0081], device='cuda:5'), covar=tensor([0.0799, 0.2951, 0.1819, 0.0427, 0.3454, 0.1895, 0.2548, 0.2858], device='cuda:5'), in_proj_covar=tensor([0.0341, 0.0364, 0.0305, 0.0318, 0.0399, 0.0401, 0.0324, 0.0427], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 21:53:59,855 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2130, 3.9077, 3.8528, 2.6708, 3.4957, 3.7928, 3.6406, 2.0295], device='cuda:5'), covar=tensor([0.0399, 0.0026, 0.0040, 0.0273, 0.0056, 0.0107, 0.0044, 0.0366], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0062, 0.0064, 0.0121, 0.0068, 0.0080, 0.0070, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 21:54:10,663 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78033.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:54:38,877 INFO [train.py:904] (5/8) Epoch 8, batch 7000, loss[loss=0.2361, simple_loss=0.326, pruned_loss=0.0731, over 16705.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3191, pruned_loss=0.08013, over 3090691.71 frames. ], batch size: 134, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:55:53,777 INFO [train.py:904] (5/8) Epoch 8, batch 7050, loss[loss=0.2304, simple_loss=0.3113, pruned_loss=0.07471, over 15312.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3185, pruned_loss=0.07872, over 3099004.67 frames. ], batch size: 190, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:56:03,901 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 3.078e+02 3.817e+02 4.566e+02 8.440e+02, threshold=7.634e+02, percent-clipped=0.0 2023-04-28 21:57:11,220 INFO [train.py:904] (5/8) Epoch 8, batch 7100, loss[loss=0.241, simple_loss=0.3235, pruned_loss=0.07925, over 15424.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3166, pruned_loss=0.07818, over 3096005.66 frames. ], batch size: 190, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:58:26,614 INFO [train.py:904] (5/8) Epoch 8, batch 7150, loss[loss=0.2246, simple_loss=0.3039, pruned_loss=0.07266, over 16413.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3151, pruned_loss=0.0782, over 3091645.28 frames. ], batch size: 68, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:58:36,173 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.673e+02 3.732e+02 4.325e+02 5.377e+02 1.184e+03, threshold=8.651e+02, percent-clipped=7.0 2023-04-28 21:58:51,625 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 21:59:09,615 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78232.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:59:39,846 INFO [train.py:904] (5/8) Epoch 8, batch 7200, loss[loss=0.1799, simple_loss=0.2786, pruned_loss=0.04059, over 16500.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3128, pruned_loss=0.07628, over 3079486.41 frames. ], batch size: 75, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:00:10,020 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:00:37,219 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9912, 3.9519, 3.9182, 3.3508, 3.9282, 1.7272, 3.7484, 3.5767], device='cuda:5'), covar=tensor([0.0078, 0.0071, 0.0114, 0.0264, 0.0069, 0.2074, 0.0098, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0107, 0.0096, 0.0142, 0.0139, 0.0113, 0.0158, 0.0127, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:00:44,925 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:00:59,557 INFO [train.py:904] (5/8) Epoch 8, batch 7250, loss[loss=0.2099, simple_loss=0.293, pruned_loss=0.06343, over 16815.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3102, pruned_loss=0.07516, over 3072845.17 frames. ], batch size: 124, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:01:10,052 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.959e+02 3.722e+02 4.557e+02 8.668e+02, threshold=7.444e+02, percent-clipped=1.0 2023-04-28 22:01:19,066 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 22:01:27,084 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:01:47,637 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:02:16,551 INFO [train.py:904] (5/8) Epoch 8, batch 7300, loss[loss=0.21, simple_loss=0.2967, pruned_loss=0.06163, over 16469.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.309, pruned_loss=0.07502, over 3082488.53 frames. ], batch size: 75, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:02:42,889 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3419, 4.5046, 4.5808, 4.6243, 4.6350, 5.1054, 4.6610, 4.4593], device='cuda:5'), covar=tensor([0.1292, 0.1553, 0.1489, 0.1550, 0.2036, 0.0857, 0.1338, 0.2514], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0429, 0.0457, 0.0386, 0.0509, 0.0484, 0.0370, 0.0521], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 22:03:02,794 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:03:34,481 INFO [train.py:904] (5/8) Epoch 8, batch 7350, loss[loss=0.2235, simple_loss=0.3035, pruned_loss=0.07175, over 16757.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3096, pruned_loss=0.07597, over 3072569.13 frames. ], batch size: 124, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:03:37,968 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9127, 2.5417, 2.5892, 1.7792, 2.7697, 2.7722, 2.3165, 2.2910], device='cuda:5'), covar=tensor([0.0790, 0.0275, 0.0238, 0.1012, 0.0082, 0.0157, 0.0470, 0.0441], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0097, 0.0082, 0.0137, 0.0066, 0.0088, 0.0118, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 22:03:45,282 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.445e+02 3.549e+02 4.373e+02 5.467e+02 2.414e+03, threshold=8.746e+02, percent-clipped=10.0 2023-04-28 22:04:55,419 INFO [train.py:904] (5/8) Epoch 8, batch 7400, loss[loss=0.2411, simple_loss=0.322, pruned_loss=0.08009, over 16713.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3108, pruned_loss=0.07636, over 3085818.23 frames. ], batch size: 89, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:05:08,456 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:06:13,356 INFO [train.py:904] (5/8) Epoch 8, batch 7450, loss[loss=0.2812, simple_loss=0.332, pruned_loss=0.1151, over 11424.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3122, pruned_loss=0.07806, over 3070767.75 frames. ], batch size: 248, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:06:26,558 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 3.338e+02 4.187e+02 5.103e+02 1.080e+03, threshold=8.375e+02, percent-clipped=2.0 2023-04-28 22:06:37,061 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1942, 3.2445, 1.6897, 3.4068, 2.3261, 3.4329, 1.8839, 2.5991], device='cuda:5'), covar=tensor([0.0209, 0.0357, 0.1547, 0.0150, 0.0848, 0.0550, 0.1467, 0.0657], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0156, 0.0179, 0.0101, 0.0163, 0.0194, 0.0188, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 22:06:47,405 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:07:34,685 INFO [train.py:904] (5/8) Epoch 8, batch 7500, loss[loss=0.2564, simple_loss=0.3163, pruned_loss=0.09819, over 11386.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.314, pruned_loss=0.07883, over 3049442.61 frames. ], batch size: 248, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:07:42,049 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:07:52,478 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7856, 5.3441, 5.4572, 5.3375, 5.4153, 5.9089, 5.4369, 5.2305], device='cuda:5'), covar=tensor([0.0847, 0.1469, 0.1596, 0.1601, 0.1819, 0.0792, 0.1082, 0.1847], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0433, 0.0465, 0.0390, 0.0514, 0.0489, 0.0377, 0.0523], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 22:08:33,362 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:08:54,412 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1542, 3.2130, 3.3532, 1.6497, 3.5014, 3.5532, 2.6895, 2.5869], device='cuda:5'), covar=tensor([0.0732, 0.0169, 0.0141, 0.1159, 0.0057, 0.0096, 0.0400, 0.0431], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0094, 0.0081, 0.0135, 0.0065, 0.0087, 0.0115, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 22:08:55,087 INFO [train.py:904] (5/8) Epoch 8, batch 7550, loss[loss=0.2257, simple_loss=0.3067, pruned_loss=0.07236, over 16198.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3134, pruned_loss=0.07915, over 3045271.58 frames. ], batch size: 165, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:09:05,666 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.406e+02 3.301e+02 4.162e+02 5.431e+02 1.258e+03, threshold=8.325e+02, percent-clipped=7.0 2023-04-28 22:09:20,065 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:09:32,582 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2173, 3.3875, 3.6388, 1.6084, 3.7761, 3.8833, 2.8341, 2.7547], device='cuda:5'), covar=tensor([0.0841, 0.0198, 0.0165, 0.1303, 0.0068, 0.0088, 0.0412, 0.0469], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0094, 0.0081, 0.0135, 0.0065, 0.0087, 0.0115, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 22:10:12,507 INFO [train.py:904] (5/8) Epoch 8, batch 7600, loss[loss=0.2475, simple_loss=0.315, pruned_loss=0.09002, over 15259.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3133, pruned_loss=0.07966, over 3036101.14 frames. ], batch size: 190, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:10:45,315 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9571, 2.7471, 2.7259, 2.0011, 2.5849, 2.1553, 2.6225, 2.8494], device='cuda:5'), covar=tensor([0.0328, 0.0650, 0.0484, 0.1556, 0.0703, 0.0842, 0.0566, 0.0735], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0134, 0.0157, 0.0142, 0.0134, 0.0124, 0.0137, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 22:11:30,207 INFO [train.py:904] (5/8) Epoch 8, batch 7650, loss[loss=0.2368, simple_loss=0.3116, pruned_loss=0.08102, over 17029.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3131, pruned_loss=0.07961, over 3041450.12 frames. ], batch size: 55, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:11:40,066 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 22:11:40,449 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 3.310e+02 4.230e+02 5.150e+02 8.626e+02, threshold=8.460e+02, percent-clipped=2.0 2023-04-28 22:12:45,891 INFO [train.py:904] (5/8) Epoch 8, batch 7700, loss[loss=0.2431, simple_loss=0.3174, pruned_loss=0.08438, over 16683.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3133, pruned_loss=0.07949, over 3056316.54 frames. ], batch size: 134, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:14:03,989 INFO [train.py:904] (5/8) Epoch 8, batch 7750, loss[loss=0.2361, simple_loss=0.3171, pruned_loss=0.07757, over 16676.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.313, pruned_loss=0.07884, over 3067932.65 frames. ], batch size: 134, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:14:17,809 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.129e+02 3.510e+02 4.111e+02 5.372e+02 9.340e+02, threshold=8.221e+02, percent-clipped=3.0 2023-04-28 22:14:26,066 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:15:06,383 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-28 22:15:19,818 INFO [train.py:904] (5/8) Epoch 8, batch 7800, loss[loss=0.2172, simple_loss=0.2981, pruned_loss=0.06819, over 16510.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3143, pruned_loss=0.07976, over 3071835.93 frames. ], batch size: 146, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:16,711 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:16:20,503 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0822, 5.0185, 4.8025, 4.2212, 4.8966, 1.8021, 4.7079, 4.7820], device='cuda:5'), covar=tensor([0.0053, 0.0051, 0.0110, 0.0280, 0.0064, 0.2013, 0.0092, 0.0124], device='cuda:5'), in_proj_covar=tensor([0.0107, 0.0094, 0.0140, 0.0137, 0.0112, 0.0158, 0.0126, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:16:38,239 INFO [train.py:904] (5/8) Epoch 8, batch 7850, loss[loss=0.2831, simple_loss=0.3441, pruned_loss=0.1111, over 11339.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3154, pruned_loss=0.07984, over 3064146.81 frames. ], batch size: 248, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:45,475 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0292, 2.5996, 2.6213, 1.8963, 2.7897, 2.7643, 2.3967, 2.3572], device='cuda:5'), covar=tensor([0.0694, 0.0187, 0.0213, 0.0908, 0.0085, 0.0185, 0.0410, 0.0422], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0096, 0.0082, 0.0137, 0.0066, 0.0089, 0.0117, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 22:16:50,963 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.229e+02 3.386e+02 3.845e+02 4.858e+02 8.286e+02, threshold=7.691e+02, percent-clipped=1.0 2023-04-28 22:16:53,961 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:17:28,017 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6188, 2.6683, 2.6538, 4.1506, 3.3731, 4.0408, 1.2248, 3.3433], device='cuda:5'), covar=tensor([0.1416, 0.0706, 0.1047, 0.0158, 0.0231, 0.0352, 0.1663, 0.0631], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0153, 0.0173, 0.0119, 0.0202, 0.0203, 0.0174, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 22:17:28,080 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3614, 2.0489, 2.1187, 4.0818, 2.0008, 2.5692, 2.1579, 2.2199], device='cuda:5'), covar=tensor([0.0890, 0.2934, 0.1932, 0.0363, 0.3541, 0.1891, 0.2557, 0.2862], device='cuda:5'), in_proj_covar=tensor([0.0340, 0.0365, 0.0303, 0.0317, 0.0400, 0.0402, 0.0323, 0.0428], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:17:29,667 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78936.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:17:34,906 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6651, 1.4295, 2.0969, 2.4615, 2.4285, 2.7235, 1.6022, 2.6160], device='cuda:5'), covar=tensor([0.0122, 0.0359, 0.0216, 0.0198, 0.0179, 0.0124, 0.0366, 0.0092], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0159, 0.0143, 0.0141, 0.0149, 0.0108, 0.0157, 0.0099], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 22:17:54,235 INFO [train.py:904] (5/8) Epoch 8, batch 7900, loss[loss=0.2315, simple_loss=0.3129, pruned_loss=0.07501, over 16712.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3144, pruned_loss=0.07924, over 3056711.92 frames. ], batch size: 134, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:18:15,505 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:18:26,137 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0491, 3.6107, 3.4470, 1.8670, 2.9966, 2.4455, 3.4351, 3.7328], device='cuda:5'), covar=tensor([0.0291, 0.0526, 0.0549, 0.1811, 0.0723, 0.0878, 0.0671, 0.0801], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0133, 0.0156, 0.0141, 0.0133, 0.0124, 0.0137, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 22:19:13,452 INFO [train.py:904] (5/8) Epoch 8, batch 7950, loss[loss=0.2358, simple_loss=0.311, pruned_loss=0.08035, over 16819.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3159, pruned_loss=0.08038, over 3048220.65 frames. ], batch size: 39, lr: 8.40e-03, grad_scale: 2.0 2023-04-28 22:19:14,082 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8197, 3.1760, 2.6681, 5.0148, 4.0236, 4.5022, 1.6662, 3.1617], device='cuda:5'), covar=tensor([0.1216, 0.0599, 0.1092, 0.0139, 0.0357, 0.0295, 0.1317, 0.0797], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0152, 0.0173, 0.0119, 0.0201, 0.0202, 0.0174, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 22:19:28,065 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.154e+02 3.667e+02 4.128e+02 4.915e+02 9.776e+02, threshold=8.255e+02, percent-clipped=2.0 2023-04-28 22:19:29,946 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-28 22:19:52,426 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:20:26,190 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7201, 2.4827, 2.1506, 3.4099, 2.4514, 3.6006, 1.3257, 2.6775], device='cuda:5'), covar=tensor([0.1329, 0.0649, 0.1224, 0.0146, 0.0190, 0.0421, 0.1629, 0.0813], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0153, 0.0173, 0.0119, 0.0201, 0.0203, 0.0174, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 22:20:32,281 INFO [train.py:904] (5/8) Epoch 8, batch 8000, loss[loss=0.2308, simple_loss=0.3135, pruned_loss=0.0741, over 16451.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3162, pruned_loss=0.08087, over 3039478.08 frames. ], batch size: 75, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:20:38,645 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3805, 3.2295, 2.5305, 2.1610, 2.2702, 2.0771, 3.2481, 3.1386], device='cuda:5'), covar=tensor([0.2418, 0.0764, 0.1470, 0.1948, 0.2040, 0.1685, 0.0498, 0.0899], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0256, 0.0278, 0.0266, 0.0281, 0.0213, 0.0262, 0.0279], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:20:46,417 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5209, 3.0976, 2.9607, 1.8895, 2.6596, 2.2537, 3.0295, 3.1765], device='cuda:5'), covar=tensor([0.0299, 0.0562, 0.0608, 0.1714, 0.0795, 0.0895, 0.0676, 0.0782], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0133, 0.0156, 0.0141, 0.0133, 0.0124, 0.0136, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 22:20:47,267 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8309, 5.3595, 5.5765, 5.4002, 5.3806, 5.9294, 5.4807, 5.1994], device='cuda:5'), covar=tensor([0.0800, 0.1329, 0.1536, 0.1587, 0.2242, 0.0824, 0.1179, 0.2276], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0433, 0.0465, 0.0392, 0.0513, 0.0485, 0.0373, 0.0527], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 22:21:48,612 INFO [train.py:904] (5/8) Epoch 8, batch 8050, loss[loss=0.2374, simple_loss=0.3177, pruned_loss=0.07848, over 16916.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3146, pruned_loss=0.07937, over 3059353.33 frames. ], batch size: 109, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:22:02,036 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.318e+02 3.198e+02 3.767e+02 4.640e+02 1.099e+03, threshold=7.534e+02, percent-clipped=1.0 2023-04-28 22:22:05,322 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 22:22:11,255 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:22:22,218 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 22:22:53,917 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8490, 3.9451, 3.1324, 2.4310, 2.8590, 2.4997, 4.2651, 3.7746], device='cuda:5'), covar=tensor([0.2253, 0.0664, 0.1318, 0.1754, 0.2040, 0.1510, 0.0332, 0.0799], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0255, 0.0278, 0.0266, 0.0282, 0.0213, 0.0261, 0.0278], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:23:05,326 INFO [train.py:904] (5/8) Epoch 8, batch 8100, loss[loss=0.2303, simple_loss=0.3074, pruned_loss=0.07658, over 16505.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3136, pruned_loss=0.0784, over 3069356.36 frames. ], batch size: 68, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:23:24,441 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:24:22,974 INFO [train.py:904] (5/8) Epoch 8, batch 8150, loss[loss=0.2054, simple_loss=0.2894, pruned_loss=0.0607, over 16869.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3113, pruned_loss=0.07761, over 3058307.49 frames. ], batch size: 116, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:24:28,433 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0845, 2.8902, 2.7624, 2.0197, 2.5149, 2.2171, 2.7754, 2.9762], device='cuda:5'), covar=tensor([0.0286, 0.0526, 0.0504, 0.1498, 0.0726, 0.0830, 0.0558, 0.0612], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0131, 0.0154, 0.0139, 0.0132, 0.0123, 0.0134, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 22:24:36,888 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 3.692e+02 4.407e+02 5.311e+02 8.589e+02, threshold=8.814e+02, percent-clipped=3.0 2023-04-28 22:24:38,596 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:25:42,525 INFO [train.py:904] (5/8) Epoch 8, batch 8200, loss[loss=0.2215, simple_loss=0.3021, pruned_loss=0.07048, over 16617.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3098, pruned_loss=0.07683, over 3084796.38 frames. ], batch size: 134, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:25:55,508 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:26:19,811 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5580, 2.6424, 2.3713, 3.6445, 2.2252, 3.8383, 1.3109, 2.6943], device='cuda:5'), covar=tensor([0.1523, 0.0653, 0.1107, 0.0161, 0.0165, 0.0403, 0.1704, 0.0821], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0152, 0.0171, 0.0118, 0.0200, 0.0199, 0.0172, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 22:26:36,858 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:27:04,314 INFO [train.py:904] (5/8) Epoch 8, batch 8250, loss[loss=0.1959, simple_loss=0.2896, pruned_loss=0.0511, over 16849.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3085, pruned_loss=0.07506, over 3049411.03 frames. ], batch size: 96, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:27:19,445 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 3.091e+02 3.915e+02 5.485e+02 9.423e+02, threshold=7.829e+02, percent-clipped=2.0 2023-04-28 22:27:21,997 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9364, 2.3986, 2.4002, 2.9893, 2.0532, 3.3743, 1.7033, 2.8418], device='cuda:5'), covar=tensor([0.1132, 0.0452, 0.0778, 0.0128, 0.0085, 0.0324, 0.1252, 0.0554], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0151, 0.0170, 0.0117, 0.0198, 0.0198, 0.0171, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 22:27:25,589 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 22:27:37,071 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:27:57,005 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:17,738 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:28:26,473 INFO [train.py:904] (5/8) Epoch 8, batch 8300, loss[loss=0.1916, simple_loss=0.2848, pruned_loss=0.04914, over 15189.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3041, pruned_loss=0.07085, over 3053163.97 frames. ], batch size: 190, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:28:46,974 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:58,683 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6414, 4.7892, 4.9154, 4.8371, 4.7702, 5.3279, 4.9367, 4.6795], device='cuda:5'), covar=tensor([0.0921, 0.1797, 0.1641, 0.1987, 0.2790, 0.1049, 0.1400, 0.2452], device='cuda:5'), in_proj_covar=tensor([0.0316, 0.0430, 0.0461, 0.0387, 0.0504, 0.0482, 0.0372, 0.0514], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 22:29:05,023 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:29:36,171 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:29:48,247 INFO [train.py:904] (5/8) Epoch 8, batch 8350, loss[loss=0.2352, simple_loss=0.3061, pruned_loss=0.0821, over 12236.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3031, pruned_loss=0.06841, over 3067810.71 frames. ], batch size: 247, lr: 8.38e-03, grad_scale: 4.0 2023-04-28 22:30:02,874 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 2.645e+02 3.311e+02 4.098e+02 6.833e+02, threshold=6.622e+02, percent-clipped=0.0 2023-04-28 22:30:26,261 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:30:43,775 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:31:09,033 INFO [train.py:904] (5/8) Epoch 8, batch 8400, loss[loss=0.1938, simple_loss=0.2708, pruned_loss=0.05834, over 11628.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.3003, pruned_loss=0.06633, over 3051102.31 frames. ], batch size: 246, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:31:11,169 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-04-28 22:32:27,073 INFO [train.py:904] (5/8) Epoch 8, batch 8450, loss[loss=0.1903, simple_loss=0.2733, pruned_loss=0.05366, over 12345.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2983, pruned_loss=0.06402, over 3064834.87 frames. ], batch size: 246, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:32:42,128 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.576e+02 3.232e+02 4.028e+02 7.324e+02, threshold=6.464e+02, percent-clipped=2.0 2023-04-28 22:32:44,043 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2259, 4.2400, 4.0688, 3.6003, 4.1283, 1.7344, 3.9271, 3.9259], device='cuda:5'), covar=tensor([0.0064, 0.0057, 0.0119, 0.0232, 0.0065, 0.2082, 0.0097, 0.0135], device='cuda:5'), in_proj_covar=tensor([0.0108, 0.0094, 0.0141, 0.0137, 0.0112, 0.0160, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:33:47,274 INFO [train.py:904] (5/8) Epoch 8, batch 8500, loss[loss=0.1895, simple_loss=0.2824, pruned_loss=0.04829, over 16416.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2932, pruned_loss=0.06072, over 3063218.06 frames. ], batch size: 75, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:34:40,484 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:35:09,524 INFO [train.py:904] (5/8) Epoch 8, batch 8550, loss[loss=0.1991, simple_loss=0.2935, pruned_loss=0.05236, over 16686.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2903, pruned_loss=0.05928, over 3044393.75 frames. ], batch size: 83, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:35:26,485 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.652e+02 3.316e+02 4.199e+02 1.038e+03, threshold=6.632e+02, percent-clipped=3.0 2023-04-28 22:35:47,551 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:36:28,522 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:36:32,414 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5557, 3.5708, 3.5682, 3.0024, 3.5125, 1.9323, 3.3394, 3.1210], device='cuda:5'), covar=tensor([0.0093, 0.0081, 0.0099, 0.0185, 0.0068, 0.1850, 0.0103, 0.0156], device='cuda:5'), in_proj_covar=tensor([0.0108, 0.0095, 0.0142, 0.0136, 0.0113, 0.0163, 0.0128, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:36:37,295 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:36:48,481 INFO [train.py:904] (5/8) Epoch 8, batch 8600, loss[loss=0.1807, simple_loss=0.2766, pruned_loss=0.04242, over 16403.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2908, pruned_loss=0.05873, over 3021325.71 frames. ], batch size: 75, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:36:59,467 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:37:24,936 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:37:50,220 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9421, 5.6513, 5.8290, 5.6735, 5.6303, 6.1334, 5.8168, 5.5684], device='cuda:5'), covar=tensor([0.0623, 0.1602, 0.1491, 0.1675, 0.2184, 0.0938, 0.1119, 0.1927], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0412, 0.0443, 0.0368, 0.0486, 0.0465, 0.0361, 0.0490], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:38:02,634 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79689.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:38:26,211 INFO [train.py:904] (5/8) Epoch 8, batch 8650, loss[loss=0.2104, simple_loss=0.2856, pruned_loss=0.0676, over 12332.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2884, pruned_loss=0.05654, over 3036031.78 frames. ], batch size: 250, lr: 8.37e-03, grad_scale: 4.0 2023-04-28 22:38:32,915 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0357, 3.8515, 4.1517, 4.2420, 4.3563, 3.8597, 4.3350, 4.3421], device='cuda:5'), covar=tensor([0.1279, 0.0958, 0.1045, 0.0573, 0.0452, 0.1205, 0.0500, 0.0491], device='cuda:5'), in_proj_covar=tensor([0.0451, 0.0551, 0.0673, 0.0564, 0.0427, 0.0421, 0.0448, 0.0492], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:38:50,268 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.014e+02 2.698e+02 3.186e+02 3.919e+02 1.176e+03, threshold=6.372e+02, percent-clipped=4.0 2023-04-28 22:39:04,737 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:09,488 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:32,938 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:40:12,074 INFO [train.py:904] (5/8) Epoch 8, batch 8700, loss[loss=0.1794, simple_loss=0.2758, pruned_loss=0.04149, over 16339.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2858, pruned_loss=0.0548, over 3051596.94 frames. ], batch size: 146, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:40:33,026 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:41:50,481 INFO [train.py:904] (5/8) Epoch 8, batch 8750, loss[loss=0.1886, simple_loss=0.2896, pruned_loss=0.04382, over 16368.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.285, pruned_loss=0.05414, over 3052001.90 frames. ], batch size: 146, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:42:06,415 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8135, 3.6724, 3.8705, 3.9570, 4.0548, 3.6344, 4.0000, 4.0773], device='cuda:5'), covar=tensor([0.1171, 0.0853, 0.1070, 0.0575, 0.0491, 0.1448, 0.0590, 0.0491], device='cuda:5'), in_proj_covar=tensor([0.0438, 0.0543, 0.0661, 0.0554, 0.0422, 0.0415, 0.0439, 0.0486], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:42:11,566 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-04-28 22:42:15,365 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.669e+02 3.166e+02 4.113e+02 7.290e+02, threshold=6.332e+02, percent-clipped=2.0 2023-04-28 22:42:44,146 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:43:01,647 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 22:43:44,540 INFO [train.py:904] (5/8) Epoch 8, batch 8800, loss[loss=0.1774, simple_loss=0.2699, pruned_loss=0.04244, over 16637.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2827, pruned_loss=0.05244, over 3065559.80 frames. ], batch size: 62, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:44:22,810 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 22:45:22,659 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7092, 1.5820, 1.9653, 2.6338, 2.5078, 2.6909, 1.9252, 2.8220], device='cuda:5'), covar=tensor([0.0128, 0.0313, 0.0244, 0.0166, 0.0171, 0.0140, 0.0305, 0.0075], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0157, 0.0142, 0.0138, 0.0145, 0.0103, 0.0155, 0.0094], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 22:45:31,367 INFO [train.py:904] (5/8) Epoch 8, batch 8850, loss[loss=0.1911, simple_loss=0.2922, pruned_loss=0.04501, over 16636.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2845, pruned_loss=0.05148, over 3052431.33 frames. ], batch size: 62, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:45:51,488 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.661e+02 3.236e+02 3.873e+02 8.211e+02, threshold=6.471e+02, percent-clipped=3.0 2023-04-28 22:46:57,779 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:46:57,810 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:47:20,943 INFO [train.py:904] (5/8) Epoch 8, batch 8900, loss[loss=0.1792, simple_loss=0.2689, pruned_loss=0.04476, over 12606.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2839, pruned_loss=0.05033, over 3047116.86 frames. ], batch size: 248, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:48:55,767 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:48:55,855 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:49:29,484 INFO [train.py:904] (5/8) Epoch 8, batch 8950, loss[loss=0.1614, simple_loss=0.2613, pruned_loss=0.03076, over 16845.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2843, pruned_loss=0.0511, over 3041752.53 frames. ], batch size: 90, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:49:40,960 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1621, 3.1075, 3.1048, 1.6822, 3.3867, 3.3802, 2.7936, 2.6555], device='cuda:5'), covar=tensor([0.0760, 0.0180, 0.0189, 0.1261, 0.0052, 0.0090, 0.0363, 0.0427], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0093, 0.0080, 0.0139, 0.0064, 0.0086, 0.0116, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 22:49:50,488 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.495e+02 3.051e+02 3.831e+02 8.360e+02, threshold=6.103e+02, percent-clipped=2.0 2023-04-28 22:49:54,999 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:08,341 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:30,244 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1795, 3.3157, 3.5706, 3.5377, 3.5504, 3.3182, 3.3841, 3.4220], device='cuda:5'), covar=tensor([0.0353, 0.0584, 0.0423, 0.0509, 0.0512, 0.0474, 0.0731, 0.0465], device='cuda:5'), in_proj_covar=tensor([0.0277, 0.0279, 0.0282, 0.0274, 0.0321, 0.0300, 0.0388, 0.0245], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 22:50:32,347 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:50:46,304 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:51:17,204 INFO [train.py:904] (5/8) Epoch 8, batch 9000, loss[loss=0.1812, simple_loss=0.2681, pruned_loss=0.04711, over 16703.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2808, pruned_loss=0.04956, over 3060778.80 frames. ], batch size: 76, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:51:17,204 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 22:51:27,538 INFO [train.py:938] (5/8) Epoch 8, validation: loss=0.1608, simple_loss=0.265, pruned_loss=0.02828, over 944034.00 frames. 2023-04-28 22:51:27,540 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 22:52:04,336 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:52:26,508 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:53:14,319 INFO [train.py:904] (5/8) Epoch 8, batch 9050, loss[loss=0.1892, simple_loss=0.2698, pruned_loss=0.05434, over 16638.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2824, pruned_loss=0.05027, over 3076638.19 frames. ], batch size: 134, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:53:35,359 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.777e+02 3.475e+02 4.225e+02 7.695e+02, threshold=6.949e+02, percent-clipped=5.0 2023-04-28 22:53:48,958 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:54:23,191 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3941, 3.2843, 3.4408, 3.5152, 3.5592, 3.2354, 3.5170, 3.5893], device='cuda:5'), covar=tensor([0.1034, 0.0850, 0.1003, 0.0590, 0.0616, 0.2498, 0.0904, 0.0757], device='cuda:5'), in_proj_covar=tensor([0.0444, 0.0549, 0.0675, 0.0563, 0.0427, 0.0421, 0.0444, 0.0491], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 22:54:50,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0453, 2.6577, 2.6647, 1.8542, 2.8787, 2.8863, 2.5135, 2.4367], device='cuda:5'), covar=tensor([0.0703, 0.0178, 0.0195, 0.0924, 0.0093, 0.0153, 0.0388, 0.0385], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0093, 0.0080, 0.0137, 0.0064, 0.0086, 0.0115, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 22:54:58,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1581, 4.1547, 4.6390, 4.5942, 4.5917, 4.2729, 4.3086, 4.1641], device='cuda:5'), covar=tensor([0.0285, 0.0424, 0.0377, 0.0415, 0.0344, 0.0326, 0.0745, 0.0419], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0281, 0.0283, 0.0276, 0.0321, 0.0301, 0.0390, 0.0246], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-28 22:54:59,319 INFO [train.py:904] (5/8) Epoch 8, batch 9100, loss[loss=0.1899, simple_loss=0.2841, pruned_loss=0.04789, over 16924.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2828, pruned_loss=0.05131, over 3079730.53 frames. ], batch size: 116, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:56:59,477 INFO [train.py:904] (5/8) Epoch 8, batch 9150, loss[loss=0.1956, simple_loss=0.2908, pruned_loss=0.05025, over 15123.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2829, pruned_loss=0.05075, over 3069463.62 frames. ], batch size: 190, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:57:20,188 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.760e+02 3.111e+02 3.901e+02 6.426e+02, threshold=6.222e+02, percent-clipped=0.0 2023-04-28 22:58:26,813 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:58:44,966 INFO [train.py:904] (5/8) Epoch 8, batch 9200, loss[loss=0.1552, simple_loss=0.2374, pruned_loss=0.03648, over 11661.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2784, pruned_loss=0.0495, over 3071649.77 frames. ], batch size: 247, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:59:43,163 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80283.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:59:55,967 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:00:22,119 INFO [train.py:904] (5/8) Epoch 8, batch 9250, loss[loss=0.163, simple_loss=0.2448, pruned_loss=0.04066, over 12367.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2779, pruned_loss=0.04956, over 3061225.35 frames. ], batch size: 248, lr: 8.34e-03, grad_scale: 4.0 2023-04-28 23:00:33,526 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8971, 5.1784, 4.9898, 4.9588, 4.5827, 4.6457, 4.6207, 5.2413], device='cuda:5'), covar=tensor([0.0820, 0.0814, 0.0878, 0.0600, 0.0747, 0.0828, 0.0764, 0.0747], device='cuda:5'), in_proj_covar=tensor([0.0452, 0.0574, 0.0481, 0.0390, 0.0357, 0.0382, 0.0478, 0.0427], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:00:42,873 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.839e+02 3.478e+02 4.215e+02 8.620e+02, threshold=6.955e+02, percent-clipped=3.0 2023-04-28 23:00:44,597 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:00:52,046 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1676, 3.4823, 3.3913, 2.2338, 3.2845, 3.5123, 3.3704, 1.8055], device='cuda:5'), covar=tensor([0.0334, 0.0024, 0.0027, 0.0284, 0.0056, 0.0039, 0.0036, 0.0364], device='cuda:5'), in_proj_covar=tensor([0.0121, 0.0060, 0.0062, 0.0117, 0.0067, 0.0076, 0.0067, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 23:01:56,365 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80344.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:02:12,084 INFO [train.py:904] (5/8) Epoch 8, batch 9300, loss[loss=0.1651, simple_loss=0.2536, pruned_loss=0.03834, over 16721.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2763, pruned_loss=0.04918, over 3044905.85 frames. ], batch size: 134, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:02:32,768 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:03:42,315 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80393.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:03:57,506 INFO [train.py:904] (5/8) Epoch 8, batch 9350, loss[loss=0.1708, simple_loss=0.2525, pruned_loss=0.04457, over 12009.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2762, pruned_loss=0.04923, over 3047603.20 frames. ], batch size: 250, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:04:07,773 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6639, 4.6690, 4.5131, 3.9950, 4.5588, 1.8440, 4.3421, 4.4734], device='cuda:5'), covar=tensor([0.0090, 0.0109, 0.0112, 0.0344, 0.0081, 0.2157, 0.0129, 0.0186], device='cuda:5'), in_proj_covar=tensor([0.0106, 0.0094, 0.0137, 0.0129, 0.0110, 0.0161, 0.0125, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:04:22,276 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.742e+02 2.575e+02 3.007e+02 3.554e+02 5.975e+02, threshold=6.013e+02, percent-clipped=0.0 2023-04-28 23:04:34,138 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:05:40,911 INFO [train.py:904] (5/8) Epoch 8, batch 9400, loss[loss=0.1934, simple_loss=0.2865, pruned_loss=0.05009, over 15287.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2759, pruned_loss=0.049, over 3041261.44 frames. ], batch size: 190, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:05:46,287 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:05,127 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:10,060 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:23,757 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9088, 3.3104, 3.3008, 2.0907, 3.1287, 3.3015, 3.1879, 1.8469], device='cuda:5'), covar=tensor([0.0374, 0.0027, 0.0031, 0.0299, 0.0059, 0.0056, 0.0052, 0.0336], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0059, 0.0061, 0.0115, 0.0066, 0.0075, 0.0066, 0.0111], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 23:07:07,320 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:07:19,866 INFO [train.py:904] (5/8) Epoch 8, batch 9450, loss[loss=0.203, simple_loss=0.2965, pruned_loss=0.05473, over 16398.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2789, pruned_loss=0.04985, over 3038440.84 frames. ], batch size: 146, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:07:38,814 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.597e+02 3.133e+02 4.086e+02 1.022e+03, threshold=6.266e+02, percent-clipped=6.0 2023-04-28 23:08:04,859 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:08:32,876 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 23:08:42,679 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6038, 3.8232, 2.0628, 4.1678, 2.7629, 4.0408, 2.1798, 2.9067], device='cuda:5'), covar=tensor([0.0189, 0.0248, 0.1572, 0.0086, 0.0814, 0.0441, 0.1537, 0.0670], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0151, 0.0180, 0.0100, 0.0161, 0.0185, 0.0189, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-28 23:08:58,875 INFO [train.py:904] (5/8) Epoch 8, batch 9500, loss[loss=0.1762, simple_loss=0.2753, pruned_loss=0.03861, over 16825.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2778, pruned_loss=0.04895, over 3054240.40 frames. ], batch size: 96, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:09:08,370 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:10:46,576 INFO [train.py:904] (5/8) Epoch 8, batch 9550, loss[loss=0.1815, simple_loss=0.2683, pruned_loss=0.04739, over 12376.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2777, pruned_loss=0.04897, over 3070690.73 frames. ], batch size: 248, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:11:10,125 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.584e+02 3.110e+02 3.705e+02 8.746e+02, threshold=6.220e+02, percent-clipped=3.0 2023-04-28 23:12:03,964 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80639.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:12:07,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3446, 5.6253, 5.4306, 5.3626, 5.0093, 4.9057, 5.1326, 5.6970], device='cuda:5'), covar=tensor([0.0882, 0.0832, 0.0804, 0.0528, 0.0659, 0.0677, 0.0817, 0.0757], device='cuda:5'), in_proj_covar=tensor([0.0445, 0.0566, 0.0466, 0.0386, 0.0349, 0.0375, 0.0470, 0.0419], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:12:10,360 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0800, 1.3795, 1.7729, 2.0753, 2.0952, 2.1846, 1.5004, 2.1637], device='cuda:5'), covar=tensor([0.0125, 0.0344, 0.0213, 0.0206, 0.0196, 0.0136, 0.0340, 0.0089], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0157, 0.0142, 0.0139, 0.0146, 0.0103, 0.0155, 0.0094], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 23:12:27,100 INFO [train.py:904] (5/8) Epoch 8, batch 9600, loss[loss=0.2062, simple_loss=0.3013, pruned_loss=0.05553, over 16263.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2788, pruned_loss=0.04957, over 3059026.90 frames. ], batch size: 165, lr: 8.32e-03, grad_scale: 8.0 2023-04-28 23:12:58,676 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 23:13:09,988 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3662, 2.0051, 2.0747, 3.7917, 1.9230, 2.4808, 2.1155, 2.0867], device='cuda:5'), covar=tensor([0.0674, 0.3111, 0.1946, 0.0364, 0.3558, 0.1846, 0.2754, 0.3245], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0348, 0.0298, 0.0305, 0.0384, 0.0382, 0.0313, 0.0408], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:14:15,061 INFO [train.py:904] (5/8) Epoch 8, batch 9650, loss[loss=0.2042, simple_loss=0.2936, pruned_loss=0.05738, over 15165.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2802, pruned_loss=0.04934, over 3071703.76 frames. ], batch size: 190, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:14:42,863 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.581e+02 3.070e+02 3.886e+02 7.619e+02, threshold=6.139e+02, percent-clipped=2.0 2023-04-28 23:15:22,852 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4430, 1.5651, 1.9789, 2.4945, 2.3736, 2.5627, 1.8196, 2.6441], device='cuda:5'), covar=tensor([0.0115, 0.0334, 0.0220, 0.0169, 0.0185, 0.0112, 0.0299, 0.0071], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0155, 0.0141, 0.0138, 0.0144, 0.0102, 0.0154, 0.0093], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 23:15:30,076 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:15:38,591 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4268, 3.3003, 2.6352, 2.1306, 2.2958, 2.1566, 3.3654, 3.1557], device='cuda:5'), covar=tensor([0.2345, 0.0716, 0.1382, 0.2023, 0.1912, 0.1656, 0.0440, 0.0844], device='cuda:5'), in_proj_covar=tensor([0.0280, 0.0241, 0.0267, 0.0253, 0.0245, 0.0203, 0.0246, 0.0257], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:15:59,675 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:16:03,280 INFO [train.py:904] (5/8) Epoch 8, batch 9700, loss[loss=0.175, simple_loss=0.27, pruned_loss=0.03999, over 16690.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2789, pruned_loss=0.04873, over 3079866.19 frames. ], batch size: 76, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:16:38,055 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3948, 3.6508, 3.7179, 1.6627, 3.8745, 4.0264, 3.0495, 2.9207], device='cuda:5'), covar=tensor([0.0774, 0.0148, 0.0151, 0.1273, 0.0067, 0.0076, 0.0338, 0.0421], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0093, 0.0079, 0.0137, 0.0065, 0.0085, 0.0114, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-28 23:16:46,546 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:03,561 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3644, 4.2883, 4.7578, 4.6904, 4.7073, 4.4431, 4.4278, 4.3152], device='cuda:5'), covar=tensor([0.0263, 0.0543, 0.0375, 0.0494, 0.0447, 0.0317, 0.0704, 0.0383], device='cuda:5'), in_proj_covar=tensor([0.0272, 0.0273, 0.0277, 0.0273, 0.0312, 0.0295, 0.0379, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-28 23:17:38,582 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:46,352 INFO [train.py:904] (5/8) Epoch 8, batch 9750, loss[loss=0.1953, simple_loss=0.2833, pruned_loss=0.05368, over 17003.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.278, pruned_loss=0.04891, over 3076052.67 frames. ], batch size: 109, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:18:08,287 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.800e+02 3.435e+02 4.029e+02 7.858e+02, threshold=6.871e+02, percent-clipped=2.0 2023-04-28 23:18:21,098 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:18:55,757 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:19:24,258 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:19:26,329 INFO [train.py:904] (5/8) Epoch 8, batch 9800, loss[loss=0.2033, simple_loss=0.2981, pruned_loss=0.05422, over 15349.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2784, pruned_loss=0.04805, over 3091667.19 frames. ], batch size: 190, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:20:31,446 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 23:21:11,977 INFO [train.py:904] (5/8) Epoch 8, batch 9850, loss[loss=0.1766, simple_loss=0.2549, pruned_loss=0.0492, over 12453.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2794, pruned_loss=0.04816, over 3080279.46 frames. ], batch size: 249, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:21:27,183 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9709, 3.6572, 3.5146, 1.8227, 2.8819, 2.4211, 3.4399, 3.5531], device='cuda:5'), covar=tensor([0.0255, 0.0451, 0.0470, 0.1696, 0.0707, 0.0878, 0.0707, 0.0695], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0124, 0.0152, 0.0138, 0.0131, 0.0123, 0.0131, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-28 23:21:33,320 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.396e+02 3.091e+02 3.765e+02 1.204e+03, threshold=6.182e+02, percent-clipped=3.0 2023-04-28 23:22:06,302 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4322, 3.0281, 2.6712, 2.2309, 2.1865, 2.0381, 2.9348, 2.8242], device='cuda:5'), covar=tensor([0.2072, 0.0690, 0.1276, 0.1988, 0.2048, 0.1757, 0.0414, 0.1003], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0240, 0.0267, 0.0253, 0.0243, 0.0204, 0.0245, 0.0256], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:22:37,355 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:23:04,053 INFO [train.py:904] (5/8) Epoch 8, batch 9900, loss[loss=0.2023, simple_loss=0.2779, pruned_loss=0.06333, over 12477.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2796, pruned_loss=0.0484, over 3064640.93 frames. ], batch size: 250, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:24:08,445 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6650, 2.7085, 1.7423, 2.8498, 2.1603, 2.8396, 1.9385, 2.4382], device='cuda:5'), covar=tensor([0.0223, 0.0320, 0.1354, 0.0155, 0.0725, 0.0447, 0.1324, 0.0548], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0152, 0.0183, 0.0100, 0.0162, 0.0186, 0.0191, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 23:24:29,640 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:25:03,329 INFO [train.py:904] (5/8) Epoch 8, batch 9950, loss[loss=0.1902, simple_loss=0.2848, pruned_loss=0.04782, over 15483.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2824, pruned_loss=0.04876, over 3089120.66 frames. ], batch size: 191, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:25:29,463 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.486e+02 3.063e+02 3.674e+02 7.431e+02, threshold=6.127e+02, percent-clipped=1.0 2023-04-28 23:25:36,241 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5040, 1.4391, 1.9451, 2.4569, 2.3192, 2.4700, 1.8289, 2.5742], device='cuda:5'), covar=tensor([0.0098, 0.0321, 0.0178, 0.0150, 0.0168, 0.0102, 0.0284, 0.0068], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0158, 0.0142, 0.0139, 0.0146, 0.0104, 0.0156, 0.0093], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 23:27:02,058 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:27:07,194 INFO [train.py:904] (5/8) Epoch 8, batch 10000, loss[loss=0.1859, simple_loss=0.2867, pruned_loss=0.04255, over 16223.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2809, pruned_loss=0.04835, over 3079112.44 frames. ], batch size: 165, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:28:28,752 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4105, 3.3091, 2.7979, 2.1156, 2.2344, 2.1595, 3.4386, 3.1592], device='cuda:5'), covar=tensor([0.2388, 0.0628, 0.1222, 0.1969, 0.2086, 0.1621, 0.0377, 0.0838], device='cuda:5'), in_proj_covar=tensor([0.0282, 0.0241, 0.0268, 0.0254, 0.0244, 0.0205, 0.0245, 0.0257], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:28:30,856 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:28:40,810 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81097.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:28:47,985 INFO [train.py:904] (5/8) Epoch 8, batch 10050, loss[loss=0.1816, simple_loss=0.2719, pruned_loss=0.04567, over 16630.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.281, pruned_loss=0.0483, over 3077594.63 frames. ], batch size: 62, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:29:08,277 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.518e+02 2.989e+02 3.596e+02 8.754e+02, threshold=5.978e+02, percent-clipped=2.0 2023-04-28 23:29:15,158 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:21,145 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:40,310 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:30:18,753 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81150.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:30:21,533 INFO [train.py:904] (5/8) Epoch 8, batch 10100, loss[loss=0.1726, simple_loss=0.2692, pruned_loss=0.03797, over 16273.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2809, pruned_loss=0.04885, over 3053042.18 frames. ], batch size: 146, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:30:51,852 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:14,359 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:37,453 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:32:08,887 INFO [train.py:904] (5/8) Epoch 9, batch 0, loss[loss=0.2196, simple_loss=0.3084, pruned_loss=0.0654, over 17109.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3084, pruned_loss=0.0654, over 17109.00 frames. ], batch size: 47, lr: 7.85e-03, grad_scale: 8.0 2023-04-28 23:32:08,887 INFO [train.py:929] (5/8) Computing validation loss 2023-04-28 23:32:16,262 INFO [train.py:938] (5/8) Epoch 9, validation: loss=0.1602, simple_loss=0.2637, pruned_loss=0.02837, over 944034.00 frames. 2023-04-28 23:32:16,263 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-28 23:32:36,786 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.828e+02 3.571e+02 4.681e+02 1.164e+03, threshold=7.142e+02, percent-clipped=9.0 2023-04-28 23:33:21,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4367, 2.5745, 2.1071, 2.4163, 2.8863, 2.7667, 3.4580, 3.2027], device='cuda:5'), covar=tensor([0.0059, 0.0236, 0.0316, 0.0261, 0.0168, 0.0218, 0.0109, 0.0136], device='cuda:5'), in_proj_covar=tensor([0.0107, 0.0184, 0.0180, 0.0179, 0.0178, 0.0182, 0.0170, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:33:25,082 INFO [train.py:904] (5/8) Epoch 9, batch 50, loss[loss=0.21, simple_loss=0.2905, pruned_loss=0.06471, over 16453.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.2983, pruned_loss=0.07327, over 747594.42 frames. ], batch size: 68, lr: 7.85e-03, grad_scale: 1.0 2023-04-28 23:34:31,174 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:34:34,759 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81301.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:34:35,505 INFO [train.py:904] (5/8) Epoch 9, batch 100, loss[loss=0.2212, simple_loss=0.2988, pruned_loss=0.07183, over 16519.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2908, pruned_loss=0.06675, over 1305915.37 frames. ], batch size: 75, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:34:54,540 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.676e+02 3.088e+02 4.301e+02 1.054e+03, threshold=6.177e+02, percent-clipped=2.0 2023-04-28 23:35:42,960 INFO [train.py:904] (5/8) Epoch 9, batch 150, loss[loss=0.1985, simple_loss=0.2926, pruned_loss=0.05222, over 17018.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2891, pruned_loss=0.06496, over 1747642.43 frames. ], batch size: 50, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:35:55,903 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:35:58,297 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:40,134 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:54,433 INFO [train.py:904] (5/8) Epoch 9, batch 200, loss[loss=0.2337, simple_loss=0.2865, pruned_loss=0.0905, over 16885.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2871, pruned_loss=0.06343, over 2095011.15 frames. ], batch size: 109, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:36:59,838 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5558, 4.6594, 4.7629, 4.7513, 4.6524, 5.3018, 4.9337, 4.5710], device='cuda:5'), covar=tensor([0.1189, 0.2032, 0.1951, 0.2045, 0.3170, 0.1115, 0.1371, 0.2529], device='cuda:5'), in_proj_covar=tensor([0.0316, 0.0448, 0.0475, 0.0394, 0.0523, 0.0495, 0.0382, 0.0526], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 23:37:13,043 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.694e+02 3.180e+02 3.951e+02 1.154e+03, threshold=6.361e+02, percent-clipped=2.0 2023-04-28 23:37:32,096 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:37:46,218 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:01,971 INFO [train.py:904] (5/8) Epoch 9, batch 250, loss[loss=0.2219, simple_loss=0.2902, pruned_loss=0.07677, over 16515.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2849, pruned_loss=0.06325, over 2368028.39 frames. ], batch size: 68, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:38:29,701 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:36,872 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:39:10,404 INFO [train.py:904] (5/8) Epoch 9, batch 300, loss[loss=0.2108, simple_loss=0.2813, pruned_loss=0.07014, over 16795.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2816, pruned_loss=0.06123, over 2585283.35 frames. ], batch size: 124, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:39:29,675 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.475e+02 2.919e+02 3.755e+02 7.155e+02, threshold=5.837e+02, percent-clipped=3.0 2023-04-28 23:39:56,606 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0460, 2.9188, 3.1246, 2.1576, 2.9433, 3.1667, 2.9529, 1.8623], device='cuda:5'), covar=tensor([0.0344, 0.0094, 0.0037, 0.0279, 0.0060, 0.0050, 0.0060, 0.0327], device='cuda:5'), in_proj_covar=tensor([0.0121, 0.0065, 0.0064, 0.0120, 0.0068, 0.0078, 0.0069, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-28 23:40:16,264 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0301, 4.3729, 3.1773, 2.3210, 2.8702, 2.4004, 4.6041, 3.9657], device='cuda:5'), covar=tensor([0.2133, 0.0587, 0.1385, 0.1987, 0.2458, 0.1657, 0.0344, 0.0794], device='cuda:5'), in_proj_covar=tensor([0.0293, 0.0253, 0.0279, 0.0267, 0.0266, 0.0215, 0.0258, 0.0277], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:40:17,835 INFO [train.py:904] (5/8) Epoch 9, batch 350, loss[loss=0.2056, simple_loss=0.2726, pruned_loss=0.06934, over 16801.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.279, pruned_loss=0.06013, over 2741970.64 frames. ], batch size: 102, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:41:23,826 INFO [train.py:904] (5/8) Epoch 9, batch 400, loss[loss=0.1894, simple_loss=0.2864, pruned_loss=0.04616, over 17256.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2777, pruned_loss=0.06008, over 2868198.34 frames. ], batch size: 52, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:41:43,592 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.386e+02 2.843e+02 3.463e+02 6.249e+02, threshold=5.687e+02, percent-clipped=1.0 2023-04-28 23:42:33,325 INFO [train.py:904] (5/8) Epoch 9, batch 450, loss[loss=0.2009, simple_loss=0.2756, pruned_loss=0.06314, over 16566.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.276, pruned_loss=0.05889, over 2971591.90 frames. ], batch size: 75, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:42:37,366 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:40,771 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:50,979 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5720, 4.9276, 4.6729, 4.7377, 4.4608, 4.3579, 4.4433, 4.9618], device='cuda:5'), covar=tensor([0.1018, 0.0755, 0.0927, 0.0537, 0.0668, 0.1100, 0.0838, 0.0753], device='cuda:5'), in_proj_covar=tensor([0.0495, 0.0634, 0.0527, 0.0431, 0.0391, 0.0412, 0.0524, 0.0467], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:42:58,897 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:43:09,172 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0506, 5.4244, 5.5824, 5.4642, 5.4268, 6.0181, 5.5953, 5.2775], device='cuda:5'), covar=tensor([0.0724, 0.1692, 0.1757, 0.1880, 0.2775, 0.0901, 0.1197, 0.2210], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0463, 0.0493, 0.0411, 0.0546, 0.0517, 0.0392, 0.0549], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 23:43:42,353 INFO [train.py:904] (5/8) Epoch 9, batch 500, loss[loss=0.1454, simple_loss=0.2374, pruned_loss=0.02665, over 16839.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.274, pruned_loss=0.05729, over 3051965.78 frames. ], batch size: 42, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:44:01,900 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.446e+02 2.865e+02 3.583e+02 8.926e+02, threshold=5.729e+02, percent-clipped=4.0 2023-04-28 23:44:22,636 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 23:44:50,831 INFO [train.py:904] (5/8) Epoch 9, batch 550, loss[loss=0.1898, simple_loss=0.2569, pruned_loss=0.0613, over 16981.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2727, pruned_loss=0.05682, over 3104265.77 frames. ], batch size: 41, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:45:09,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7120, 4.8181, 5.2406, 5.2493, 5.2355, 4.9073, 4.8495, 4.6418], device='cuda:5'), covar=tensor([0.0238, 0.0380, 0.0455, 0.0399, 0.0365, 0.0298, 0.0762, 0.0362], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0308, 0.0308, 0.0301, 0.0346, 0.0328, 0.0421, 0.0264], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 23:45:19,997 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:46:02,612 INFO [train.py:904] (5/8) Epoch 9, batch 600, loss[loss=0.2139, simple_loss=0.2876, pruned_loss=0.07007, over 16731.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2721, pruned_loss=0.0568, over 3154917.00 frames. ], batch size: 124, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:46:15,454 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7850, 4.8678, 5.3699, 5.3630, 5.3385, 5.0056, 4.9225, 4.6630], device='cuda:5'), covar=tensor([0.0276, 0.0412, 0.0430, 0.0385, 0.0418, 0.0317, 0.0902, 0.0405], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0310, 0.0312, 0.0303, 0.0348, 0.0330, 0.0425, 0.0266], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-28 23:46:20,959 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5714, 2.0564, 2.3345, 4.1713, 2.1672, 2.5998, 2.1807, 2.3185], device='cuda:5'), covar=tensor([0.0856, 0.2938, 0.1802, 0.0382, 0.3145, 0.1840, 0.2767, 0.2554], device='cuda:5'), in_proj_covar=tensor([0.0347, 0.0369, 0.0311, 0.0323, 0.0399, 0.0408, 0.0330, 0.0434], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:46:21,575 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.670e+02 3.296e+02 4.052e+02 8.860e+02, threshold=6.592e+02, percent-clipped=6.0 2023-04-28 23:46:26,576 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:46:28,733 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4036, 3.8422, 3.9135, 2.1472, 3.1831, 2.6733, 3.8501, 3.8960], device='cuda:5'), covar=tensor([0.0278, 0.0682, 0.0462, 0.1661, 0.0716, 0.0844, 0.0596, 0.1100], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0135, 0.0157, 0.0143, 0.0135, 0.0125, 0.0133, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-28 23:47:09,565 INFO [train.py:904] (5/8) Epoch 9, batch 650, loss[loss=0.1858, simple_loss=0.2555, pruned_loss=0.05806, over 16866.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2709, pruned_loss=0.05602, over 3184612.53 frames. ], batch size: 90, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:47:17,335 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4352, 3.0039, 2.6357, 2.2872, 2.2370, 2.2006, 2.8850, 2.8152], device='cuda:5'), covar=tensor([0.2029, 0.0777, 0.1279, 0.1684, 0.1904, 0.1623, 0.0447, 0.0873], device='cuda:5'), in_proj_covar=tensor([0.0295, 0.0256, 0.0281, 0.0269, 0.0274, 0.0217, 0.0259, 0.0284], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:47:33,629 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2054, 5.6525, 5.9140, 5.5984, 5.8731, 6.3016, 5.9150, 5.5752], device='cuda:5'), covar=tensor([0.0733, 0.1606, 0.1466, 0.1742, 0.1846, 0.0772, 0.1143, 0.2125], device='cuda:5'), in_proj_covar=tensor([0.0325, 0.0458, 0.0488, 0.0406, 0.0540, 0.0513, 0.0387, 0.0541], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 23:48:18,162 INFO [train.py:904] (5/8) Epoch 9, batch 700, loss[loss=0.2014, simple_loss=0.2659, pruned_loss=0.06842, over 16837.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2704, pruned_loss=0.05588, over 3220783.83 frames. ], batch size: 116, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:48:23,320 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 23:48:24,277 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 23:48:37,201 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.553e+02 3.037e+02 4.738e+02 2.852e+03, threshold=6.073e+02, percent-clipped=12.0 2023-04-28 23:49:25,164 INFO [train.py:904] (5/8) Epoch 9, batch 750, loss[loss=0.184, simple_loss=0.2658, pruned_loss=0.05112, over 16955.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2711, pruned_loss=0.05654, over 3241841.43 frames. ], batch size: 41, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:49:29,102 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:49:31,615 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:05,308 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1666, 5.7032, 5.8874, 5.5589, 5.6557, 6.1809, 5.8069, 5.5043], device='cuda:5'), covar=tensor([0.0743, 0.1704, 0.1766, 0.2043, 0.2582, 0.0947, 0.1207, 0.2260], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0465, 0.0497, 0.0412, 0.0550, 0.0522, 0.0395, 0.0547], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-28 23:50:14,424 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:31,368 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8059, 3.3793, 2.9079, 5.1032, 4.1781, 4.6786, 1.8695, 3.4411], device='cuda:5'), covar=tensor([0.1428, 0.0587, 0.1087, 0.0123, 0.0338, 0.0311, 0.1454, 0.0674], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0153, 0.0174, 0.0124, 0.0190, 0.0208, 0.0173, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-28 23:50:37,840 INFO [train.py:904] (5/8) Epoch 9, batch 800, loss[loss=0.1958, simple_loss=0.2667, pruned_loss=0.06246, over 16867.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2703, pruned_loss=0.05616, over 3254775.91 frames. ], batch size: 96, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:50:39,973 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82003.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:42,265 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:56,933 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.619e+02 2.968e+02 3.371e+02 5.495e+02, threshold=5.935e+02, percent-clipped=0.0 2023-04-28 23:51:10,871 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 23:51:40,961 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82048.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:51:43,261 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0529, 1.8184, 2.4150, 2.7807, 2.5935, 3.3802, 2.3893, 3.1806], device='cuda:5'), covar=tensor([0.0141, 0.0332, 0.0214, 0.0197, 0.0204, 0.0109, 0.0254, 0.0100], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0165, 0.0150, 0.0149, 0.0156, 0.0112, 0.0163, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-28 23:51:45,045 INFO [train.py:904] (5/8) Epoch 9, batch 850, loss[loss=0.1998, simple_loss=0.2645, pruned_loss=0.06755, over 16896.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2692, pruned_loss=0.05557, over 3272662.18 frames. ], batch size: 109, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:51:49,643 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:51:56,998 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 2023-04-28 23:52:54,669 INFO [train.py:904] (5/8) Epoch 9, batch 900, loss[loss=0.1814, simple_loss=0.2681, pruned_loss=0.04741, over 16546.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2689, pruned_loss=0.05489, over 3288171.45 frames. ], batch size: 75, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:53:13,860 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.604e+02 3.157e+02 3.589e+02 5.638e+02, threshold=6.315e+02, percent-clipped=0.0 2023-04-28 23:53:15,039 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:53:32,095 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9652, 3.8206, 4.0179, 4.1598, 4.2359, 3.8061, 4.0664, 4.2108], device='cuda:5'), covar=tensor([0.1177, 0.0906, 0.1174, 0.0565, 0.0548, 0.1454, 0.1329, 0.0560], device='cuda:5'), in_proj_covar=tensor([0.0512, 0.0626, 0.0785, 0.0637, 0.0478, 0.0474, 0.0504, 0.0554], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:54:03,816 INFO [train.py:904] (5/8) Epoch 9, batch 950, loss[loss=0.1997, simple_loss=0.2845, pruned_loss=0.05742, over 17063.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2689, pruned_loss=0.05518, over 3291459.03 frames. ], batch size: 53, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:11,105 INFO [train.py:904] (5/8) Epoch 9, batch 1000, loss[loss=0.177, simple_loss=0.2637, pruned_loss=0.04517, over 17135.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2682, pruned_loss=0.0545, over 3304327.64 frames. ], batch size: 47, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:31,952 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 2.342e+02 2.920e+02 3.464e+02 5.943e+02, threshold=5.839e+02, percent-clipped=0.0 2023-04-28 23:55:37,153 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2713, 1.9762, 2.1928, 3.6881, 2.0528, 2.3909, 2.0978, 2.1571], device='cuda:5'), covar=tensor([0.0916, 0.3065, 0.1970, 0.0485, 0.3171, 0.1969, 0.3016, 0.2777], device='cuda:5'), in_proj_covar=tensor([0.0353, 0.0374, 0.0316, 0.0327, 0.0401, 0.0416, 0.0333, 0.0439], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:56:11,982 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9317, 2.3137, 2.4398, 4.6427, 2.2192, 2.7799, 2.4585, 2.6282], device='cuda:5'), covar=tensor([0.0736, 0.3051, 0.1867, 0.0318, 0.3446, 0.2169, 0.2601, 0.3078], device='cuda:5'), in_proj_covar=tensor([0.0353, 0.0374, 0.0316, 0.0327, 0.0400, 0.0416, 0.0333, 0.0440], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:56:17,992 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7579, 2.4742, 1.8685, 2.2491, 2.9043, 2.7225, 3.1093, 2.9999], device='cuda:5'), covar=tensor([0.0108, 0.0246, 0.0348, 0.0309, 0.0129, 0.0198, 0.0149, 0.0158], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0194, 0.0189, 0.0189, 0.0189, 0.0192, 0.0194, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-28 23:56:20,449 INFO [train.py:904] (5/8) Epoch 9, batch 1050, loss[loss=0.1816, simple_loss=0.2614, pruned_loss=0.05083, over 16568.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.268, pruned_loss=0.05484, over 3308539.34 frames. ], batch size: 75, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:28,540 INFO [train.py:904] (5/8) Epoch 9, batch 1100, loss[loss=0.1918, simple_loss=0.257, pruned_loss=0.06331, over 16924.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2674, pruned_loss=0.05457, over 3309832.45 frames. ], batch size: 109, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:47,217 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.514e+02 3.056e+02 3.616e+02 1.290e+03, threshold=6.113e+02, percent-clipped=7.0 2023-04-28 23:58:01,857 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:24,287 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:28,067 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:35,315 INFO [train.py:904] (5/8) Epoch 9, batch 1150, loss[loss=0.1934, simple_loss=0.2617, pruned_loss=0.06258, over 16432.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2666, pruned_loss=0.05414, over 3307991.81 frames. ], batch size: 146, lr: 7.79e-03, grad_scale: 4.0 2023-04-28 23:59:04,896 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:44,525 INFO [train.py:904] (5/8) Epoch 9, batch 1200, loss[loss=0.1892, simple_loss=0.2713, pruned_loss=0.05359, over 17188.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2659, pruned_loss=0.05374, over 3314765.03 frames. ], batch size: 46, lr: 7.79e-03, grad_scale: 8.0 2023-04-28 23:59:50,722 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:55,282 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82411.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:00:02,699 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.375e+02 3.047e+02 3.937e+02 1.504e+03, threshold=6.095e+02, percent-clipped=2.0 2023-04-29 00:00:50,237 INFO [train.py:904] (5/8) Epoch 9, batch 1250, loss[loss=0.1722, simple_loss=0.2676, pruned_loss=0.03844, over 17100.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2658, pruned_loss=0.05381, over 3316946.72 frames. ], batch size: 49, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:01:11,926 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-29 00:01:47,143 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:01:47,200 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1785, 3.6488, 3.4795, 2.1117, 2.9745, 2.4440, 3.5912, 3.7262], device='cuda:5'), covar=tensor([0.0277, 0.0607, 0.0503, 0.1528, 0.0708, 0.0875, 0.0558, 0.0711], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0139, 0.0156, 0.0142, 0.0135, 0.0125, 0.0136, 0.0148], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:01:58,481 INFO [train.py:904] (5/8) Epoch 9, batch 1300, loss[loss=0.1891, simple_loss=0.2614, pruned_loss=0.05842, over 16740.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2658, pruned_loss=0.05357, over 3319267.88 frames. ], batch size: 83, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:02:18,041 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.404e+02 2.965e+02 3.987e+02 6.881e+02, threshold=5.930e+02, percent-clipped=4.0 2023-04-29 00:02:30,941 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0502, 4.1880, 1.9479, 4.6375, 2.8541, 4.5379, 2.1721, 3.2187], device='cuda:5'), covar=tensor([0.0168, 0.0223, 0.1600, 0.0095, 0.0734, 0.0347, 0.1427, 0.0590], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0163, 0.0186, 0.0116, 0.0166, 0.0203, 0.0195, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:02:38,203 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-29 00:03:05,242 INFO [train.py:904] (5/8) Epoch 9, batch 1350, loss[loss=0.1721, simple_loss=0.2545, pruned_loss=0.04492, over 16979.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2663, pruned_loss=0.05377, over 3328690.17 frames. ], batch size: 41, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:03:08,025 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:03:29,832 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:04:12,579 INFO [train.py:904] (5/8) Epoch 9, batch 1400, loss[loss=0.2003, simple_loss=0.2726, pruned_loss=0.06404, over 16919.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2656, pruned_loss=0.05356, over 3318359.52 frames. ], batch size: 109, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:04:31,646 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6525, 4.6130, 4.5444, 4.0403, 4.5407, 1.8973, 4.3062, 4.3985], device='cuda:5'), covar=tensor([0.0066, 0.0064, 0.0117, 0.0276, 0.0065, 0.2057, 0.0105, 0.0130], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0109, 0.0159, 0.0153, 0.0128, 0.0176, 0.0143, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:04:33,569 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.260e+02 2.970e+02 4.048e+02 1.598e+03, threshold=5.940e+02, percent-clipped=4.0 2023-04-29 00:04:38,294 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0229, 4.2512, 2.1838, 4.7924, 3.0461, 4.6760, 2.3982, 3.3485], device='cuda:5'), covar=tensor([0.0193, 0.0289, 0.1652, 0.0134, 0.0767, 0.0421, 0.1450, 0.0620], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0164, 0.0186, 0.0117, 0.0166, 0.0206, 0.0196, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:04:53,128 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:09,615 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:22,666 INFO [train.py:904] (5/8) Epoch 9, batch 1450, loss[loss=0.1809, simple_loss=0.2627, pruned_loss=0.04956, over 16786.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2644, pruned_loss=0.05301, over 3320956.32 frames. ], batch size: 57, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:05:57,635 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 00:06:07,178 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8771, 2.1922, 2.1757, 4.6291, 2.1345, 2.8297, 2.3229, 2.4397], device='cuda:5'), covar=tensor([0.0735, 0.3174, 0.2057, 0.0297, 0.3481, 0.2146, 0.2553, 0.3222], device='cuda:5'), in_proj_covar=tensor([0.0352, 0.0374, 0.0316, 0.0325, 0.0399, 0.0419, 0.0333, 0.0440], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:06:16,064 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:29,720 INFO [train.py:904] (5/8) Epoch 9, batch 1500, loss[loss=0.1815, simple_loss=0.279, pruned_loss=0.04201, over 17051.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2644, pruned_loss=0.05313, over 3321556.42 frames. ], batch size: 50, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:06:30,001 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:42,874 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:51,618 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.540e+02 3.048e+02 3.562e+02 8.559e+02, threshold=6.096e+02, percent-clipped=4.0 2023-04-29 00:07:39,192 INFO [train.py:904] (5/8) Epoch 9, batch 1550, loss[loss=0.1771, simple_loss=0.271, pruned_loss=0.04156, over 17127.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2658, pruned_loss=0.05449, over 3318744.29 frames. ], batch size: 48, lr: 7.77e-03, grad_scale: 4.0 2023-04-29 00:07:49,787 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:07:58,601 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 00:08:12,324 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7138, 4.2405, 2.7720, 2.1468, 2.9642, 2.2935, 4.4680, 3.9223], device='cuda:5'), covar=tensor([0.2633, 0.0610, 0.1758, 0.2175, 0.2398, 0.1740, 0.0395, 0.0907], device='cuda:5'), in_proj_covar=tensor([0.0292, 0.0255, 0.0276, 0.0265, 0.0274, 0.0214, 0.0257, 0.0285], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:08:29,476 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8137, 2.7279, 2.1215, 2.3902, 3.0541, 2.8934, 3.7159, 3.3532], device='cuda:5'), covar=tensor([0.0056, 0.0275, 0.0355, 0.0327, 0.0173, 0.0232, 0.0126, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0199, 0.0192, 0.0194, 0.0194, 0.0197, 0.0200, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:08:48,753 INFO [train.py:904] (5/8) Epoch 9, batch 1600, loss[loss=0.1959, simple_loss=0.2805, pruned_loss=0.05569, over 16488.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2688, pruned_loss=0.05532, over 3307821.62 frames. ], batch size: 68, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:09:09,711 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.591e+02 3.276e+02 4.037e+02 8.145e+02, threshold=6.551e+02, percent-clipped=5.0 2023-04-29 00:09:52,309 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:09:56,033 INFO [train.py:904] (5/8) Epoch 9, batch 1650, loss[loss=0.1894, simple_loss=0.2845, pruned_loss=0.04709, over 17237.00 frames. ], tot_loss[loss=0.192, simple_loss=0.271, pruned_loss=0.05646, over 3296473.37 frames. ], batch size: 45, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:11:05,717 INFO [train.py:904] (5/8) Epoch 9, batch 1700, loss[loss=0.2062, simple_loss=0.2841, pruned_loss=0.06416, over 16878.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2739, pruned_loss=0.05765, over 3295499.35 frames. ], batch size: 96, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:11:24,536 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.504e+02 2.790e+02 3.377e+02 4.170e+02 9.157e+02, threshold=6.754e+02, percent-clipped=3.0 2023-04-29 00:11:37,133 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:12:13,202 INFO [train.py:904] (5/8) Epoch 9, batch 1750, loss[loss=0.178, simple_loss=0.2586, pruned_loss=0.04872, over 16894.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2746, pruned_loss=0.05727, over 3299283.65 frames. ], batch size: 90, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:12:15,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-29 00:12:28,890 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:12:55,129 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 00:13:19,615 INFO [train.py:904] (5/8) Epoch 9, batch 1800, loss[loss=0.2115, simple_loss=0.2876, pruned_loss=0.06773, over 16446.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2754, pruned_loss=0.05694, over 3310164.41 frames. ], batch size: 75, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:13:20,633 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:13:40,266 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.500e+02 2.970e+02 3.668e+02 6.624e+02, threshold=5.940e+02, percent-clipped=0.0 2023-04-29 00:13:50,274 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:14:26,662 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83050.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:14:28,693 INFO [train.py:904] (5/8) Epoch 9, batch 1850, loss[loss=0.193, simple_loss=0.2718, pruned_loss=0.05713, over 16822.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2749, pruned_loss=0.05589, over 3314208.06 frames. ], batch size: 96, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:37,693 INFO [train.py:904] (5/8) Epoch 9, batch 1900, loss[loss=0.1966, simple_loss=0.2885, pruned_loss=0.05236, over 17072.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.275, pruned_loss=0.05577, over 3314062.05 frames. ], batch size: 53, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:59,154 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.476e+02 2.876e+02 3.479e+02 6.930e+02, threshold=5.751e+02, percent-clipped=2.0 2023-04-29 00:16:36,186 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6972, 3.3572, 2.6564, 5.0919, 4.3672, 4.6117, 1.5640, 3.4192], device='cuda:5'), covar=tensor([0.1332, 0.0567, 0.1139, 0.0187, 0.0288, 0.0386, 0.1507, 0.0674], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0152, 0.0175, 0.0126, 0.0195, 0.0210, 0.0171, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 00:16:42,428 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:16:46,840 INFO [train.py:904] (5/8) Epoch 9, batch 1950, loss[loss=0.2084, simple_loss=0.2941, pruned_loss=0.06139, over 16744.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2736, pruned_loss=0.05441, over 3323594.90 frames. ], batch size: 124, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:17:48,346 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:17:49,340 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:17:55,091 INFO [train.py:904] (5/8) Epoch 9, batch 2000, loss[loss=0.1777, simple_loss=0.2632, pruned_loss=0.04605, over 16839.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2743, pruned_loss=0.05445, over 3327765.38 frames. ], batch size: 42, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:18:05,275 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3613, 2.9922, 2.6049, 2.1603, 2.1459, 2.0908, 2.9042, 2.8374], device='cuda:5'), covar=tensor([0.2143, 0.0753, 0.1326, 0.2003, 0.2082, 0.1706, 0.0556, 0.0910], device='cuda:5'), in_proj_covar=tensor([0.0294, 0.0256, 0.0281, 0.0269, 0.0280, 0.0217, 0.0263, 0.0289], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:18:17,515 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.480e+02 2.857e+02 3.443e+02 6.900e+02, threshold=5.715e+02, percent-clipped=1.0 2023-04-29 00:18:21,654 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0836, 3.2806, 3.0997, 1.9803, 2.6837, 2.1383, 3.5846, 3.3335], device='cuda:5'), covar=tensor([0.0250, 0.0687, 0.0563, 0.1610, 0.0767, 0.0983, 0.0491, 0.0823], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0140, 0.0155, 0.0142, 0.0133, 0.0124, 0.0135, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:18:26,850 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:18:33,592 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 00:19:04,143 INFO [train.py:904] (5/8) Epoch 9, batch 2050, loss[loss=0.2028, simple_loss=0.2715, pruned_loss=0.06708, over 16638.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2751, pruned_loss=0.05588, over 3326657.82 frames. ], batch size: 134, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:19:12,738 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:33,983 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:48,950 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:20:13,199 INFO [train.py:904] (5/8) Epoch 9, batch 2100, loss[loss=0.2317, simple_loss=0.2947, pruned_loss=0.08437, over 16863.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2759, pruned_loss=0.05662, over 3324975.60 frames. ], batch size: 109, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:20:35,024 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 2.645e+02 3.321e+02 4.089e+02 1.048e+03, threshold=6.642e+02, percent-clipped=7.0 2023-04-29 00:20:37,113 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:11,839 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:20,740 INFO [train.py:904] (5/8) Epoch 9, batch 2150, loss[loss=0.2137, simple_loss=0.2867, pruned_loss=0.07038, over 16712.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2771, pruned_loss=0.05736, over 3308833.01 frames. ], batch size: 89, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:22:26,666 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0177, 2.5506, 2.6337, 1.8556, 2.7971, 2.7758, 2.3478, 2.3886], device='cuda:5'), covar=tensor([0.0676, 0.0188, 0.0201, 0.0866, 0.0087, 0.0175, 0.0431, 0.0397], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0098, 0.0087, 0.0142, 0.0070, 0.0098, 0.0121, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 00:22:31,375 INFO [train.py:904] (5/8) Epoch 9, batch 2200, loss[loss=0.2017, simple_loss=0.2904, pruned_loss=0.0565, over 17245.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.278, pruned_loss=0.05833, over 3306682.02 frames. ], batch size: 52, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:22:54,055 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.797e+02 3.398e+02 4.283e+02 9.169e+02, threshold=6.797e+02, percent-clipped=3.0 2023-04-29 00:23:21,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5345, 4.4326, 4.4137, 4.1695, 4.1003, 4.4751, 4.2198, 4.1702], device='cuda:5'), covar=tensor([0.0537, 0.0438, 0.0196, 0.0214, 0.0802, 0.0360, 0.0433, 0.0588], device='cuda:5'), in_proj_covar=tensor([0.0245, 0.0294, 0.0284, 0.0258, 0.0312, 0.0294, 0.0196, 0.0329], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 00:23:41,056 INFO [train.py:904] (5/8) Epoch 9, batch 2250, loss[loss=0.182, simple_loss=0.2833, pruned_loss=0.04038, over 17146.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2785, pruned_loss=0.05843, over 3303907.28 frames. ], batch size: 49, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:24:49,205 INFO [train.py:904] (5/8) Epoch 9, batch 2300, loss[loss=0.1869, simple_loss=0.2766, pruned_loss=0.04861, over 17123.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2785, pruned_loss=0.05803, over 3309825.53 frames. ], batch size: 49, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:24:58,707 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7984, 3.0899, 2.5671, 4.4758, 3.7615, 4.3527, 1.4023, 3.0874], device='cuda:5'), covar=tensor([0.1175, 0.0517, 0.0974, 0.0130, 0.0218, 0.0309, 0.1354, 0.0658], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0154, 0.0175, 0.0128, 0.0199, 0.0210, 0.0172, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 00:25:12,018 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.674e+02 3.137e+02 4.026e+02 1.366e+03, threshold=6.274e+02, percent-clipped=5.0 2023-04-29 00:25:59,052 INFO [train.py:904] (5/8) Epoch 9, batch 2350, loss[loss=0.1832, simple_loss=0.2753, pruned_loss=0.04552, over 16782.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2783, pruned_loss=0.05818, over 3315593.36 frames. ], batch size: 57, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:59,331 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:26:37,430 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6608, 4.6058, 4.5462, 4.0934, 4.6077, 1.9158, 4.3486, 4.4099], device='cuda:5'), covar=tensor([0.0073, 0.0063, 0.0119, 0.0246, 0.0063, 0.1953, 0.0100, 0.0131], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0110, 0.0160, 0.0153, 0.0129, 0.0174, 0.0144, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:26:47,059 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 00:27:06,421 INFO [train.py:904] (5/8) Epoch 9, batch 2400, loss[loss=0.1683, simple_loss=0.2545, pruned_loss=0.04101, over 17036.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2785, pruned_loss=0.05808, over 3316775.44 frames. ], batch size: 41, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:27:29,675 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.500e+02 2.959e+02 3.579e+02 6.970e+02, threshold=5.919e+02, percent-clipped=2.0 2023-04-29 00:27:31,283 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:27:59,176 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:28:14,439 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8571, 3.7615, 2.5692, 5.0879, 4.3437, 4.5894, 1.8198, 2.9364], device='cuda:5'), covar=tensor([0.1305, 0.0480, 0.1130, 0.0155, 0.0291, 0.0390, 0.1399, 0.0855], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0153, 0.0174, 0.0128, 0.0198, 0.0210, 0.0172, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 00:28:14,969 INFO [train.py:904] (5/8) Epoch 9, batch 2450, loss[loss=0.2088, simple_loss=0.2905, pruned_loss=0.06358, over 17218.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2787, pruned_loss=0.05726, over 3323519.19 frames. ], batch size: 43, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:28:35,100 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83667.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:29:22,522 INFO [train.py:904] (5/8) Epoch 9, batch 2500, loss[loss=0.1893, simple_loss=0.2846, pruned_loss=0.04701, over 17269.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2791, pruned_loss=0.05734, over 3313610.42 frames. ], batch size: 52, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:29:23,071 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8411, 2.1053, 2.2632, 4.6517, 2.0796, 2.8319, 2.2515, 2.4421], device='cuda:5'), covar=tensor([0.0771, 0.3208, 0.2002, 0.0298, 0.3549, 0.2023, 0.2673, 0.3166], device='cuda:5'), in_proj_covar=tensor([0.0353, 0.0374, 0.0316, 0.0323, 0.0398, 0.0424, 0.0334, 0.0442], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:29:44,592 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.505e+02 2.999e+02 3.639e+02 7.354e+02, threshold=5.999e+02, percent-clipped=3.0 2023-04-29 00:30:28,785 INFO [train.py:904] (5/8) Epoch 9, batch 2550, loss[loss=0.1683, simple_loss=0.2545, pruned_loss=0.04107, over 17221.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.279, pruned_loss=0.05671, over 3318536.81 frames. ], batch size: 45, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:30:41,843 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:30:43,275 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0658, 2.5544, 2.5761, 4.9110, 2.3173, 3.1927, 2.5814, 2.8112], device='cuda:5'), covar=tensor([0.0681, 0.2881, 0.1788, 0.0279, 0.3332, 0.1728, 0.2443, 0.2719], device='cuda:5'), in_proj_covar=tensor([0.0354, 0.0375, 0.0316, 0.0324, 0.0398, 0.0425, 0.0334, 0.0442], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:31:05,940 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2876, 5.7213, 5.8746, 5.6295, 5.7553, 6.2220, 5.8363, 5.5557], device='cuda:5'), covar=tensor([0.0685, 0.1543, 0.1437, 0.1735, 0.2239, 0.0880, 0.1055, 0.1991], device='cuda:5'), in_proj_covar=tensor([0.0336, 0.0473, 0.0504, 0.0416, 0.0551, 0.0528, 0.0399, 0.0556], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 00:31:20,068 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4677, 5.8145, 5.5859, 5.6807, 5.2252, 5.1067, 5.3381, 5.9758], device='cuda:5'), covar=tensor([0.0913, 0.0836, 0.0901, 0.0553, 0.0711, 0.0597, 0.0807, 0.0735], device='cuda:5'), in_proj_covar=tensor([0.0509, 0.0648, 0.0543, 0.0440, 0.0400, 0.0410, 0.0532, 0.0478], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:31:38,618 INFO [train.py:904] (5/8) Epoch 9, batch 2600, loss[loss=0.1992, simple_loss=0.279, pruned_loss=0.05973, over 16517.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2792, pruned_loss=0.0569, over 3317328.14 frames. ], batch size: 68, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:31:59,382 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.494e+02 3.064e+02 3.833e+02 6.863e+02, threshold=6.129e+02, percent-clipped=2.0 2023-04-29 00:32:04,218 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:32:40,652 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9073, 4.0363, 4.2787, 2.9813, 3.7109, 4.2625, 3.9027, 2.4746], device='cuda:5'), covar=tensor([0.0299, 0.0060, 0.0025, 0.0249, 0.0072, 0.0055, 0.0044, 0.0305], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0069, 0.0068, 0.0125, 0.0075, 0.0084, 0.0075, 0.0118], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 00:32:45,554 INFO [train.py:904] (5/8) Epoch 9, batch 2650, loss[loss=0.1734, simple_loss=0.2685, pruned_loss=0.03913, over 17158.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2795, pruned_loss=0.05711, over 3314302.70 frames. ], batch size: 47, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:32:45,859 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:32:54,757 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7289, 4.6800, 5.2545, 5.2078, 5.2310, 4.7475, 4.7775, 4.4830], device='cuda:5'), covar=tensor([0.0257, 0.0422, 0.0286, 0.0325, 0.0405, 0.0343, 0.0909, 0.0428], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0328, 0.0327, 0.0314, 0.0373, 0.0349, 0.0455, 0.0276], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 00:33:29,449 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7169, 2.2147, 2.2063, 4.4481, 1.9905, 2.7511, 2.2968, 2.4598], device='cuda:5'), covar=tensor([0.0788, 0.3066, 0.1909, 0.0355, 0.3400, 0.1956, 0.2690, 0.2933], device='cuda:5'), in_proj_covar=tensor([0.0354, 0.0375, 0.0315, 0.0324, 0.0397, 0.0426, 0.0333, 0.0443], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:33:51,932 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83900.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:33:53,987 INFO [train.py:904] (5/8) Epoch 9, batch 2700, loss[loss=0.2144, simple_loss=0.286, pruned_loss=0.07138, over 16891.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2794, pruned_loss=0.05649, over 3316264.48 frames. ], batch size: 96, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:34:17,451 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.501e+02 2.856e+02 3.411e+02 7.585e+02, threshold=5.711e+02, percent-clipped=1.0 2023-04-29 00:34:34,764 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 00:34:47,055 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:34:58,137 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:02,470 INFO [train.py:904] (5/8) Epoch 9, batch 2750, loss[loss=0.188, simple_loss=0.2763, pruned_loss=0.04985, over 17112.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2787, pruned_loss=0.05522, over 3323298.13 frames. ], batch size: 47, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:35:48,430 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:51,692 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:56,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3619, 1.9762, 2.1831, 4.1403, 1.9652, 2.5368, 2.1106, 2.2344], device='cuda:5'), covar=tensor([0.0968, 0.3108, 0.1872, 0.0347, 0.3243, 0.1995, 0.2894, 0.2528], device='cuda:5'), in_proj_covar=tensor([0.0356, 0.0377, 0.0317, 0.0326, 0.0401, 0.0428, 0.0335, 0.0446], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:36:15,277 INFO [train.py:904] (5/8) Epoch 9, batch 2800, loss[loss=0.1865, simple_loss=0.2664, pruned_loss=0.05327, over 16281.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2783, pruned_loss=0.05465, over 3325152.73 frames. ], batch size: 165, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:36:20,967 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0440, 3.3849, 3.1175, 5.1360, 4.3638, 4.6712, 1.9551, 3.5819], device='cuda:5'), covar=tensor([0.1070, 0.0542, 0.0886, 0.0125, 0.0278, 0.0341, 0.1176, 0.0626], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0153, 0.0174, 0.0129, 0.0199, 0.0210, 0.0172, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 00:36:25,637 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:36:36,299 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.751e+02 3.290e+02 3.928e+02 7.223e+02, threshold=6.581e+02, percent-clipped=1.0 2023-04-29 00:36:38,770 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3991, 3.3895, 3.8522, 2.6909, 3.4864, 3.7686, 3.6165, 2.1583], device='cuda:5'), covar=tensor([0.0351, 0.0127, 0.0028, 0.0242, 0.0062, 0.0079, 0.0053, 0.0314], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0069, 0.0067, 0.0124, 0.0075, 0.0084, 0.0075, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 00:37:16,074 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:37:23,287 INFO [train.py:904] (5/8) Epoch 9, batch 2850, loss[loss=0.175, simple_loss=0.2622, pruned_loss=0.04392, over 17020.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2778, pruned_loss=0.0548, over 3319486.94 frames. ], batch size: 50, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:38:26,005 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:38:32,592 INFO [train.py:904] (5/8) Epoch 9, batch 2900, loss[loss=0.2371, simple_loss=0.2939, pruned_loss=0.09014, over 15569.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2765, pruned_loss=0.05528, over 3323043.78 frames. ], batch size: 191, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:38:47,859 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 00:38:52,443 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:38:54,445 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.515e+02 3.057e+02 3.611e+02 6.139e+02, threshold=6.114e+02, percent-clipped=0.0 2023-04-29 00:39:23,226 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8458, 5.1223, 4.8922, 4.8638, 4.6352, 4.5598, 4.6335, 5.2461], device='cuda:5'), covar=tensor([0.0936, 0.0933, 0.0915, 0.0684, 0.0761, 0.0927, 0.0892, 0.0786], device='cuda:5'), in_proj_covar=tensor([0.0513, 0.0654, 0.0546, 0.0444, 0.0404, 0.0416, 0.0537, 0.0483], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:39:27,465 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3342, 3.8401, 3.7148, 2.1287, 2.8581, 2.2747, 3.6818, 3.8811], device='cuda:5'), covar=tensor([0.0291, 0.0628, 0.0541, 0.1674, 0.0794, 0.1024, 0.0630, 0.0882], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0143, 0.0157, 0.0141, 0.0134, 0.0124, 0.0136, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:39:30,628 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-29 00:39:43,190 INFO [train.py:904] (5/8) Epoch 9, batch 2950, loss[loss=0.1735, simple_loss=0.2547, pruned_loss=0.04613, over 15766.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2767, pruned_loss=0.05702, over 3321974.33 frames. ], batch size: 35, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:39:51,201 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:39:54,467 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-29 00:39:56,982 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1880, 4.8631, 5.1470, 5.3604, 5.5441, 4.8032, 5.4421, 5.4664], device='cuda:5'), covar=tensor([0.1292, 0.1081, 0.1412, 0.0622, 0.0482, 0.0623, 0.0465, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0535, 0.0654, 0.0817, 0.0667, 0.0507, 0.0507, 0.0522, 0.0581], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:40:47,995 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:40:52,814 INFO [train.py:904] (5/8) Epoch 9, batch 3000, loss[loss=0.1979, simple_loss=0.291, pruned_loss=0.05241, over 17203.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2771, pruned_loss=0.05795, over 3313933.84 frames. ], batch size: 52, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:40:52,814 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 00:41:02,063 INFO [train.py:938] (5/8) Epoch 9, validation: loss=0.1444, simple_loss=0.2507, pruned_loss=0.019, over 944034.00 frames. 2023-04-29 00:41:02,064 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 00:41:23,149 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.086e+02 2.665e+02 3.010e+02 4.048e+02 8.692e+02, threshold=6.019e+02, percent-clipped=1.0 2023-04-29 00:41:27,740 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6817, 3.1340, 2.6165, 4.7417, 3.9473, 4.5226, 1.5329, 3.1939], device='cuda:5'), covar=tensor([0.1318, 0.0571, 0.1072, 0.0164, 0.0303, 0.0353, 0.1440, 0.0701], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0153, 0.0175, 0.0129, 0.0200, 0.0210, 0.0172, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 00:41:47,963 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:09,947 INFO [train.py:904] (5/8) Epoch 9, batch 3050, loss[loss=0.1594, simple_loss=0.2418, pruned_loss=0.03847, over 16972.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2771, pruned_loss=0.0579, over 3321395.13 frames. ], batch size: 41, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:42:20,929 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:35,830 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5235, 3.6956, 2.0803, 3.8606, 2.7168, 3.8806, 2.0125, 2.8593], device='cuda:5'), covar=tensor([0.0197, 0.0296, 0.1320, 0.0185, 0.0650, 0.0476, 0.1218, 0.0547], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0165, 0.0185, 0.0121, 0.0165, 0.0208, 0.0192, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:42:51,682 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4420, 5.4000, 5.2282, 4.6697, 5.2247, 2.1905, 4.9932, 5.3038], device='cuda:5'), covar=tensor([0.0054, 0.0057, 0.0126, 0.0313, 0.0069, 0.1895, 0.0087, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0114, 0.0165, 0.0158, 0.0133, 0.0178, 0.0149, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:43:11,776 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:43:17,186 INFO [train.py:904] (5/8) Epoch 9, batch 3100, loss[loss=0.2074, simple_loss=0.278, pruned_loss=0.06841, over 16810.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2765, pruned_loss=0.05778, over 3326732.55 frames. ], batch size: 102, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:43:22,115 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:43:39,992 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.846e+02 2.519e+02 3.109e+02 3.888e+02 1.112e+03, threshold=6.217e+02, percent-clipped=5.0 2023-04-29 00:43:44,776 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5496, 3.6589, 3.8191, 1.9478, 3.9472, 3.9680, 3.1376, 3.0005], device='cuda:5'), covar=tensor([0.0693, 0.0137, 0.0154, 0.1089, 0.0060, 0.0108, 0.0372, 0.0368], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0097, 0.0089, 0.0142, 0.0071, 0.0101, 0.0122, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 00:44:06,178 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:12,271 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:23,031 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6420, 4.2964, 4.2050, 2.2931, 3.2323, 2.9816, 4.0361, 4.2940], device='cuda:5'), covar=tensor([0.0286, 0.0566, 0.0452, 0.1588, 0.0758, 0.0755, 0.0555, 0.0820], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0143, 0.0157, 0.0140, 0.0134, 0.0124, 0.0136, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:44:28,378 INFO [train.py:904] (5/8) Epoch 9, batch 3150, loss[loss=0.1934, simple_loss=0.284, pruned_loss=0.05143, over 17061.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2757, pruned_loss=0.05711, over 3319672.44 frames. ], batch size: 55, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:45:30,249 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:36,495 INFO [train.py:904] (5/8) Epoch 9, batch 3200, loss[loss=0.1768, simple_loss=0.2515, pruned_loss=0.05107, over 16536.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2744, pruned_loss=0.05645, over 3322383.31 frames. ], batch size: 75, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:45:55,669 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-29 00:45:56,389 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:56,568 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2183, 3.5190, 3.2624, 2.0584, 2.8471, 2.3476, 3.5988, 3.6186], device='cuda:5'), covar=tensor([0.0217, 0.0660, 0.0573, 0.1680, 0.0724, 0.0986, 0.0515, 0.0731], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0144, 0.0157, 0.0141, 0.0135, 0.0125, 0.0136, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:45:59,093 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.601e+02 3.159e+02 4.189e+02 1.012e+03, threshold=6.318e+02, percent-clipped=3.0 2023-04-29 00:46:45,891 INFO [train.py:904] (5/8) Epoch 9, batch 3250, loss[loss=0.1996, simple_loss=0.2775, pruned_loss=0.0609, over 16811.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2748, pruned_loss=0.05687, over 3324777.57 frames. ], batch size: 102, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:46:47,472 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:03,266 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:38,049 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-04-29 00:47:44,424 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-29 00:47:53,019 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-04-29 00:47:55,758 INFO [train.py:904] (5/8) Epoch 9, batch 3300, loss[loss=0.1691, simple_loss=0.2608, pruned_loss=0.03868, over 17141.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2755, pruned_loss=0.05696, over 3322267.45 frames. ], batch size: 47, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:48:18,091 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.660e+02 3.024e+02 3.740e+02 8.260e+02, threshold=6.049e+02, percent-clipped=1.0 2023-04-29 00:48:27,943 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:48:30,821 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8136, 4.7987, 5.3126, 5.3560, 5.3301, 4.9836, 4.9267, 4.6739], device='cuda:5'), covar=tensor([0.0259, 0.0436, 0.0401, 0.0356, 0.0338, 0.0277, 0.0769, 0.0347], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0325, 0.0328, 0.0314, 0.0370, 0.0345, 0.0452, 0.0275], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 00:49:04,556 INFO [train.py:904] (5/8) Epoch 9, batch 3350, loss[loss=0.1556, simple_loss=0.2344, pruned_loss=0.03836, over 17050.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.275, pruned_loss=0.05637, over 3321507.02 frames. ], batch size: 41, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:49:10,242 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:46,812 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.41 vs. limit=5.0 2023-04-29 00:49:53,298 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:01,527 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:16,103 INFO [train.py:904] (5/8) Epoch 9, batch 3400, loss[loss=0.1693, simple_loss=0.2574, pruned_loss=0.04055, over 17144.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2747, pruned_loss=0.05581, over 3323628.31 frames. ], batch size: 47, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:50:21,026 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:38,678 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.352e+02 2.993e+02 3.807e+02 8.102e+02, threshold=5.985e+02, percent-clipped=2.0 2023-04-29 00:50:59,462 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:10,739 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:11,093 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 00:51:24,593 INFO [train.py:904] (5/8) Epoch 9, batch 3450, loss[loss=0.1662, simple_loss=0.2491, pruned_loss=0.04168, over 17034.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2728, pruned_loss=0.05496, over 3328850.37 frames. ], batch size: 41, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:51:26,679 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:57,043 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7353, 4.2415, 4.2826, 1.7995, 4.5378, 4.7663, 3.3490, 3.5432], device='cuda:5'), covar=tensor([0.0935, 0.0122, 0.0178, 0.1321, 0.0093, 0.0069, 0.0351, 0.0422], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0096, 0.0088, 0.0140, 0.0070, 0.0101, 0.0121, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 00:52:16,067 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:20,954 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:24,204 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:35,104 INFO [train.py:904] (5/8) Epoch 9, batch 3500, loss[loss=0.2082, simple_loss=0.2936, pruned_loss=0.06137, over 17079.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2714, pruned_loss=0.0543, over 3336174.76 frames. ], batch size: 53, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:52:37,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7759, 4.7160, 4.6115, 4.3571, 4.2787, 4.7340, 4.6033, 4.4398], device='cuda:5'), covar=tensor([0.0518, 0.0520, 0.0279, 0.0263, 0.0974, 0.0380, 0.0391, 0.0609], device='cuda:5'), in_proj_covar=tensor([0.0250, 0.0306, 0.0296, 0.0267, 0.0323, 0.0305, 0.0202, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 00:52:56,980 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.287e+02 2.792e+02 3.603e+02 8.093e+02, threshold=5.583e+02, percent-clipped=2.0 2023-04-29 00:53:44,901 INFO [train.py:904] (5/8) Epoch 9, batch 3550, loss[loss=0.1663, simple_loss=0.2549, pruned_loss=0.03891, over 16815.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2709, pruned_loss=0.05425, over 3326312.43 frames. ], batch size: 42, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:53:47,111 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:30,628 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6461, 2.2925, 1.7524, 2.0881, 2.7158, 2.5395, 2.9263, 2.8869], device='cuda:5'), covar=tensor([0.0111, 0.0266, 0.0362, 0.0318, 0.0151, 0.0240, 0.0127, 0.0146], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0194, 0.0191, 0.0191, 0.0191, 0.0197, 0.0201, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 00:54:52,985 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:54,906 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:55,846 INFO [train.py:904] (5/8) Epoch 9, batch 3600, loss[loss=0.1865, simple_loss=0.271, pruned_loss=0.05097, over 17208.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2707, pruned_loss=0.05443, over 3312652.69 frames. ], batch size: 44, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:55:17,926 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.544e+02 2.999e+02 3.772e+02 8.043e+02, threshold=5.998e+02, percent-clipped=5.0 2023-04-29 00:55:48,328 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5999, 2.7802, 2.4244, 3.7798, 3.2385, 3.9992, 1.5518, 2.7311], device='cuda:5'), covar=tensor([0.1378, 0.0563, 0.1107, 0.0158, 0.0190, 0.0330, 0.1470, 0.0812], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0155, 0.0178, 0.0131, 0.0204, 0.0215, 0.0176, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:56:07,151 INFO [train.py:904] (5/8) Epoch 9, batch 3650, loss[loss=0.1875, simple_loss=0.2531, pruned_loss=0.06093, over 16880.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2694, pruned_loss=0.05451, over 3314054.86 frames. ], batch size: 109, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:56:12,816 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:21,093 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:50,622 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:07,336 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84892.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:21,453 INFO [train.py:904] (5/8) Epoch 9, batch 3700, loss[loss=0.1769, simple_loss=0.2516, pruned_loss=0.05114, over 16831.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2683, pruned_loss=0.05612, over 3302956.59 frames. ], batch size: 124, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:57:23,534 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:40,824 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9437, 5.4285, 5.5740, 5.3818, 5.3702, 5.9454, 5.5075, 5.2562], device='cuda:5'), covar=tensor([0.0796, 0.1501, 0.1228, 0.1486, 0.2051, 0.0779, 0.0980, 0.1902], device='cuda:5'), in_proj_covar=tensor([0.0335, 0.0467, 0.0495, 0.0414, 0.0544, 0.0515, 0.0395, 0.0554], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 00:57:46,411 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.579e+02 2.947e+02 3.773e+02 6.988e+02, threshold=5.894e+02, percent-clipped=1.0 2023-04-29 00:58:19,042 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:28,279 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8352, 4.1083, 4.1309, 2.2581, 3.2379, 2.9434, 4.3077, 4.1254], device='cuda:5'), covar=tensor([0.0170, 0.0483, 0.0447, 0.1598, 0.0727, 0.0688, 0.0395, 0.0778], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0143, 0.0155, 0.0140, 0.0134, 0.0123, 0.0134, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 00:58:35,038 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:35,663 INFO [train.py:904] (5/8) Epoch 9, batch 3750, loss[loss=0.2018, simple_loss=0.2703, pruned_loss=0.06663, over 16406.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2696, pruned_loss=0.05775, over 3286840.34 frames. ], batch size: 75, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 00:58:41,008 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-29 00:58:50,419 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 00:59:28,400 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:34,140 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:47,559 INFO [train.py:904] (5/8) Epoch 9, batch 3800, loss[loss=0.2029, simple_loss=0.2828, pruned_loss=0.06156, over 17028.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2707, pruned_loss=0.05942, over 3273519.77 frames. ], batch size: 55, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:00:04,142 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:00:11,020 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.825e+02 2.585e+02 2.931e+02 3.484e+02 5.921e+02, threshold=5.862e+02, percent-clipped=1.0 2023-04-29 01:00:43,937 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:01:00,355 INFO [train.py:904] (5/8) Epoch 9, batch 3850, loss[loss=0.1922, simple_loss=0.2646, pruned_loss=0.05994, over 16462.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2708, pruned_loss=0.06026, over 3268893.90 frames. ], batch size: 75, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:02:02,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6222, 2.3358, 1.7664, 2.1015, 2.7819, 2.5966, 2.9747, 2.9452], device='cuda:5'), covar=tensor([0.0105, 0.0240, 0.0349, 0.0296, 0.0123, 0.0197, 0.0136, 0.0135], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0189, 0.0186, 0.0185, 0.0185, 0.0190, 0.0194, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:02:13,001 INFO [train.py:904] (5/8) Epoch 9, batch 3900, loss[loss=0.1969, simple_loss=0.2695, pruned_loss=0.06219, over 16441.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2705, pruned_loss=0.06061, over 3274935.60 frames. ], batch size: 146, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:02:21,761 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:32,584 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:36,237 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.478e+02 2.844e+02 3.462e+02 6.399e+02, threshold=5.687e+02, percent-clipped=3.0 2023-04-29 01:03:09,312 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 01:03:24,368 INFO [train.py:904] (5/8) Epoch 9, batch 3950, loss[loss=0.2038, simple_loss=0.2692, pruned_loss=0.06918, over 16890.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2704, pruned_loss=0.06158, over 3272078.18 frames. ], batch size: 109, lr: 7.66e-03, grad_scale: 8.0 2023-04-29 01:03:28,805 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4422, 2.3759, 1.8203, 2.0930, 2.7362, 2.5557, 2.8134, 2.9320], device='cuda:5'), covar=tensor([0.0104, 0.0205, 0.0318, 0.0264, 0.0114, 0.0172, 0.0129, 0.0119], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0191, 0.0188, 0.0187, 0.0188, 0.0191, 0.0196, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:03:32,290 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1195, 5.4373, 5.1173, 5.2264, 4.8033, 4.6937, 4.8517, 5.4868], device='cuda:5'), covar=tensor([0.0987, 0.0815, 0.1029, 0.0626, 0.0843, 0.0895, 0.0856, 0.0845], device='cuda:5'), in_proj_covar=tensor([0.0516, 0.0658, 0.0540, 0.0448, 0.0410, 0.0418, 0.0544, 0.0488], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:03:32,294 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:03:50,083 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:04:00,693 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:07,273 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:37,100 INFO [train.py:904] (5/8) Epoch 9, batch 4000, loss[loss=0.1782, simple_loss=0.2625, pruned_loss=0.04699, over 16679.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2691, pruned_loss=0.06096, over 3283368.46 frames. ], batch size: 89, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:05:00,774 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.490e+02 3.071e+02 3.606e+02 5.108e+02, threshold=6.141e+02, percent-clipped=0.0 2023-04-29 01:05:17,723 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:05:51,196 INFO [train.py:904] (5/8) Epoch 9, batch 4050, loss[loss=0.1782, simple_loss=0.2622, pruned_loss=0.04713, over 16306.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2697, pruned_loss=0.05968, over 3274973.47 frames. ], batch size: 165, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:06:40,947 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7932, 3.2741, 3.2003, 1.8465, 2.7314, 2.1487, 3.2975, 3.3005], device='cuda:5'), covar=tensor([0.0268, 0.0532, 0.0546, 0.1719, 0.0767, 0.0890, 0.0623, 0.0750], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0145, 0.0158, 0.0143, 0.0137, 0.0126, 0.0137, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:06:45,650 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:06:59,709 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-29 01:07:04,646 INFO [train.py:904] (5/8) Epoch 9, batch 4100, loss[loss=0.2251, simple_loss=0.3084, pruned_loss=0.07089, over 16844.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2712, pruned_loss=0.0589, over 3281398.26 frames. ], batch size: 116, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:07:06,672 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0371, 3.4228, 3.2137, 1.9237, 2.6996, 2.0956, 3.4470, 3.4639], device='cuda:5'), covar=tensor([0.0225, 0.0628, 0.0590, 0.1816, 0.0862, 0.0982, 0.0659, 0.0951], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0146, 0.0159, 0.0144, 0.0138, 0.0127, 0.0138, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:07:09,909 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2480, 5.5697, 5.2553, 5.3710, 4.8997, 4.7701, 4.9417, 5.6518], device='cuda:5'), covar=tensor([0.0892, 0.0745, 0.0898, 0.0598, 0.0769, 0.0684, 0.0828, 0.0753], device='cuda:5'), in_proj_covar=tensor([0.0506, 0.0644, 0.0533, 0.0438, 0.0401, 0.0410, 0.0535, 0.0479], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:07:12,436 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:28,053 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.051e+02 2.369e+02 2.790e+02 6.834e+02, threshold=4.737e+02, percent-clipped=1.0 2023-04-29 01:07:57,985 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:59,369 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0064, 1.8746, 2.4614, 2.9661, 2.8559, 3.4893, 1.9531, 3.2391], device='cuda:5'), covar=tensor([0.0132, 0.0311, 0.0184, 0.0167, 0.0160, 0.0087, 0.0301, 0.0066], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0165, 0.0152, 0.0153, 0.0158, 0.0117, 0.0164, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 01:08:20,249 INFO [train.py:904] (5/8) Epoch 9, batch 4150, loss[loss=0.2354, simple_loss=0.3155, pruned_loss=0.07764, over 17028.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2789, pruned_loss=0.06202, over 3260371.96 frames. ], batch size: 41, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:08:38,533 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3339, 3.2850, 3.3239, 3.4479, 3.4843, 3.2418, 3.4589, 3.5315], device='cuda:5'), covar=tensor([0.0901, 0.0680, 0.0950, 0.0542, 0.0532, 0.2575, 0.0901, 0.0568], device='cuda:5'), in_proj_covar=tensor([0.0505, 0.0612, 0.0764, 0.0624, 0.0473, 0.0478, 0.0490, 0.0546], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:08:57,311 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7249, 1.7702, 1.5333, 1.5216, 1.8242, 1.6527, 1.7865, 1.9674], device='cuda:5'), covar=tensor([0.0084, 0.0175, 0.0244, 0.0217, 0.0120, 0.0166, 0.0110, 0.0139], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0192, 0.0189, 0.0188, 0.0189, 0.0192, 0.0196, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:09:37,281 INFO [train.py:904] (5/8) Epoch 9, batch 4200, loss[loss=0.2194, simple_loss=0.3023, pruned_loss=0.06822, over 16679.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.286, pruned_loss=0.06361, over 3243494.97 frames. ], batch size: 134, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:02,497 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.639e+02 3.175e+02 3.973e+02 9.081e+02, threshold=6.349e+02, percent-clipped=14.0 2023-04-29 01:10:17,806 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:10:26,711 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0109, 2.4419, 2.3066, 2.9445, 2.3457, 3.2823, 1.6711, 2.7058], device='cuda:5'), covar=tensor([0.0964, 0.0493, 0.0887, 0.0123, 0.0138, 0.0353, 0.1279, 0.0603], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0154, 0.0175, 0.0128, 0.0202, 0.0208, 0.0175, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 01:10:38,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8450, 2.9289, 2.7976, 5.0036, 3.8084, 4.4094, 1.5646, 3.3299], device='cuda:5'), covar=tensor([0.1083, 0.0620, 0.0936, 0.0078, 0.0261, 0.0356, 0.1418, 0.0668], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0154, 0.0175, 0.0128, 0.0202, 0.0208, 0.0175, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 01:10:52,280 INFO [train.py:904] (5/8) Epoch 9, batch 4250, loss[loss=0.1972, simple_loss=0.2955, pruned_loss=0.04947, over 16718.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2898, pruned_loss=0.06454, over 3195946.49 frames. ], batch size: 76, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:59,212 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:07,289 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9704, 4.7350, 5.0075, 5.1663, 5.3214, 4.7376, 5.3128, 5.3259], device='cuda:5'), covar=tensor([0.1218, 0.0874, 0.1172, 0.0502, 0.0436, 0.0655, 0.0429, 0.0401], device='cuda:5'), in_proj_covar=tensor([0.0493, 0.0598, 0.0742, 0.0608, 0.0463, 0.0468, 0.0479, 0.0533], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:11:08,362 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:11:20,279 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:38,976 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:48,388 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:12:07,380 INFO [train.py:904] (5/8) Epoch 9, batch 4300, loss[loss=0.2249, simple_loss=0.3067, pruned_loss=0.07157, over 16618.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2913, pruned_loss=0.06356, over 3185282.38 frames. ], batch size: 62, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:12:11,406 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:12:24,467 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6255, 3.9804, 4.0933, 2.1202, 3.3186, 2.3614, 4.0101, 3.8817], device='cuda:5'), covar=tensor([0.0210, 0.0536, 0.0436, 0.1795, 0.0721, 0.0890, 0.0506, 0.0815], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0142, 0.0157, 0.0142, 0.0136, 0.0125, 0.0136, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:12:30,793 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 2.412e+02 2.949e+02 3.511e+02 9.601e+02, threshold=5.898e+02, percent-clipped=2.0 2023-04-29 01:12:49,301 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:12:59,947 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4636, 3.7492, 3.9008, 1.8583, 4.2447, 4.2633, 3.0115, 3.1926], device='cuda:5'), covar=tensor([0.0828, 0.0177, 0.0203, 0.1164, 0.0041, 0.0061, 0.0413, 0.0379], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0095, 0.0084, 0.0136, 0.0067, 0.0095, 0.0118, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 01:13:07,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4502, 4.4301, 4.2560, 3.7226, 4.3449, 1.6085, 4.1472, 4.0325], device='cuda:5'), covar=tensor([0.0057, 0.0052, 0.0105, 0.0260, 0.0064, 0.2343, 0.0082, 0.0155], device='cuda:5'), in_proj_covar=tensor([0.0121, 0.0108, 0.0157, 0.0149, 0.0126, 0.0170, 0.0143, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:13:10,402 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:13:21,370 INFO [train.py:904] (5/8) Epoch 9, batch 4350, loss[loss=0.2234, simple_loss=0.3006, pruned_loss=0.07309, over 11934.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2948, pruned_loss=0.0648, over 3181893.25 frames. ], batch size: 246, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:14:19,053 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:34,746 INFO [train.py:904] (5/8) Epoch 9, batch 4400, loss[loss=0.2082, simple_loss=0.2986, pruned_loss=0.05896, over 17145.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2969, pruned_loss=0.06581, over 3171378.53 frames. ], batch size: 49, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:14:41,691 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:56,972 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.712e+02 3.095e+02 3.599e+02 6.298e+02, threshold=6.190e+02, percent-clipped=2.0 2023-04-29 01:15:06,884 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:15:40,191 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7004, 5.0501, 5.2359, 5.0869, 5.1395, 5.7097, 5.1884, 4.7991], device='cuda:5'), covar=tensor([0.0868, 0.1652, 0.1392, 0.1497, 0.2092, 0.0819, 0.1024, 0.2189], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0458, 0.0485, 0.0401, 0.0532, 0.0513, 0.0391, 0.0545], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 01:15:43,472 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3363, 3.1985, 2.5589, 2.0814, 2.2118, 1.9669, 3.2691, 3.0299], device='cuda:5'), covar=tensor([0.2453, 0.0726, 0.1489, 0.2020, 0.2141, 0.1862, 0.0486, 0.0936], device='cuda:5'), in_proj_covar=tensor([0.0294, 0.0254, 0.0279, 0.0272, 0.0285, 0.0216, 0.0263, 0.0286], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:15:46,021 INFO [train.py:904] (5/8) Epoch 9, batch 4450, loss[loss=0.2223, simple_loss=0.3057, pruned_loss=0.06947, over 17070.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2998, pruned_loss=0.06641, over 3199581.28 frames. ], batch size: 53, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:15:50,461 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:16:17,409 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 01:16:34,319 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:16:56,356 INFO [train.py:904] (5/8) Epoch 9, batch 4500, loss[loss=0.2043, simple_loss=0.2975, pruned_loss=0.05558, over 17210.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2996, pruned_loss=0.06656, over 3213022.08 frames. ], batch size: 44, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:17:08,447 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0723, 2.9893, 3.0892, 1.6309, 3.3265, 3.3504, 2.6821, 2.5878], device='cuda:5'), covar=tensor([0.0788, 0.0201, 0.0177, 0.1160, 0.0060, 0.0107, 0.0391, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0095, 0.0084, 0.0137, 0.0067, 0.0095, 0.0118, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 01:17:20,319 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.131e+02 2.570e+02 3.015e+02 5.229e+02, threshold=5.140e+02, percent-clipped=0.0 2023-04-29 01:18:07,072 INFO [train.py:904] (5/8) Epoch 9, batch 4550, loss[loss=0.2309, simple_loss=0.3012, pruned_loss=0.0803, over 11982.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3005, pruned_loss=0.06741, over 3197982.79 frames. ], batch size: 246, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:18:23,757 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:18:34,920 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:18:54,428 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:20,638 INFO [train.py:904] (5/8) Epoch 9, batch 4600, loss[loss=0.2073, simple_loss=0.2964, pruned_loss=0.05912, over 17197.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3008, pruned_loss=0.06741, over 3200453.69 frames. ], batch size: 46, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:19:32,654 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:42,426 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.222e+02 2.575e+02 3.036e+02 5.036e+02, threshold=5.150e+02, percent-clipped=0.0 2023-04-29 01:19:44,976 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:20:13,016 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:20:30,869 INFO [train.py:904] (5/8) Epoch 9, batch 4650, loss[loss=0.2009, simple_loss=0.2839, pruned_loss=0.05894, over 16703.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2987, pruned_loss=0.06665, over 3208833.58 frames. ], batch size: 89, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:20:36,077 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4004, 3.4452, 2.7236, 2.0106, 2.3294, 2.0864, 3.6557, 3.1442], device='cuda:5'), covar=tensor([0.2654, 0.0731, 0.1497, 0.2103, 0.2235, 0.1874, 0.0451, 0.0934], device='cuda:5'), in_proj_covar=tensor([0.0296, 0.0253, 0.0279, 0.0272, 0.0284, 0.0216, 0.0263, 0.0285], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:21:02,425 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4237, 4.4233, 4.2781, 3.2407, 4.3493, 1.5251, 4.0336, 3.8494], device='cuda:5'), covar=tensor([0.0085, 0.0060, 0.0128, 0.0498, 0.0077, 0.2805, 0.0105, 0.0223], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0106, 0.0154, 0.0146, 0.0123, 0.0167, 0.0139, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:21:04,793 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:20,451 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:42,724 INFO [train.py:904] (5/8) Epoch 9, batch 4700, loss[loss=0.2124, simple_loss=0.2952, pruned_loss=0.0648, over 16402.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2953, pruned_loss=0.06485, over 3212735.95 frames. ], batch size: 146, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:22:06,317 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.169e+02 2.505e+02 2.903e+02 6.033e+02, threshold=5.010e+02, percent-clipped=1.0 2023-04-29 01:22:25,241 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:22:33,582 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:22:55,524 INFO [train.py:904] (5/8) Epoch 9, batch 4750, loss[loss=0.1688, simple_loss=0.254, pruned_loss=0.0418, over 16781.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2909, pruned_loss=0.06254, over 3219513.97 frames. ], batch size: 83, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:23:04,372 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:23:39,923 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:23:55,336 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:24:13,778 INFO [train.py:904] (5/8) Epoch 9, batch 4800, loss[loss=0.2068, simple_loss=0.3003, pruned_loss=0.05663, over 16671.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2876, pruned_loss=0.06041, over 3218571.39 frames. ], batch size: 134, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:24:14,301 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:24:37,223 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.079e+02 2.389e+02 2.964e+02 6.421e+02, threshold=4.777e+02, percent-clipped=3.0 2023-04-29 01:24:38,381 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:28,116 INFO [train.py:904] (5/8) Epoch 9, batch 4850, loss[loss=0.2138, simple_loss=0.2915, pruned_loss=0.068, over 16578.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2891, pruned_loss=0.06032, over 3185440.77 frames. ], batch size: 62, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:25:28,531 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:45,081 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:54,694 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:16,482 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:29,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6611, 2.7397, 1.8701, 2.8037, 2.1823, 2.7987, 2.0505, 2.3857], device='cuda:5'), covar=tensor([0.0203, 0.0320, 0.1195, 0.0118, 0.0631, 0.0464, 0.1092, 0.0563], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0160, 0.0182, 0.0109, 0.0162, 0.0198, 0.0190, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:26:42,276 INFO [train.py:904] (5/8) Epoch 9, batch 4900, loss[loss=0.2044, simple_loss=0.2832, pruned_loss=0.06278, over 16680.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2885, pruned_loss=0.05916, over 3180225.62 frames. ], batch size: 57, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:26:59,820 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86113.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:02,015 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5672, 5.9885, 5.1645, 5.9044, 5.2831, 5.0108, 5.6412, 5.8370], device='cuda:5'), covar=tensor([0.1871, 0.1174, 0.2596, 0.0801, 0.1329, 0.1132, 0.1565, 0.1551], device='cuda:5'), in_proj_covar=tensor([0.0474, 0.0597, 0.0498, 0.0406, 0.0373, 0.0385, 0.0497, 0.0451], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:27:05,787 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.278e+02 2.647e+02 3.027e+02 5.480e+02, threshold=5.294e+02, percent-clipped=1.0 2023-04-29 01:27:23,849 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:26,528 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:37,275 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:55,685 INFO [train.py:904] (5/8) Epoch 9, batch 4950, loss[loss=0.2261, simple_loss=0.3086, pruned_loss=0.07184, over 16397.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2884, pruned_loss=0.05892, over 3173813.48 frames. ], batch size: 68, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:28:45,906 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:28:47,116 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:08,685 INFO [train.py:904] (5/8) Epoch 9, batch 5000, loss[loss=0.1967, simple_loss=0.2836, pruned_loss=0.05493, over 16516.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2898, pruned_loss=0.05849, over 3190972.49 frames. ], batch size: 68, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:29:32,035 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.365e+02 2.931e+02 3.414e+02 6.733e+02, threshold=5.862e+02, percent-clipped=2.0 2023-04-29 01:29:50,445 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:54,585 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:11,881 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86246.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:19,727 INFO [train.py:904] (5/8) Epoch 9, batch 5050, loss[loss=0.2096, simple_loss=0.3017, pruned_loss=0.05873, over 16904.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2905, pruned_loss=0.05828, over 3201308.77 frames. ], batch size: 109, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:31:01,237 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:08,001 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:31:29,532 INFO [train.py:904] (5/8) Epoch 9, batch 5100, loss[loss=0.1978, simple_loss=0.2798, pruned_loss=0.05792, over 17032.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2877, pruned_loss=0.0574, over 3205387.78 frames. ], batch size: 55, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:31:37,421 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:45,608 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86313.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:52,765 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.422e+02 2.762e+02 3.223e+02 5.276e+02, threshold=5.525e+02, percent-clipped=0.0 2023-04-29 01:32:00,412 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7145, 3.8091, 2.0011, 4.4439, 2.7377, 4.2568, 2.1351, 3.0503], device='cuda:5'), covar=tensor([0.0201, 0.0296, 0.1765, 0.0059, 0.0840, 0.0388, 0.1646, 0.0668], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0159, 0.0182, 0.0108, 0.0163, 0.0198, 0.0189, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:32:08,054 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:32:41,557 INFO [train.py:904] (5/8) Epoch 9, batch 5150, loss[loss=0.2266, simple_loss=0.314, pruned_loss=0.06962, over 16319.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2877, pruned_loss=0.05691, over 3196108.53 frames. ], batch size: 165, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:32:50,290 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:33:27,387 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6639, 4.6160, 4.4912, 4.3011, 4.0395, 4.5924, 4.4345, 4.2647], device='cuda:5'), covar=tensor([0.0475, 0.0365, 0.0258, 0.0222, 0.0978, 0.0346, 0.0321, 0.0562], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0272, 0.0266, 0.0239, 0.0288, 0.0274, 0.0181, 0.0305], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:33:36,395 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 01:33:48,189 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7317, 3.7948, 2.1510, 4.3346, 2.6172, 4.1839, 1.9591, 2.8764], device='cuda:5'), covar=tensor([0.0159, 0.0270, 0.1347, 0.0078, 0.0761, 0.0334, 0.1551, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0160, 0.0184, 0.0108, 0.0164, 0.0200, 0.0191, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:33:54,445 INFO [train.py:904] (5/8) Epoch 9, batch 5200, loss[loss=0.2132, simple_loss=0.3014, pruned_loss=0.06247, over 16649.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2871, pruned_loss=0.05703, over 3191990.48 frames. ], batch size: 134, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:34:02,861 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:17,302 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 2.298e+02 2.670e+02 3.115e+02 5.719e+02, threshold=5.340e+02, percent-clipped=1.0 2023-04-29 01:34:27,691 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:42,632 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6860, 2.7609, 1.8172, 2.8850, 2.2379, 2.8673, 1.9343, 2.3906], device='cuda:5'), covar=tensor([0.0213, 0.0296, 0.1166, 0.0117, 0.0617, 0.0387, 0.1174, 0.0552], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0160, 0.0184, 0.0109, 0.0164, 0.0200, 0.0192, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:34:50,136 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:57,891 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:35:06,332 INFO [train.py:904] (5/8) Epoch 9, batch 5250, loss[loss=0.2067, simple_loss=0.2944, pruned_loss=0.05949, over 16316.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.285, pruned_loss=0.05692, over 3193943.67 frames. ], batch size: 165, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:35:55,471 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5430, 3.4969, 3.4430, 2.9313, 3.3436, 2.0495, 3.1816, 2.8593], device='cuda:5'), covar=tensor([0.0112, 0.0105, 0.0103, 0.0236, 0.0073, 0.1960, 0.0126, 0.0175], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0108, 0.0156, 0.0148, 0.0124, 0.0171, 0.0140, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:36:16,370 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3342, 3.2901, 3.3457, 3.5206, 3.5419, 3.2221, 3.5107, 3.5929], device='cuda:5'), covar=tensor([0.0996, 0.0902, 0.1145, 0.0595, 0.0581, 0.2285, 0.0838, 0.0607], device='cuda:5'), in_proj_covar=tensor([0.0496, 0.0604, 0.0750, 0.0615, 0.0466, 0.0467, 0.0483, 0.0536], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:36:17,600 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86501.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:36:19,018 INFO [train.py:904] (5/8) Epoch 9, batch 5300, loss[loss=0.1663, simple_loss=0.2442, pruned_loss=0.04422, over 17210.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2825, pruned_loss=0.05627, over 3176617.12 frames. ], batch size: 43, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:36:25,624 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:36:25,991 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-29 01:36:42,028 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.448e+02 2.767e+02 3.292e+02 5.111e+02, threshold=5.534e+02, percent-clipped=0.0 2023-04-29 01:36:46,553 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3549, 3.7754, 3.4585, 2.0286, 3.0530, 2.5731, 3.6562, 3.7824], device='cuda:5'), covar=tensor([0.0212, 0.0528, 0.0564, 0.1661, 0.0690, 0.0809, 0.0579, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0139, 0.0156, 0.0141, 0.0134, 0.0124, 0.0136, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:37:01,906 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:37:32,907 INFO [train.py:904] (5/8) Epoch 9, batch 5350, loss[loss=0.2104, simple_loss=0.2973, pruned_loss=0.06175, over 16778.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2805, pruned_loss=0.05523, over 3180541.92 frames. ], batch size: 124, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:11,779 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:13,789 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:23,759 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:38:45,868 INFO [train.py:904] (5/8) Epoch 9, batch 5400, loss[loss=0.2002, simple_loss=0.2931, pruned_loss=0.05364, over 16182.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2838, pruned_loss=0.05652, over 3159550.67 frames. ], batch size: 165, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:46,257 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:01,786 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:09,040 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.152e+02 2.423e+02 2.907e+02 4.710e+02, threshold=4.847e+02, percent-clipped=0.0 2023-04-29 01:39:32,123 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:39:36,765 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 01:39:41,860 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:00,486 INFO [train.py:904] (5/8) Epoch 9, batch 5450, loss[loss=0.2334, simple_loss=0.3145, pruned_loss=0.07616, over 16549.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2868, pruned_loss=0.05814, over 3154067.06 frames. ], batch size: 68, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:40:10,941 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:15,979 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86661.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:18,961 INFO [train.py:904] (5/8) Epoch 9, batch 5500, loss[loss=0.2656, simple_loss=0.3307, pruned_loss=0.1003, over 11664.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2946, pruned_loss=0.06371, over 3117324.84 frames. ], batch size: 248, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:41:24,419 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:27,704 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:43,584 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 3.483e+02 4.212e+02 5.366e+02 1.166e+03, threshold=8.424e+02, percent-clipped=36.0 2023-04-29 01:41:54,568 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:42:36,880 INFO [train.py:904] (5/8) Epoch 9, batch 5550, loss[loss=0.2179, simple_loss=0.3049, pruned_loss=0.06542, over 16632.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3023, pruned_loss=0.06945, over 3109095.72 frames. ], batch size: 76, lr: 7.59e-03, grad_scale: 16.0 2023-04-29 01:42:39,283 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6080, 4.5912, 4.4099, 4.2212, 4.0523, 4.5145, 4.4040, 4.2610], device='cuda:5'), covar=tensor([0.0543, 0.0437, 0.0278, 0.0250, 0.0922, 0.0414, 0.0351, 0.0636], device='cuda:5'), in_proj_covar=tensor([0.0231, 0.0279, 0.0271, 0.0246, 0.0294, 0.0283, 0.0184, 0.0314], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:42:43,145 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:03,221 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 01:43:09,699 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:45,367 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:53,952 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:43:54,689 INFO [train.py:904] (5/8) Epoch 9, batch 5600, loss[loss=0.3634, simple_loss=0.3993, pruned_loss=0.1638, over 11163.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3087, pruned_loss=0.07514, over 3068371.35 frames. ], batch size: 247, lr: 7.59e-03, grad_scale: 8.0 2023-04-29 01:44:14,692 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:44:21,511 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.583e+02 3.990e+02 4.962e+02 6.279e+02 1.585e+03, threshold=9.923e+02, percent-clipped=6.0 2023-04-29 01:44:29,694 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-29 01:44:42,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9487, 4.9483, 5.4143, 5.3923, 5.4359, 4.9688, 4.9881, 4.6315], device='cuda:5'), covar=tensor([0.0233, 0.0372, 0.0278, 0.0354, 0.0361, 0.0267, 0.0816, 0.0385], device='cuda:5'), in_proj_covar=tensor([0.0301, 0.0301, 0.0312, 0.0299, 0.0351, 0.0325, 0.0429, 0.0264], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 01:45:17,290 INFO [train.py:904] (5/8) Epoch 9, batch 5650, loss[loss=0.2057, simple_loss=0.2927, pruned_loss=0.05936, over 16762.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3141, pruned_loss=0.07935, over 3061021.38 frames. ], batch size: 89, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:45:53,414 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9277, 4.6743, 4.9237, 5.1481, 5.3077, 4.7245, 5.2707, 5.2568], device='cuda:5'), covar=tensor([0.1517, 0.1133, 0.1400, 0.0524, 0.0452, 0.0681, 0.0470, 0.0530], device='cuda:5'), in_proj_covar=tensor([0.0485, 0.0594, 0.0728, 0.0600, 0.0458, 0.0459, 0.0478, 0.0527], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:45:53,516 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:45:58,527 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:33,129 INFO [train.py:904] (5/8) Epoch 9, batch 5700, loss[loss=0.2343, simple_loss=0.324, pruned_loss=0.07229, over 16413.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3154, pruned_loss=0.08067, over 3055567.44 frames. ], batch size: 68, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:46:33,488 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:59,885 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.388e+02 3.612e+02 4.435e+02 5.746e+02 9.391e+02, threshold=8.870e+02, percent-clipped=0.0 2023-04-29 01:47:25,189 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:31,056 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:48,292 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:50,855 INFO [train.py:904] (5/8) Epoch 9, batch 5750, loss[loss=0.2274, simple_loss=0.308, pruned_loss=0.07339, over 16255.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3182, pruned_loss=0.0825, over 3017064.78 frames. ], batch size: 165, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:48:02,534 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3382, 3.2732, 3.3228, 3.4662, 3.4851, 3.2291, 3.4430, 3.5274], device='cuda:5'), covar=tensor([0.0943, 0.0828, 0.0967, 0.0490, 0.0565, 0.2029, 0.0873, 0.0612], device='cuda:5'), in_proj_covar=tensor([0.0480, 0.0589, 0.0720, 0.0595, 0.0453, 0.0455, 0.0474, 0.0522], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:49:03,360 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3607, 1.8785, 2.0594, 3.9501, 1.8767, 2.3605, 2.0276, 2.0742], device='cuda:5'), covar=tensor([0.0942, 0.3688, 0.2144, 0.0406, 0.4208, 0.2260, 0.3277, 0.3484], device='cuda:5'), in_proj_covar=tensor([0.0350, 0.0373, 0.0311, 0.0318, 0.0402, 0.0422, 0.0333, 0.0441], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:49:12,518 INFO [train.py:904] (5/8) Epoch 9, batch 5800, loss[loss=0.2017, simple_loss=0.2863, pruned_loss=0.05851, over 16484.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3171, pruned_loss=0.08045, over 3041275.75 frames. ], batch size: 68, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:49:13,824 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:49:19,650 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 01:49:40,722 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 3.181e+02 3.841e+02 4.855e+02 7.782e+02, threshold=7.681e+02, percent-clipped=0.0 2023-04-29 01:50:30,440 INFO [train.py:904] (5/8) Epoch 9, batch 5850, loss[loss=0.2159, simple_loss=0.3045, pruned_loss=0.06366, over 16880.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3148, pruned_loss=0.07847, over 3050808.76 frames. ], batch size: 96, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:50:47,481 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0348, 3.0682, 1.7155, 3.2709, 2.2351, 3.2568, 1.9464, 2.4919], device='cuda:5'), covar=tensor([0.0233, 0.0348, 0.1644, 0.0131, 0.0829, 0.0525, 0.1419, 0.0664], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0161, 0.0185, 0.0111, 0.0165, 0.0201, 0.0192, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 01:50:47,506 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:51:42,235 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:51:50,059 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:51:50,881 INFO [train.py:904] (5/8) Epoch 9, batch 5900, loss[loss=0.2726, simple_loss=0.326, pruned_loss=0.1096, over 11775.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3146, pruned_loss=0.0779, over 3072996.55 frames. ], batch size: 248, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:52:22,907 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.063e+02 3.002e+02 3.675e+02 4.500e+02 8.851e+02, threshold=7.350e+02, percent-clipped=1.0 2023-04-29 01:52:33,902 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9861, 2.3849, 2.3301, 2.9068, 2.2737, 3.2520, 1.7232, 2.6694], device='cuda:5'), covar=tensor([0.1110, 0.0468, 0.0872, 0.0139, 0.0148, 0.0342, 0.1206, 0.0635], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0152, 0.0176, 0.0126, 0.0201, 0.0204, 0.0173, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 01:53:01,034 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87144.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:09,082 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:13,005 INFO [train.py:904] (5/8) Epoch 9, batch 5950, loss[loss=0.2468, simple_loss=0.3161, pruned_loss=0.0887, over 16913.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3152, pruned_loss=0.07709, over 3043108.71 frames. ], batch size: 109, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:53:34,389 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1146, 3.3297, 3.5569, 3.5266, 3.5296, 3.2921, 3.3392, 3.3863], device='cuda:5'), covar=tensor([0.0411, 0.0590, 0.0444, 0.0498, 0.0504, 0.0496, 0.0854, 0.0518], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0311, 0.0322, 0.0306, 0.0361, 0.0332, 0.0443, 0.0273], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 01:53:42,084 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:54:07,689 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3889, 3.2506, 2.5362, 2.0577, 2.4365, 2.0711, 3.2786, 3.2078], device='cuda:5'), covar=tensor([0.2746, 0.0751, 0.1682, 0.2015, 0.2164, 0.1885, 0.0583, 0.0941], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0255, 0.0281, 0.0273, 0.0283, 0.0215, 0.0265, 0.0282], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:54:33,108 INFO [train.py:904] (5/8) Epoch 9, batch 6000, loss[loss=0.231, simple_loss=0.3062, pruned_loss=0.07791, over 16652.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3139, pruned_loss=0.07647, over 3056666.71 frames. ], batch size: 57, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:54:33,108 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 01:54:44,302 INFO [train.py:938] (5/8) Epoch 9, validation: loss=0.1674, simple_loss=0.2809, pruned_loss=0.02692, over 944034.00 frames. 2023-04-29 01:54:44,303 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 01:54:57,437 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8049, 2.1950, 1.7662, 2.0124, 2.6019, 2.3187, 2.7008, 2.8057], device='cuda:5'), covar=tensor([0.0069, 0.0264, 0.0349, 0.0292, 0.0158, 0.0247, 0.0160, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0189, 0.0186, 0.0186, 0.0186, 0.0190, 0.0189, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 01:55:11,804 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 3.247e+02 3.795e+02 4.870e+02 1.523e+03, threshold=7.589e+02, percent-clipped=3.0 2023-04-29 01:55:28,614 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87230.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:55:34,801 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:55:37,609 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:56:03,549 INFO [train.py:904] (5/8) Epoch 9, batch 6050, loss[loss=0.2322, simple_loss=0.324, pruned_loss=0.07021, over 16682.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3117, pruned_loss=0.07512, over 3078806.07 frames. ], batch size: 89, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:56:25,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2285, 3.2621, 2.9798, 5.3403, 4.0993, 4.7827, 2.1118, 3.6350], device='cuda:5'), covar=tensor([0.1184, 0.0640, 0.1009, 0.0094, 0.0394, 0.0316, 0.1324, 0.0698], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0153, 0.0177, 0.0126, 0.0203, 0.0205, 0.0174, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 01:56:56,178 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:57:07,057 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:57:11,073 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 01:57:23,359 INFO [train.py:904] (5/8) Epoch 9, batch 6100, loss[loss=0.2083, simple_loss=0.2963, pruned_loss=0.06011, over 16552.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3106, pruned_loss=0.07355, over 3092069.50 frames. ], batch size: 68, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:57:52,503 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.938e+02 3.700e+02 4.608e+02 9.243e+02, threshold=7.400e+02, percent-clipped=2.0 2023-04-29 01:57:55,707 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-29 01:58:42,519 INFO [train.py:904] (5/8) Epoch 9, batch 6150, loss[loss=0.2316, simple_loss=0.3111, pruned_loss=0.07608, over 15241.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3087, pruned_loss=0.07288, over 3102661.88 frames. ], batch size: 190, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:58:48,832 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-29 01:58:52,858 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:58:59,194 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:00:00,636 INFO [train.py:904] (5/8) Epoch 9, batch 6200, loss[loss=0.2269, simple_loss=0.308, pruned_loss=0.0729, over 16826.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3083, pruned_loss=0.07351, over 3096086.94 frames. ], batch size: 116, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 02:00:28,305 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.203e+02 3.670e+02 4.311e+02 5.768e+02 9.493e+02, threshold=8.622e+02, percent-clipped=7.0 2023-04-29 02:00:29,711 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5118, 3.4822, 3.4354, 2.8562, 3.4364, 1.9688, 3.2284, 2.8995], device='cuda:5'), covar=tensor([0.0115, 0.0101, 0.0138, 0.0253, 0.0086, 0.1982, 0.0121, 0.0205], device='cuda:5'), in_proj_covar=tensor([0.0117, 0.0105, 0.0153, 0.0147, 0.0121, 0.0168, 0.0138, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:00:34,895 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:18,833 INFO [train.py:904] (5/8) Epoch 9, batch 6250, loss[loss=0.2289, simple_loss=0.3074, pruned_loss=0.07525, over 15470.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3077, pruned_loss=0.07301, over 3096461.31 frames. ], batch size: 190, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:01:21,200 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0106, 2.8075, 2.6374, 1.9357, 2.5352, 2.2175, 2.7522, 2.9318], device='cuda:5'), covar=tensor([0.0300, 0.0617, 0.0574, 0.1642, 0.0755, 0.0809, 0.0575, 0.0584], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0140, 0.0159, 0.0143, 0.0136, 0.0126, 0.0136, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 02:01:32,680 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:45,569 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:24,269 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1988, 4.0252, 4.0397, 2.7737, 3.6245, 3.9936, 3.7947, 2.1528], device='cuda:5'), covar=tensor([0.0406, 0.0023, 0.0024, 0.0256, 0.0060, 0.0065, 0.0041, 0.0345], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0064, 0.0065, 0.0122, 0.0071, 0.0083, 0.0071, 0.0115], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 02:02:33,532 INFO [train.py:904] (5/8) Epoch 9, batch 6300, loss[loss=0.2391, simple_loss=0.3152, pruned_loss=0.0815, over 16478.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3073, pruned_loss=0.07204, over 3110385.31 frames. ], batch size: 68, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:02:38,925 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:59,938 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87518.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:04,116 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 3.209e+02 3.966e+02 4.812e+02 1.231e+03, threshold=7.932e+02, percent-clipped=2.0 2023-04-29 02:03:06,737 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:24,698 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:51,232 INFO [train.py:904] (5/8) Epoch 9, batch 6350, loss[loss=0.2284, simple_loss=0.309, pruned_loss=0.07393, over 16498.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3076, pruned_loss=0.07273, over 3123987.51 frames. ], batch size: 68, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:04:13,038 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 02:04:37,492 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:40,719 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:44,427 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:50,120 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:08,853 INFO [train.py:904] (5/8) Epoch 9, batch 6400, loss[loss=0.3105, simple_loss=0.3682, pruned_loss=0.1264, over 11236.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3082, pruned_loss=0.07447, over 3106366.48 frames. ], batch size: 248, lr: 7.56e-03, grad_scale: 8.0 2023-04-29 02:05:13,522 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:15,228 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1693, 4.1507, 4.0767, 3.4548, 4.0808, 1.5554, 3.8935, 3.7933], device='cuda:5'), covar=tensor([0.0082, 0.0069, 0.0117, 0.0292, 0.0080, 0.2343, 0.0101, 0.0163], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0104, 0.0151, 0.0144, 0.0121, 0.0165, 0.0136, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:05:37,572 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.591e+02 3.429e+02 4.257e+02 5.158e+02 9.236e+02, threshold=8.515e+02, percent-clipped=3.0 2023-04-29 02:06:15,014 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 02:06:18,038 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4431, 4.3958, 4.3133, 3.7354, 4.3433, 1.5643, 4.1159, 4.0519], device='cuda:5'), covar=tensor([0.0072, 0.0066, 0.0109, 0.0274, 0.0066, 0.2247, 0.0099, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0104, 0.0151, 0.0145, 0.0121, 0.0166, 0.0136, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:06:22,246 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:06:24,605 INFO [train.py:904] (5/8) Epoch 9, batch 6450, loss[loss=0.2363, simple_loss=0.3158, pruned_loss=0.07836, over 17045.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3084, pruned_loss=0.0739, over 3101331.34 frames. ], batch size: 50, lr: 7.55e-03, grad_scale: 4.0 2023-04-29 02:06:34,914 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:06:46,384 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:07:25,263 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-29 02:07:41,898 INFO [train.py:904] (5/8) Epoch 9, batch 6500, loss[loss=0.1986, simple_loss=0.2841, pruned_loss=0.0566, over 16752.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3063, pruned_loss=0.07298, over 3092530.05 frames. ], batch size: 76, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:07:42,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5291, 3.5913, 2.8188, 2.1269, 2.4853, 2.2084, 3.7140, 3.3174], device='cuda:5'), covar=tensor([0.2538, 0.0686, 0.1455, 0.2067, 0.2190, 0.1753, 0.0378, 0.0940], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0253, 0.0280, 0.0272, 0.0282, 0.0214, 0.0263, 0.0280], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:07:48,425 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:08:06,570 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:08:13,256 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.348e+02 3.236e+02 4.021e+02 5.179e+02 1.078e+03, threshold=8.043e+02, percent-clipped=2.0 2023-04-29 02:09:00,853 INFO [train.py:904] (5/8) Epoch 9, batch 6550, loss[loss=0.2216, simple_loss=0.3206, pruned_loss=0.06126, over 16472.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3091, pruned_loss=0.07478, over 3075682.82 frames. ], batch size: 146, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:10:18,529 INFO [train.py:904] (5/8) Epoch 9, batch 6600, loss[loss=0.2285, simple_loss=0.315, pruned_loss=0.07101, over 16668.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3114, pruned_loss=0.0753, over 3088623.45 frames. ], batch size: 62, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:10:42,570 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:10:51,647 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 3.184e+02 4.033e+02 5.145e+02 1.254e+03, threshold=8.065e+02, percent-clipped=5.0 2023-04-29 02:11:32,526 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 02:11:36,982 INFO [train.py:904] (5/8) Epoch 9, batch 6650, loss[loss=0.2202, simple_loss=0.3017, pruned_loss=0.06937, over 16907.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3122, pruned_loss=0.07636, over 3083236.67 frames. ], batch size: 109, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:11:51,507 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:12:29,322 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:12:53,911 INFO [train.py:904] (5/8) Epoch 9, batch 6700, loss[loss=0.2216, simple_loss=0.3065, pruned_loss=0.06834, over 16174.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3107, pruned_loss=0.07628, over 3076028.73 frames. ], batch size: 165, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:13:26,705 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 3.383e+02 4.194e+02 5.361e+02 9.838e+02, threshold=8.388e+02, percent-clipped=4.0 2023-04-29 02:13:44,326 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:13:44,502 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3402, 3.9274, 3.9828, 2.5762, 3.5978, 3.9027, 3.6862, 2.1203], device='cuda:5'), covar=tensor([0.0379, 0.0028, 0.0025, 0.0281, 0.0057, 0.0061, 0.0045, 0.0343], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0063, 0.0063, 0.0120, 0.0070, 0.0082, 0.0071, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 02:13:52,499 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:14:00,296 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:14:10,410 INFO [train.py:904] (5/8) Epoch 9, batch 6750, loss[loss=0.2121, simple_loss=0.2947, pruned_loss=0.06478, over 16811.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3098, pruned_loss=0.07658, over 3076607.75 frames. ], batch size: 96, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:14:15,834 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 02:14:24,857 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:15:29,305 INFO [train.py:904] (5/8) Epoch 9, batch 6800, loss[loss=0.2172, simple_loss=0.3019, pruned_loss=0.06627, over 16894.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3094, pruned_loss=0.07552, over 3103867.85 frames. ], batch size: 116, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:15:54,267 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:16:02,239 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 3.136e+02 3.707e+02 4.804e+02 8.151e+02, threshold=7.414e+02, percent-clipped=0.0 2023-04-29 02:16:06,267 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-29 02:16:45,314 INFO [train.py:904] (5/8) Epoch 9, batch 6850, loss[loss=0.2833, simple_loss=0.3413, pruned_loss=0.1126, over 11676.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3108, pruned_loss=0.07575, over 3104174.08 frames. ], batch size: 246, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:16:56,650 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6914, 2.5303, 2.2522, 3.4385, 2.5249, 3.6950, 1.3387, 2.7225], device='cuda:5'), covar=tensor([0.1359, 0.0609, 0.1147, 0.0121, 0.0184, 0.0366, 0.1585, 0.0779], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0152, 0.0176, 0.0127, 0.0200, 0.0204, 0.0173, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 02:17:06,663 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:17:18,934 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7480, 4.7160, 4.5975, 3.8163, 4.5487, 1.6410, 4.2942, 4.3825], device='cuda:5'), covar=tensor([0.0094, 0.0081, 0.0149, 0.0399, 0.0099, 0.2418, 0.0152, 0.0181], device='cuda:5'), in_proj_covar=tensor([0.0116, 0.0104, 0.0151, 0.0144, 0.0120, 0.0166, 0.0135, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:17:47,803 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8925, 3.1873, 2.8086, 4.9411, 3.7283, 4.6144, 1.6750, 3.2763], device='cuda:5'), covar=tensor([0.1179, 0.0541, 0.0991, 0.0095, 0.0294, 0.0298, 0.1322, 0.0664], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0153, 0.0177, 0.0127, 0.0201, 0.0204, 0.0173, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 02:17:59,110 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:01,010 INFO [train.py:904] (5/8) Epoch 9, batch 6900, loss[loss=0.251, simple_loss=0.3283, pruned_loss=0.0868, over 16141.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3126, pruned_loss=0.07431, over 3124561.10 frames. ], batch size: 165, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:18:24,720 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:33,104 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 3.211e+02 3.844e+02 4.463e+02 9.504e+02, threshold=7.687e+02, percent-clipped=4.0 2023-04-29 02:19:18,082 INFO [train.py:904] (5/8) Epoch 9, batch 6950, loss[loss=0.2352, simple_loss=0.3122, pruned_loss=0.07913, over 16489.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3142, pruned_loss=0.07591, over 3127420.54 frames. ], batch size: 75, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:19:32,170 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:19:32,246 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:19:38,180 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:20:31,304 INFO [train.py:904] (5/8) Epoch 9, batch 7000, loss[loss=0.2079, simple_loss=0.3017, pruned_loss=0.05708, over 16815.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3135, pruned_loss=0.07508, over 3123421.24 frames. ], batch size: 102, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:20:37,441 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9151, 2.0206, 2.3497, 3.1426, 2.0934, 2.2724, 2.2420, 2.0926], device='cuda:5'), covar=tensor([0.0874, 0.2788, 0.1620, 0.0479, 0.3392, 0.2084, 0.2540, 0.2835], device='cuda:5'), in_proj_covar=tensor([0.0351, 0.0374, 0.0311, 0.0318, 0.0404, 0.0423, 0.0334, 0.0437], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:20:43,257 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:03,409 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 3.320e+02 4.209e+02 4.909e+02 8.638e+02, threshold=8.417e+02, percent-clipped=2.0 2023-04-29 02:21:12,422 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0523, 1.9193, 2.1147, 3.4951, 1.8678, 2.2821, 2.0377, 2.0423], device='cuda:5'), covar=tensor([0.0982, 0.3057, 0.2072, 0.0457, 0.3775, 0.2108, 0.2888, 0.2952], device='cuda:5'), in_proj_covar=tensor([0.0352, 0.0375, 0.0312, 0.0319, 0.0405, 0.0424, 0.0336, 0.0437], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:21:12,545 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.21 vs. limit=5.0 2023-04-29 02:21:25,428 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4270, 2.8647, 2.6820, 2.2410, 2.2992, 2.1002, 2.8159, 2.8454], device='cuda:5'), covar=tensor([0.2010, 0.0839, 0.1305, 0.1791, 0.1762, 0.1694, 0.0459, 0.0888], device='cuda:5'), in_proj_covar=tensor([0.0296, 0.0254, 0.0280, 0.0270, 0.0279, 0.0214, 0.0263, 0.0281], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:21:29,044 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:36,659 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88245.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:47,207 INFO [train.py:904] (5/8) Epoch 9, batch 7050, loss[loss=0.2223, simple_loss=0.3092, pruned_loss=0.06769, over 15313.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3147, pruned_loss=0.07527, over 3121061.12 frames. ], batch size: 190, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:22:00,401 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:22:41,150 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:22:45,559 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:22:47,895 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:23:02,087 INFO [train.py:904] (5/8) Epoch 9, batch 7100, loss[loss=0.2356, simple_loss=0.3109, pruned_loss=0.08016, over 16973.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3129, pruned_loss=0.07495, over 3122584.49 frames. ], batch size: 41, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:23:12,649 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:23:33,103 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.153e+02 3.325e+02 4.085e+02 5.101e+02 9.859e+02, threshold=8.169e+02, percent-clipped=1.0 2023-04-29 02:23:43,686 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3416, 1.8879, 2.7419, 3.3813, 3.1836, 3.8829, 2.4043, 3.6399], device='cuda:5'), covar=tensor([0.0124, 0.0322, 0.0191, 0.0121, 0.0152, 0.0060, 0.0274, 0.0062], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0159, 0.0143, 0.0143, 0.0152, 0.0111, 0.0160, 0.0103], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 02:24:15,985 INFO [train.py:904] (5/8) Epoch 9, batch 7150, loss[loss=0.1971, simple_loss=0.2893, pruned_loss=0.05243, over 16738.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3111, pruned_loss=0.07488, over 3108122.33 frames. ], batch size: 83, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:24:17,083 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:25:28,100 INFO [train.py:904] (5/8) Epoch 9, batch 7200, loss[loss=0.1893, simple_loss=0.2793, pruned_loss=0.04972, over 16731.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3085, pruned_loss=0.07331, over 3086873.56 frames. ], batch size: 134, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:00,149 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.048e+02 3.005e+02 3.439e+02 4.228e+02 8.504e+02, threshold=6.879e+02, percent-clipped=1.0 2023-04-29 02:26:47,217 INFO [train.py:904] (5/8) Epoch 9, batch 7250, loss[loss=0.2478, simple_loss=0.31, pruned_loss=0.09281, over 11429.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3053, pruned_loss=0.07124, over 3094722.70 frames. ], batch size: 248, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:53,324 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:28:00,998 INFO [train.py:904] (5/8) Epoch 9, batch 7300, loss[loss=0.2021, simple_loss=0.2917, pruned_loss=0.05626, over 17083.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3051, pruned_loss=0.07148, over 3079694.60 frames. ], batch size: 53, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:28:33,454 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 3.255e+02 4.092e+02 5.788e+02 1.345e+03, threshold=8.184e+02, percent-clipped=12.0 2023-04-29 02:29:14,120 INFO [train.py:904] (5/8) Epoch 9, batch 7350, loss[loss=0.2166, simple_loss=0.2956, pruned_loss=0.06882, over 16675.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3068, pruned_loss=0.07328, over 3044904.70 frames. ], batch size: 62, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:29:17,669 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3030, 4.2897, 4.1981, 2.7914, 3.7924, 4.1310, 3.8414, 2.5123], device='cuda:5'), covar=tensor([0.0393, 0.0016, 0.0022, 0.0263, 0.0049, 0.0070, 0.0038, 0.0295], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0063, 0.0065, 0.0123, 0.0072, 0.0083, 0.0072, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 02:29:27,468 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 02:29:41,156 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-29 02:30:27,932 INFO [train.py:904] (5/8) Epoch 9, batch 7400, loss[loss=0.2411, simple_loss=0.3215, pruned_loss=0.0803, over 16505.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3072, pruned_loss=0.07355, over 3048328.28 frames. ], batch size: 68, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:30:38,098 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 02:30:47,488 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 02:31:01,755 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.901e+02 3.181e+02 3.584e+02 4.494e+02 7.659e+02, threshold=7.169e+02, percent-clipped=0.0 2023-04-29 02:31:38,478 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:31:44,629 INFO [train.py:904] (5/8) Epoch 9, batch 7450, loss[loss=0.293, simple_loss=0.3449, pruned_loss=0.1205, over 11722.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3084, pruned_loss=0.07475, over 3051630.89 frames. ], batch size: 248, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:02,971 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7087, 3.8360, 3.1276, 2.2378, 2.8385, 2.4166, 4.1023, 3.6420], device='cuda:5'), covar=tensor([0.2509, 0.0674, 0.1445, 0.2023, 0.2136, 0.1677, 0.0431, 0.0896], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0256, 0.0281, 0.0270, 0.0281, 0.0215, 0.0264, 0.0285], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:33:03,524 INFO [train.py:904] (5/8) Epoch 9, batch 7500, loss[loss=0.2097, simple_loss=0.2951, pruned_loss=0.06218, over 16508.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3092, pruned_loss=0.07431, over 3052685.85 frames. ], batch size: 75, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:36,933 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 3.551e+02 4.142e+02 5.326e+02 1.060e+03, threshold=8.283e+02, percent-clipped=5.0 2023-04-29 02:33:46,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6543, 4.4762, 4.7205, 4.9496, 5.0309, 4.5371, 5.0290, 4.9943], device='cuda:5'), covar=tensor([0.1538, 0.1007, 0.1310, 0.0495, 0.0486, 0.0743, 0.0472, 0.0524], device='cuda:5'), in_proj_covar=tensor([0.0473, 0.0581, 0.0720, 0.0587, 0.0456, 0.0452, 0.0475, 0.0525], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:33:58,008 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6444, 2.6215, 1.8466, 2.7861, 2.1806, 2.7523, 2.0654, 2.3696], device='cuda:5'), covar=tensor([0.0200, 0.0360, 0.1128, 0.0141, 0.0616, 0.0496, 0.1044, 0.0521], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0159, 0.0184, 0.0111, 0.0165, 0.0199, 0.0191, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 02:34:18,167 INFO [train.py:904] (5/8) Epoch 9, batch 7550, loss[loss=0.2193, simple_loss=0.2953, pruned_loss=0.07163, over 15340.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3084, pruned_loss=0.0741, over 3065493.71 frames. ], batch size: 190, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:34:23,598 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:35:33,598 INFO [train.py:904] (5/8) Epoch 9, batch 7600, loss[loss=0.2232, simple_loss=0.3118, pruned_loss=0.06726, over 16234.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3075, pruned_loss=0.07404, over 3078219.60 frames. ], batch size: 165, lr: 7.51e-03, grad_scale: 8.0 2023-04-29 02:35:37,865 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:36:05,555 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 3.313e+02 3.789e+02 4.639e+02 1.052e+03, threshold=7.577e+02, percent-clipped=2.0 2023-04-29 02:36:42,882 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8262, 2.7654, 2.4895, 4.5020, 3.3874, 4.1852, 1.4815, 2.9716], device='cuda:5'), covar=tensor([0.1298, 0.0701, 0.1176, 0.0145, 0.0333, 0.0366, 0.1540, 0.0832], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0155, 0.0178, 0.0127, 0.0203, 0.0205, 0.0176, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 02:36:45,921 INFO [train.py:904] (5/8) Epoch 9, batch 7650, loss[loss=0.2249, simple_loss=0.3065, pruned_loss=0.07167, over 16698.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3083, pruned_loss=0.07494, over 3073780.73 frames. ], batch size: 134, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:38:01,881 INFO [train.py:904] (5/8) Epoch 9, batch 7700, loss[loss=0.2225, simple_loss=0.3044, pruned_loss=0.07037, over 17056.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3078, pruned_loss=0.07472, over 3103132.16 frames. ], batch size: 55, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:38:27,596 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 02:38:34,566 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 3.502e+02 4.245e+02 5.386e+02 9.706e+02, threshold=8.489e+02, percent-clipped=2.0 2023-04-29 02:39:00,073 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-29 02:39:08,844 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:39:16,838 INFO [train.py:904] (5/8) Epoch 9, batch 7750, loss[loss=0.2082, simple_loss=0.2941, pruned_loss=0.06117, over 17022.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3088, pruned_loss=0.07509, over 3101414.22 frames. ], batch size: 53, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:40:01,645 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 02:40:20,785 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:40:29,356 INFO [train.py:904] (5/8) Epoch 9, batch 7800, loss[loss=0.2073, simple_loss=0.2959, pruned_loss=0.05931, over 16679.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3103, pruned_loss=0.07614, over 3094529.89 frames. ], batch size: 89, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:40:46,639 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9608, 3.4877, 2.9856, 1.6924, 2.6305, 2.0890, 3.3099, 3.6121], device='cuda:5'), covar=tensor([0.0278, 0.0559, 0.0743, 0.2041, 0.0955, 0.1138, 0.0729, 0.0792], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0139, 0.0157, 0.0141, 0.0133, 0.0125, 0.0136, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 02:40:58,353 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9064, 1.9829, 2.2883, 3.1445, 2.1491, 2.3210, 2.2167, 2.1360], device='cuda:5'), covar=tensor([0.0861, 0.2845, 0.1659, 0.0491, 0.3408, 0.1960, 0.2510, 0.2607], device='cuda:5'), in_proj_covar=tensor([0.0350, 0.0379, 0.0315, 0.0319, 0.0410, 0.0428, 0.0338, 0.0440], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:41:02,676 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.440e+02 3.396e+02 4.306e+02 5.337e+02 1.534e+03, threshold=8.611e+02, percent-clipped=4.0 2023-04-29 02:41:06,984 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 02:41:07,736 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:41:44,468 INFO [train.py:904] (5/8) Epoch 9, batch 7850, loss[loss=0.2132, simple_loss=0.3079, pruned_loss=0.05928, over 16442.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3111, pruned_loss=0.07556, over 3105944.78 frames. ], batch size: 68, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:42:38,994 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:42:57,238 INFO [train.py:904] (5/8) Epoch 9, batch 7900, loss[loss=0.3041, simple_loss=0.3607, pruned_loss=0.1237, over 11399.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3099, pruned_loss=0.07515, over 3093152.98 frames. ], batch size: 247, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:43:28,447 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.998e+02 3.084e+02 3.809e+02 4.466e+02 8.463e+02, threshold=7.617e+02, percent-clipped=0.0 2023-04-29 02:44:12,934 INFO [train.py:904] (5/8) Epoch 9, batch 7950, loss[loss=0.2022, simple_loss=0.2889, pruned_loss=0.05771, over 17213.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3094, pruned_loss=0.07511, over 3099625.74 frames. ], batch size: 44, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:44:23,008 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-29 02:44:39,203 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:45:26,772 INFO [train.py:904] (5/8) Epoch 9, batch 8000, loss[loss=0.2713, simple_loss=0.3371, pruned_loss=0.1027, over 11450.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3101, pruned_loss=0.076, over 3088780.48 frames. ], batch size: 247, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:45:59,151 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-04-29 02:45:59,521 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.399e+02 3.668e+02 4.137e+02 4.605e+02 8.803e+02, threshold=8.275e+02, percent-clipped=3.0 2023-04-29 02:46:09,113 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:46:40,089 INFO [train.py:904] (5/8) Epoch 9, batch 8050, loss[loss=0.3169, simple_loss=0.359, pruned_loss=0.1374, over 11841.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3099, pruned_loss=0.0754, over 3105235.94 frames. ], batch size: 249, lr: 7.49e-03, grad_scale: 4.0 2023-04-29 02:47:55,825 INFO [train.py:904] (5/8) Epoch 9, batch 8100, loss[loss=0.2514, simple_loss=0.3204, pruned_loss=0.09117, over 11692.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3091, pruned_loss=0.07423, over 3111086.12 frames. ], batch size: 246, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:48:29,434 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.308e+02 2.948e+02 3.621e+02 4.670e+02 8.254e+02, threshold=7.243e+02, percent-clipped=0.0 2023-04-29 02:48:48,843 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2293, 1.8975, 1.6028, 1.6854, 2.1771, 1.9903, 2.1637, 2.3504], device='cuda:5'), covar=tensor([0.0097, 0.0232, 0.0325, 0.0308, 0.0153, 0.0225, 0.0150, 0.0158], device='cuda:5'), in_proj_covar=tensor([0.0123, 0.0191, 0.0190, 0.0189, 0.0190, 0.0191, 0.0192, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:49:10,983 INFO [train.py:904] (5/8) Epoch 9, batch 8150, loss[loss=0.2737, simple_loss=0.3255, pruned_loss=0.1109, over 11591.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3073, pruned_loss=0.07371, over 3090209.91 frames. ], batch size: 247, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:49:15,213 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1176, 3.0375, 3.1226, 1.7724, 3.3312, 3.3911, 2.6975, 2.5188], device='cuda:5'), covar=tensor([0.0805, 0.0172, 0.0179, 0.1139, 0.0062, 0.0118, 0.0431, 0.0444], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0100, 0.0088, 0.0142, 0.0068, 0.0096, 0.0120, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 02:49:19,359 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7922, 4.8012, 4.6255, 4.4031, 4.1841, 4.7068, 4.6249, 4.3004], device='cuda:5'), covar=tensor([0.0601, 0.0405, 0.0293, 0.0274, 0.1076, 0.0392, 0.0317, 0.0701], device='cuda:5'), in_proj_covar=tensor([0.0228, 0.0275, 0.0262, 0.0243, 0.0284, 0.0275, 0.0181, 0.0308], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:49:45,580 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:49:58,380 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7982, 3.8695, 2.4097, 4.4566, 2.9269, 4.3981, 2.4177, 3.1105], device='cuda:5'), covar=tensor([0.0197, 0.0317, 0.1419, 0.0100, 0.0714, 0.0409, 0.1406, 0.0591], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0160, 0.0185, 0.0112, 0.0165, 0.0202, 0.0194, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 02:49:59,366 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:50:28,017 INFO [train.py:904] (5/8) Epoch 9, batch 8200, loss[loss=0.1863, simple_loss=0.2703, pruned_loss=0.05116, over 17009.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.304, pruned_loss=0.07227, over 3100439.49 frames. ], batch size: 50, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:51:05,381 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89425.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:51:06,154 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.992e+02 3.313e+02 4.005e+02 4.576e+02 8.683e+02, threshold=8.011e+02, percent-clipped=3.0 2023-04-29 02:51:06,839 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0478, 2.8483, 2.8215, 2.0973, 2.7217, 2.1706, 2.8766, 2.9348], device='cuda:5'), covar=tensor([0.0296, 0.0593, 0.0434, 0.1587, 0.0648, 0.0941, 0.0517, 0.0575], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0139, 0.0157, 0.0142, 0.0134, 0.0125, 0.0136, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 02:51:20,861 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:51:27,542 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:51:47,441 INFO [train.py:904] (5/8) Epoch 9, batch 8250, loss[loss=0.2241, simple_loss=0.3177, pruned_loss=0.06522, over 16118.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.303, pruned_loss=0.06995, over 3090117.81 frames. ], batch size: 165, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:52:01,291 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2504, 5.5472, 5.3186, 5.3531, 4.9722, 4.8238, 4.9819, 5.6307], device='cuda:5'), covar=tensor([0.0984, 0.0796, 0.1060, 0.0619, 0.0762, 0.0686, 0.0919, 0.0823], device='cuda:5'), in_proj_covar=tensor([0.0493, 0.0623, 0.0526, 0.0424, 0.0387, 0.0407, 0.0522, 0.0472], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:52:30,978 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0362, 4.2222, 3.8983, 3.7948, 3.1786, 4.1406, 3.9418, 3.7469], device='cuda:5'), covar=tensor([0.0789, 0.0516, 0.0476, 0.0370, 0.1667, 0.0487, 0.0823, 0.0752], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0271, 0.0260, 0.0240, 0.0280, 0.0271, 0.0180, 0.0304], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 02:52:43,724 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:06,312 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:53:08,501 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 02:53:08,811 INFO [train.py:904] (5/8) Epoch 9, batch 8300, loss[loss=0.1873, simple_loss=0.2904, pruned_loss=0.04209, over 16886.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2999, pruned_loss=0.06647, over 3082190.10 frames. ], batch size: 96, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:53:21,167 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:47,803 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89525.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:48,530 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.763e+02 2.745e+02 3.164e+02 3.682e+02 6.265e+02, threshold=6.329e+02, percent-clipped=0.0 2023-04-29 02:54:31,544 INFO [train.py:904] (5/8) Epoch 9, batch 8350, loss[loss=0.1995, simple_loss=0.2868, pruned_loss=0.05612, over 15135.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2988, pruned_loss=0.06461, over 3067272.66 frames. ], batch size: 190, lr: 7.47e-03, grad_scale: 2.0 2023-04-29 02:55:01,170 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:55:11,752 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 02:55:52,211 INFO [train.py:904] (5/8) Epoch 9, batch 8400, loss[loss=0.1793, simple_loss=0.2796, pruned_loss=0.03949, over 16831.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2953, pruned_loss=0.06171, over 3068685.56 frames. ], batch size: 102, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:56:31,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.203e+02 2.904e+02 3.369e+02 3.930e+02 8.032e+02, threshold=6.737e+02, percent-clipped=2.0 2023-04-29 02:56:47,996 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8996, 2.2627, 2.2098, 2.9001, 1.9798, 3.3080, 1.6324, 2.7138], device='cuda:5'), covar=tensor([0.1218, 0.0634, 0.0981, 0.0121, 0.0114, 0.0366, 0.1352, 0.0632], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0150, 0.0173, 0.0125, 0.0195, 0.0201, 0.0173, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 02:57:13,783 INFO [train.py:904] (5/8) Epoch 9, batch 8450, loss[loss=0.1944, simple_loss=0.2729, pruned_loss=0.05789, over 12274.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2938, pruned_loss=0.06023, over 3065367.30 frames. ], batch size: 248, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:58:00,200 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:04,749 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:34,966 INFO [train.py:904] (5/8) Epoch 9, batch 8500, loss[loss=0.1751, simple_loss=0.2697, pruned_loss=0.04025, over 16897.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2894, pruned_loss=0.05759, over 3056576.12 frames. ], batch size: 96, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:59:14,101 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.444e+02 3.113e+02 3.731e+02 7.658e+02, threshold=6.225e+02, percent-clipped=1.0 2023-04-29 02:59:20,987 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89730.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:22,203 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:39,468 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:58,603 INFO [train.py:904] (5/8) Epoch 9, batch 8550, loss[loss=0.2105, simple_loss=0.2993, pruned_loss=0.06091, over 15255.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.287, pruned_loss=0.05627, over 3055217.39 frames. ], batch size: 190, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 03:00:21,688 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 03:00:55,442 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:01:24,775 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:01:37,962 INFO [train.py:904] (5/8) Epoch 9, batch 8600, loss[loss=0.2122, simple_loss=0.3043, pruned_loss=0.0601, over 16647.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2877, pruned_loss=0.05536, over 3079285.21 frames. ], batch size: 134, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:02:23,608 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3571, 1.8965, 1.6011, 1.6174, 2.1729, 1.9744, 2.1359, 2.3284], device='cuda:5'), covar=tensor([0.0070, 0.0246, 0.0334, 0.0314, 0.0157, 0.0227, 0.0127, 0.0166], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0188, 0.0186, 0.0187, 0.0186, 0.0187, 0.0184, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:02:25,747 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:02:26,427 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.695e+02 3.313e+02 4.176e+02 8.025e+02, threshold=6.625e+02, percent-clipped=2.0 2023-04-29 03:03:15,783 INFO [train.py:904] (5/8) Epoch 9, batch 8650, loss[loss=0.1823, simple_loss=0.2686, pruned_loss=0.04802, over 12281.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2854, pruned_loss=0.05403, over 3047217.04 frames. ], batch size: 248, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:03:47,703 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:47,862 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:49,632 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2610, 1.9372, 2.1162, 3.6637, 1.8920, 2.3661, 2.1026, 2.0633], device='cuda:5'), covar=tensor([0.0775, 0.3606, 0.2259, 0.0393, 0.4104, 0.2098, 0.3100, 0.3348], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0367, 0.0308, 0.0307, 0.0398, 0.0411, 0.0327, 0.0427], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:03:57,753 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9102, 2.3017, 2.2205, 2.9204, 2.0233, 3.2873, 1.5618, 2.8038], device='cuda:5'), covar=tensor([0.1252, 0.0587, 0.0991, 0.0113, 0.0098, 0.0337, 0.1414, 0.0604], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0150, 0.0174, 0.0124, 0.0192, 0.0202, 0.0173, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 03:04:05,357 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:04:42,285 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 03:05:02,366 INFO [train.py:904] (5/8) Epoch 9, batch 8700, loss[loss=0.1811, simple_loss=0.2659, pruned_loss=0.04817, over 16533.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2821, pruned_loss=0.05268, over 3032769.05 frames. ], batch size: 68, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:05:45,067 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.547e+02 3.087e+02 3.551e+02 5.785e+02, threshold=6.175e+02, percent-clipped=0.0 2023-04-29 03:05:45,881 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 03:06:36,271 INFO [train.py:904] (5/8) Epoch 9, batch 8750, loss[loss=0.1802, simple_loss=0.2661, pruned_loss=0.04716, over 12134.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2814, pruned_loss=0.05175, over 3050352.46 frames. ], batch size: 248, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:07:43,922 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 03:08:32,648 INFO [train.py:904] (5/8) Epoch 9, batch 8800, loss[loss=0.1903, simple_loss=0.2834, pruned_loss=0.0486, over 16815.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2794, pruned_loss=0.0498, over 3077279.17 frames. ], batch size: 90, lr: 7.46e-03, grad_scale: 8.0 2023-04-29 03:09:21,969 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.546e+02 3.019e+02 3.613e+02 7.330e+02, threshold=6.037e+02, percent-clipped=3.0 2023-04-29 03:09:31,765 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:09:44,381 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:10:17,862 INFO [train.py:904] (5/8) Epoch 9, batch 8850, loss[loss=0.1771, simple_loss=0.2816, pruned_loss=0.03632, over 15397.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2818, pruned_loss=0.04922, over 3062960.43 frames. ], batch size: 192, lr: 7.45e-03, grad_scale: 8.0 2023-04-29 03:11:12,094 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:18,201 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:50,057 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:12:02,867 INFO [train.py:904] (5/8) Epoch 9, batch 8900, loss[loss=0.2009, simple_loss=0.2925, pruned_loss=0.05467, over 15421.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2829, pruned_loss=0.04904, over 3059369.43 frames. ], batch size: 191, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:12:18,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4335, 3.0301, 2.5956, 2.1778, 2.2309, 2.1284, 2.9916, 2.9679], device='cuda:5'), covar=tensor([0.2140, 0.0624, 0.1397, 0.2179, 0.1995, 0.1646, 0.0477, 0.0836], device='cuda:5'), in_proj_covar=tensor([0.0288, 0.0243, 0.0271, 0.0260, 0.0256, 0.0206, 0.0252, 0.0266], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:12:57,514 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.859e+02 2.641e+02 3.095e+02 3.699e+02 6.742e+02, threshold=6.190e+02, percent-clipped=4.0 2023-04-29 03:13:07,419 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:13:10,567 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:13:17,353 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1872, 1.9619, 2.1737, 3.6454, 1.9537, 2.4031, 2.1177, 2.1250], device='cuda:5'), covar=tensor([0.0807, 0.3032, 0.1961, 0.0384, 0.3693, 0.1988, 0.2810, 0.2845], device='cuda:5'), in_proj_covar=tensor([0.0339, 0.0367, 0.0311, 0.0308, 0.0400, 0.0411, 0.0329, 0.0426], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:13:48,685 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 03:14:08,714 INFO [train.py:904] (5/8) Epoch 9, batch 8950, loss[loss=0.1701, simple_loss=0.2658, pruned_loss=0.03721, over 16848.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2817, pruned_loss=0.04907, over 3051504.09 frames. ], batch size: 76, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:14:10,678 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 03:14:14,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0594, 3.9215, 4.1371, 4.2910, 4.3694, 3.9693, 4.3613, 4.3643], device='cuda:5'), covar=tensor([0.1434, 0.1009, 0.1218, 0.0516, 0.0501, 0.1012, 0.0489, 0.0481], device='cuda:5'), in_proj_covar=tensor([0.0458, 0.0562, 0.0683, 0.0569, 0.0441, 0.0438, 0.0458, 0.0514], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:14:38,312 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:15:35,022 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:15:57,373 INFO [train.py:904] (5/8) Epoch 9, batch 9000, loss[loss=0.1789, simple_loss=0.2644, pruned_loss=0.04668, over 16680.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2784, pruned_loss=0.04743, over 3082252.30 frames. ], batch size: 134, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:15:57,373 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 03:16:07,533 INFO [train.py:938] (5/8) Epoch 9, validation: loss=0.1581, simple_loss=0.2623, pruned_loss=0.02697, over 944034.00 frames. 2023-04-29 03:16:07,535 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 03:16:31,272 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:16:47,984 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 03:16:53,582 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 03:16:58,707 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.701e+02 2.410e+02 2.881e+02 3.616e+02 7.746e+02, threshold=5.761e+02, percent-clipped=2.0 2023-04-29 03:17:09,258 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:17:09,298 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7602, 3.1917, 2.8063, 4.7589, 3.7270, 4.4968, 1.5989, 3.3846], device='cuda:5'), covar=tensor([0.1288, 0.0531, 0.0963, 0.0091, 0.0144, 0.0252, 0.1383, 0.0552], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0151, 0.0175, 0.0124, 0.0188, 0.0201, 0.0174, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 03:17:51,888 INFO [train.py:904] (5/8) Epoch 9, batch 9050, loss[loss=0.1982, simple_loss=0.2826, pruned_loss=0.05693, over 16208.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2794, pruned_loss=0.04831, over 3077254.18 frames. ], batch size: 165, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:17:55,149 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8677, 3.0561, 3.1174, 2.0715, 2.8635, 3.1094, 2.9908, 1.9584], device='cuda:5'), covar=tensor([0.0437, 0.0037, 0.0042, 0.0313, 0.0064, 0.0077, 0.0058, 0.0370], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0063, 0.0064, 0.0122, 0.0072, 0.0081, 0.0072, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 03:18:10,415 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 03:19:19,979 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:19:36,856 INFO [train.py:904] (5/8) Epoch 9, batch 9100, loss[loss=0.178, simple_loss=0.2626, pruned_loss=0.04675, over 12315.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2791, pruned_loss=0.04867, over 3090650.12 frames. ], batch size: 247, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:19:54,555 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 03:20:34,111 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.617e+02 3.274e+02 4.462e+02 7.495e+02, threshold=6.548e+02, percent-clipped=8.0 2023-04-29 03:20:38,293 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2010, 3.9782, 4.2129, 4.3862, 4.5153, 4.0393, 4.5037, 4.5050], device='cuda:5'), covar=tensor([0.1398, 0.0990, 0.1346, 0.0580, 0.0464, 0.1097, 0.0479, 0.0466], device='cuda:5'), in_proj_covar=tensor([0.0459, 0.0566, 0.0686, 0.0571, 0.0439, 0.0439, 0.0456, 0.0514], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:20:44,018 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 03:20:51,083 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3015, 3.3458, 1.9198, 3.6729, 2.3849, 3.5999, 1.9495, 2.7324], device='cuda:5'), covar=tensor([0.0213, 0.0294, 0.1434, 0.0112, 0.0831, 0.0393, 0.1555, 0.0613], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0152, 0.0178, 0.0106, 0.0157, 0.0188, 0.0186, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 03:20:58,617 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:21:34,632 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90351.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:21:35,365 INFO [train.py:904] (5/8) Epoch 9, batch 9150, loss[loss=0.1687, simple_loss=0.2608, pruned_loss=0.03828, over 16655.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2797, pruned_loss=0.04829, over 3092462.25 frames. ], batch size: 57, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:22:10,782 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4118, 3.3554, 3.4395, 3.5406, 3.5738, 3.2555, 3.5542, 3.5951], device='cuda:5'), covar=tensor([0.0977, 0.0726, 0.0888, 0.0519, 0.0494, 0.1950, 0.0674, 0.0562], device='cuda:5'), in_proj_covar=tensor([0.0458, 0.0564, 0.0683, 0.0570, 0.0437, 0.0439, 0.0455, 0.0513], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:22:13,020 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1165, 2.6479, 2.6595, 1.8151, 2.8684, 2.9110, 2.5254, 2.4023], device='cuda:5'), covar=tensor([0.0642, 0.0175, 0.0168, 0.0909, 0.0080, 0.0139, 0.0355, 0.0383], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0095, 0.0081, 0.0137, 0.0065, 0.0090, 0.0116, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-29 03:22:45,427 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:23:22,760 INFO [train.py:904] (5/8) Epoch 9, batch 9200, loss[loss=0.1861, simple_loss=0.2755, pruned_loss=0.04835, over 15325.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2751, pruned_loss=0.04697, over 3097825.67 frames. ], batch size: 191, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:23:42,443 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90412.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:24:07,628 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.608e+02 2.952e+02 3.728e+02 7.219e+02, threshold=5.904e+02, percent-clipped=3.0 2023-04-29 03:24:58,585 INFO [train.py:904] (5/8) Epoch 9, batch 9250, loss[loss=0.1686, simple_loss=0.2653, pruned_loss=0.03597, over 16925.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2748, pruned_loss=0.0471, over 3109448.63 frames. ], batch size: 109, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:25:14,817 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:13,042 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:38,467 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:49,746 INFO [train.py:904] (5/8) Epoch 9, batch 9300, loss[loss=0.1682, simple_loss=0.257, pruned_loss=0.03972, over 15206.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2728, pruned_loss=0.04634, over 3091041.44 frames. ], batch size: 190, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:27:35,121 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:27:37,003 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:27:46,917 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.420e+02 2.895e+02 3.675e+02 7.033e+02, threshold=5.790e+02, percent-clipped=1.0 2023-04-29 03:28:36,085 INFO [train.py:904] (5/8) Epoch 9, batch 9350, loss[loss=0.189, simple_loss=0.2671, pruned_loss=0.05541, over 12565.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2724, pruned_loss=0.0463, over 3080925.49 frames. ], batch size: 247, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:28:48,558 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:14,438 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:50,764 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:30:18,708 INFO [train.py:904] (5/8) Epoch 9, batch 9400, loss[loss=0.1689, simple_loss=0.2593, pruned_loss=0.03927, over 12659.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2728, pruned_loss=0.046, over 3078324.28 frames. ], batch size: 248, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:30:46,266 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6609, 2.6862, 1.7887, 2.8744, 2.1583, 2.8430, 1.9849, 2.4217], device='cuda:5'), covar=tensor([0.0189, 0.0331, 0.1323, 0.0157, 0.0685, 0.0464, 0.1242, 0.0546], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0153, 0.0180, 0.0108, 0.0159, 0.0187, 0.0187, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 03:31:09,568 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.474e+02 2.959e+02 3.714e+02 8.908e+02, threshold=5.918e+02, percent-clipped=5.0 2023-04-29 03:32:00,901 INFO [train.py:904] (5/8) Epoch 9, batch 9450, loss[loss=0.1778, simple_loss=0.2775, pruned_loss=0.03903, over 16823.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2741, pruned_loss=0.04589, over 3076362.03 frames. ], batch size: 90, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:32:43,731 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9087, 5.1863, 4.9704, 4.9906, 4.6637, 4.6651, 4.6541, 5.2509], device='cuda:5'), covar=tensor([0.0823, 0.0744, 0.0866, 0.0576, 0.0668, 0.0738, 0.0884, 0.0774], device='cuda:5'), in_proj_covar=tensor([0.0467, 0.0594, 0.0485, 0.0406, 0.0369, 0.0387, 0.0496, 0.0442], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:33:43,980 INFO [train.py:904] (5/8) Epoch 9, batch 9500, loss[loss=0.1723, simple_loss=0.2641, pruned_loss=0.04027, over 16691.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2735, pruned_loss=0.0455, over 3096781.46 frames. ], batch size: 134, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:33:58,747 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:34:35,333 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.298e+02 2.954e+02 3.660e+02 6.291e+02, threshold=5.908e+02, percent-clipped=3.0 2023-04-29 03:35:30,044 INFO [train.py:904] (5/8) Epoch 9, batch 9550, loss[loss=0.1798, simple_loss=0.2632, pruned_loss=0.04819, over 12750.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.274, pruned_loss=0.04628, over 3098857.48 frames. ], batch size: 248, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:36:43,111 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:37:11,192 INFO [train.py:904] (5/8) Epoch 9, batch 9600, loss[loss=0.2114, simple_loss=0.3019, pruned_loss=0.06043, over 16747.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2758, pruned_loss=0.04716, over 3100866.13 frames. ], batch size: 83, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:37:24,958 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 03:37:37,099 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:37:59,122 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.542e+02 3.036e+02 4.023e+02 8.440e+02, threshold=6.073e+02, percent-clipped=4.0 2023-04-29 03:38:13,544 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6189, 3.7495, 4.0095, 1.8673, 4.2014, 4.2046, 3.2461, 3.0165], device='cuda:5'), covar=tensor([0.0672, 0.0150, 0.0131, 0.1105, 0.0041, 0.0072, 0.0279, 0.0393], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0093, 0.0079, 0.0135, 0.0064, 0.0090, 0.0114, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-29 03:38:17,844 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:38:55,243 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:39:01,501 INFO [train.py:904] (5/8) Epoch 9, batch 9650, loss[loss=0.2054, simple_loss=0.2937, pruned_loss=0.05858, over 15454.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2771, pruned_loss=0.04739, over 3081726.75 frames. ], batch size: 193, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:39:03,594 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:40:19,141 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:40:48,113 INFO [train.py:904] (5/8) Epoch 9, batch 9700, loss[loss=0.1932, simple_loss=0.2799, pruned_loss=0.05318, over 16424.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.276, pruned_loss=0.04714, over 3076548.51 frames. ], batch size: 147, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:41:04,622 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:41:09,042 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9813, 2.4240, 2.3377, 3.1003, 2.2591, 3.2984, 1.6758, 2.8372], device='cuda:5'), covar=tensor([0.1155, 0.0503, 0.0943, 0.0112, 0.0085, 0.0365, 0.1307, 0.0596], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0149, 0.0174, 0.0121, 0.0179, 0.0199, 0.0173, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 03:41:25,772 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5426, 3.5570, 3.4958, 3.0278, 3.4785, 1.9920, 3.3156, 3.0848], device='cuda:5'), covar=tensor([0.0094, 0.0081, 0.0112, 0.0188, 0.0074, 0.1813, 0.0094, 0.0167], device='cuda:5'), in_proj_covar=tensor([0.0111, 0.0099, 0.0144, 0.0132, 0.0115, 0.0165, 0.0130, 0.0134], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-29 03:41:40,524 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.339e+02 3.062e+02 3.716e+02 7.920e+02, threshold=6.123e+02, percent-clipped=1.0 2023-04-29 03:41:46,261 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-04-29 03:41:49,891 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 03:42:00,641 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:42:23,254 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 03:42:31,495 INFO [train.py:904] (5/8) Epoch 9, batch 9750, loss[loss=0.1783, simple_loss=0.2629, pruned_loss=0.04691, over 12569.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2752, pruned_loss=0.04729, over 3076755.23 frames. ], batch size: 248, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:43:40,095 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4074, 2.9468, 2.7127, 2.2465, 2.2115, 2.1934, 2.9103, 2.8874], device='cuda:5'), covar=tensor([0.2236, 0.0683, 0.1258, 0.1917, 0.2204, 0.1690, 0.0389, 0.0990], device='cuda:5'), in_proj_covar=tensor([0.0289, 0.0244, 0.0269, 0.0262, 0.0247, 0.0208, 0.0253, 0.0265], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:44:05,950 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3750, 4.4301, 4.5583, 4.4616, 4.4829, 4.9888, 4.5226, 4.2079], device='cuda:5'), covar=tensor([0.1207, 0.1808, 0.1598, 0.1592, 0.2323, 0.0936, 0.1463, 0.2396], device='cuda:5'), in_proj_covar=tensor([0.0304, 0.0432, 0.0462, 0.0379, 0.0498, 0.0483, 0.0378, 0.0501], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:44:10,876 INFO [train.py:904] (5/8) Epoch 9, batch 9800, loss[loss=0.1705, simple_loss=0.2669, pruned_loss=0.03699, over 16553.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2747, pruned_loss=0.04643, over 3073976.93 frames. ], batch size: 62, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:44:21,912 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:44:57,918 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.313e+02 2.751e+02 3.431e+02 5.847e+02, threshold=5.502e+02, percent-clipped=0.0 2023-04-29 03:45:50,624 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9242, 2.2273, 2.2626, 4.7661, 2.1341, 2.7652, 2.2607, 2.5890], device='cuda:5'), covar=tensor([0.0645, 0.3097, 0.2080, 0.0248, 0.3668, 0.1978, 0.2870, 0.2905], device='cuda:5'), in_proj_covar=tensor([0.0336, 0.0363, 0.0309, 0.0306, 0.0395, 0.0403, 0.0326, 0.0420], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:45:57,961 INFO [train.py:904] (5/8) Epoch 9, batch 9850, loss[loss=0.1854, simple_loss=0.2754, pruned_loss=0.04774, over 16905.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2756, pruned_loss=0.04625, over 3053982.47 frames. ], batch size: 116, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:46:05,847 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:47:48,716 INFO [train.py:904] (5/8) Epoch 9, batch 9900, loss[loss=0.1721, simple_loss=0.2549, pruned_loss=0.04469, over 12302.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2759, pruned_loss=0.04625, over 3037707.35 frames. ], batch size: 248, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:48:20,575 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:48:47,936 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.434e+02 3.232e+02 4.008e+02 9.044e+02, threshold=6.464e+02, percent-clipped=5.0 2023-04-29 03:49:47,998 INFO [train.py:904] (5/8) Epoch 9, batch 9950, loss[loss=0.1853, simple_loss=0.2737, pruned_loss=0.04851, over 12365.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2781, pruned_loss=0.04709, over 3016946.33 frames. ], batch size: 249, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:49:49,185 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:50:14,538 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:50:44,993 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6841, 3.2464, 3.0100, 1.7915, 2.6647, 2.1241, 3.2042, 3.2732], device='cuda:5'), covar=tensor([0.0233, 0.0500, 0.0672, 0.1760, 0.0742, 0.0988, 0.0615, 0.0630], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0129, 0.0152, 0.0139, 0.0130, 0.0122, 0.0130, 0.0140], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 03:50:45,018 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4199, 3.4460, 2.6645, 2.1421, 2.2132, 2.1638, 3.4811, 3.2735], device='cuda:5'), covar=tensor([0.2541, 0.0588, 0.1564, 0.2379, 0.1860, 0.1589, 0.0476, 0.0746], device='cuda:5'), in_proj_covar=tensor([0.0285, 0.0242, 0.0266, 0.0258, 0.0243, 0.0205, 0.0250, 0.0262], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:51:44,349 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:51:47,252 INFO [train.py:904] (5/8) Epoch 9, batch 10000, loss[loss=0.1894, simple_loss=0.2956, pruned_loss=0.04157, over 15357.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2764, pruned_loss=0.04653, over 3046268.99 frames. ], batch size: 191, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:51:53,878 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91205.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:52:35,744 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.574e+02 2.985e+02 3.622e+02 7.238e+02, threshold=5.969e+02, percent-clipped=2.0 2023-04-29 03:53:27,231 INFO [train.py:904] (5/8) Epoch 9, batch 10050, loss[loss=0.2208, simple_loss=0.3102, pruned_loss=0.06567, over 16253.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2764, pruned_loss=0.04645, over 3062664.56 frames. ], batch size: 165, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:54:00,661 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5342, 3.5421, 3.4785, 2.9606, 3.4603, 1.8680, 3.2208, 2.9004], device='cuda:5'), covar=tensor([0.0101, 0.0087, 0.0129, 0.0187, 0.0076, 0.2111, 0.0120, 0.0182], device='cuda:5'), in_proj_covar=tensor([0.0115, 0.0102, 0.0147, 0.0135, 0.0117, 0.0170, 0.0134, 0.0137], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:54:30,781 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91285.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:55:01,294 INFO [train.py:904] (5/8) Epoch 9, batch 10100, loss[loss=0.1702, simple_loss=0.2626, pruned_loss=0.03887, over 16833.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2766, pruned_loss=0.04636, over 3067997.92 frames. ], batch size: 124, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:55:52,060 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.558e+02 3.237e+02 3.853e+02 8.628e+02, threshold=6.474e+02, percent-clipped=2.0 2023-04-29 03:56:14,318 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:56:45,269 INFO [train.py:904] (5/8) Epoch 10, batch 0, loss[loss=0.3087, simple_loss=0.3463, pruned_loss=0.1355, over 16765.00 frames. ], tot_loss[loss=0.3087, simple_loss=0.3463, pruned_loss=0.1355, over 16765.00 frames. ], batch size: 124, lr: 7.04e-03, grad_scale: 8.0 2023-04-29 03:56:45,269 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 03:56:50,399 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2374, 4.5484, 4.4138, 4.4632, 4.1598, 4.3299, 4.1358, 4.5646], device='cuda:5'), covar=tensor([0.0798, 0.0688, 0.0645, 0.0467, 0.0709, 0.0364, 0.0744, 0.0622], device='cuda:5'), in_proj_covar=tensor([0.0467, 0.0596, 0.0485, 0.0407, 0.0373, 0.0388, 0.0496, 0.0443], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:56:52,896 INFO [train.py:938] (5/8) Epoch 10, validation: loss=0.158, simple_loss=0.2614, pruned_loss=0.02732, over 944034.00 frames. 2023-04-29 03:56:52,898 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 03:57:19,012 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1276, 5.6728, 5.7947, 5.6075, 5.6653, 6.1349, 5.6687, 5.4611], device='cuda:5'), covar=tensor([0.0725, 0.1621, 0.1791, 0.1887, 0.2448, 0.0986, 0.1252, 0.2119], device='cuda:5'), in_proj_covar=tensor([0.0308, 0.0440, 0.0466, 0.0387, 0.0504, 0.0489, 0.0382, 0.0507], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-29 03:57:55,181 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7453, 4.2386, 4.4577, 3.0949, 3.7788, 4.2864, 4.0024, 2.5855], device='cuda:5'), covar=tensor([0.0338, 0.0032, 0.0024, 0.0251, 0.0076, 0.0046, 0.0047, 0.0338], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0065, 0.0067, 0.0125, 0.0074, 0.0082, 0.0074, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 03:58:02,512 INFO [train.py:904] (5/8) Epoch 10, batch 50, loss[loss=0.2391, simple_loss=0.3015, pruned_loss=0.0883, over 16745.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2962, pruned_loss=0.07234, over 749790.41 frames. ], batch size: 124, lr: 7.04e-03, grad_scale: 2.0 2023-04-29 03:58:39,950 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.984e+02 3.603e+02 4.569e+02 8.591e+02, threshold=7.207e+02, percent-clipped=1.0 2023-04-29 03:58:42,969 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0472, 1.8224, 2.3224, 2.9246, 2.7324, 3.2385, 2.3811, 3.2510], device='cuda:5'), covar=tensor([0.0139, 0.0340, 0.0238, 0.0186, 0.0179, 0.0118, 0.0287, 0.0103], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0165, 0.0149, 0.0149, 0.0160, 0.0115, 0.0166, 0.0102], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 03:58:54,309 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3181, 3.8617, 3.7101, 1.9123, 2.8797, 2.5435, 3.7365, 3.8823], device='cuda:5'), covar=tensor([0.0307, 0.0658, 0.0542, 0.1817, 0.0798, 0.0917, 0.0669, 0.1063], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0130, 0.0154, 0.0140, 0.0131, 0.0123, 0.0131, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 03:59:08,810 INFO [train.py:904] (5/8) Epoch 10, batch 100, loss[loss=0.2251, simple_loss=0.2922, pruned_loss=0.07897, over 16482.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2883, pruned_loss=0.06526, over 1324518.08 frames. ], batch size: 75, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 03:59:12,056 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 03:59:31,420 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4904, 3.4877, 3.4081, 2.8847, 3.4353, 1.9934, 3.1604, 2.8803], device='cuda:5'), covar=tensor([0.0096, 0.0090, 0.0115, 0.0175, 0.0073, 0.1930, 0.0112, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0118, 0.0104, 0.0151, 0.0138, 0.0121, 0.0173, 0.0138, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:00:09,447 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6351, 3.8114, 2.7516, 2.2257, 2.5814, 2.0966, 3.8114, 3.4428], device='cuda:5'), covar=tensor([0.2604, 0.0587, 0.1635, 0.2329, 0.2170, 0.1789, 0.0496, 0.1481], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0248, 0.0274, 0.0265, 0.0251, 0.0211, 0.0258, 0.0274], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:00:16,894 INFO [train.py:904] (5/8) Epoch 10, batch 150, loss[loss=0.2095, simple_loss=0.2908, pruned_loss=0.06409, over 16631.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2836, pruned_loss=0.06153, over 1766513.90 frames. ], batch size: 57, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:00:22,936 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:00:37,219 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0718, 3.4159, 2.9008, 5.2813, 4.5452, 4.8335, 1.9774, 3.4422], device='cuda:5'), covar=tensor([0.1156, 0.0555, 0.0986, 0.0107, 0.0223, 0.0268, 0.1242, 0.0643], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0153, 0.0176, 0.0126, 0.0183, 0.0205, 0.0176, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 04:00:42,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4183, 2.9060, 2.6102, 2.1858, 2.2082, 2.1295, 2.9251, 2.8286], device='cuda:5'), covar=tensor([0.2051, 0.0685, 0.1327, 0.1900, 0.1938, 0.1563, 0.0501, 0.0908], device='cuda:5'), in_proj_covar=tensor([0.0291, 0.0249, 0.0275, 0.0265, 0.0252, 0.0211, 0.0259, 0.0275], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:00:56,398 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.683e+02 3.349e+02 4.064e+02 6.042e+02, threshold=6.698e+02, percent-clipped=0.0 2023-04-29 04:01:03,445 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 04:01:12,430 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:01:26,572 INFO [train.py:904] (5/8) Epoch 10, batch 200, loss[loss=0.2037, simple_loss=0.294, pruned_loss=0.05671, over 17154.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2833, pruned_loss=0.06064, over 2110806.93 frames. ], batch size: 48, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:01:28,066 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:02:34,684 INFO [train.py:904] (5/8) Epoch 10, batch 250, loss[loss=0.1859, simple_loss=0.2706, pruned_loss=0.05058, over 16666.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2798, pruned_loss=0.05938, over 2389451.08 frames. ], batch size: 57, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:02:36,465 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 04:02:39,963 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0707, 5.0729, 4.8656, 4.5292, 4.4528, 4.9573, 4.9660, 4.5325], device='cuda:5'), covar=tensor([0.0519, 0.0359, 0.0292, 0.0271, 0.1162, 0.0366, 0.0273, 0.0642], device='cuda:5'), in_proj_covar=tensor([0.0231, 0.0278, 0.0268, 0.0246, 0.0289, 0.0280, 0.0183, 0.0312], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:03:06,164 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-04-29 04:03:11,344 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.504e+02 3.085e+02 3.691e+02 6.376e+02, threshold=6.169e+02, percent-clipped=0.0 2023-04-29 04:03:28,101 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:03:42,043 INFO [train.py:904] (5/8) Epoch 10, batch 300, loss[loss=0.1968, simple_loss=0.2899, pruned_loss=0.0518, over 17074.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2768, pruned_loss=0.05759, over 2598800.71 frames. ], batch size: 53, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:03:42,603 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3451, 3.6343, 3.3866, 2.1516, 2.9342, 2.4746, 3.7168, 3.6269], device='cuda:5'), covar=tensor([0.0208, 0.0599, 0.0576, 0.1510, 0.0695, 0.0859, 0.0465, 0.0951], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0132, 0.0154, 0.0139, 0.0131, 0.0122, 0.0132, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 04:04:29,081 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1200, 1.7781, 2.4759, 2.9848, 2.8119, 3.2359, 2.2654, 3.2533], device='cuda:5'), covar=tensor([0.0166, 0.0369, 0.0218, 0.0213, 0.0194, 0.0172, 0.0309, 0.0104], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0166, 0.0151, 0.0152, 0.0161, 0.0117, 0.0167, 0.0104], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 04:04:51,294 INFO [train.py:904] (5/8) Epoch 10, batch 350, loss[loss=0.1665, simple_loss=0.257, pruned_loss=0.03806, over 17043.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2744, pruned_loss=0.05587, over 2758935.39 frames. ], batch size: 50, lr: 7.02e-03, grad_scale: 1.0 2023-04-29 04:05:28,619 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.384e+02 2.940e+02 3.585e+02 5.710e+02, threshold=5.881e+02, percent-clipped=0.0 2023-04-29 04:05:35,393 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 04:05:59,434 INFO [train.py:904] (5/8) Epoch 10, batch 400, loss[loss=0.1841, simple_loss=0.2739, pruned_loss=0.04715, over 17162.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2735, pruned_loss=0.05602, over 2885771.02 frames. ], batch size: 46, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:11,421 INFO [train.py:904] (5/8) Epoch 10, batch 450, loss[loss=0.2079, simple_loss=0.307, pruned_loss=0.05436, over 16687.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2719, pruned_loss=0.05551, over 2986518.38 frames. ], batch size: 57, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:50,760 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 2.266e+02 2.904e+02 3.599e+02 6.333e+02, threshold=5.808e+02, percent-clipped=1.0 2023-04-29 04:08:20,253 INFO [train.py:904] (5/8) Epoch 10, batch 500, loss[loss=0.2107, simple_loss=0.2884, pruned_loss=0.06645, over 16790.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2712, pruned_loss=0.0555, over 3050643.22 frames. ], batch size: 89, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:09:15,084 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:09:23,341 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-29 04:09:23,882 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:09:29,475 INFO [train.py:904] (5/8) Epoch 10, batch 550, loss[loss=0.205, simple_loss=0.2793, pruned_loss=0.06535, over 15715.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2699, pruned_loss=0.05517, over 3111356.19 frames. ], batch size: 191, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:09:59,354 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0463, 4.2185, 4.4901, 2.3351, 4.8096, 4.8052, 3.2857, 3.7993], device='cuda:5'), covar=tensor([0.0646, 0.0189, 0.0172, 0.1007, 0.0042, 0.0096, 0.0371, 0.0324], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0099, 0.0085, 0.0142, 0.0069, 0.0099, 0.0121, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 04:10:07,831 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.517e+02 3.166e+02 3.975e+02 1.037e+03, threshold=6.333e+02, percent-clipped=5.0 2023-04-29 04:10:22,741 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:38,349 INFO [train.py:904] (5/8) Epoch 10, batch 600, loss[loss=0.1877, simple_loss=0.2802, pruned_loss=0.04759, over 17035.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2686, pruned_loss=0.054, over 3167105.37 frames. ], batch size: 50, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:10:38,885 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:42,851 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:19,578 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:11:30,063 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:51,309 INFO [train.py:904] (5/8) Epoch 10, batch 650, loss[loss=0.1786, simple_loss=0.271, pruned_loss=0.04309, over 17223.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2674, pruned_loss=0.05343, over 3205135.39 frames. ], batch size: 45, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:12:12,088 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:12:31,523 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-29 04:12:31,800 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.417e+02 3.017e+02 3.569e+02 8.138e+02, threshold=6.033e+02, percent-clipped=1.0 2023-04-29 04:12:42,040 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1081, 5.0918, 4.9106, 4.3961, 4.9287, 1.8558, 4.6863, 4.8519], device='cuda:5'), covar=tensor([0.0074, 0.0055, 0.0138, 0.0337, 0.0090, 0.2263, 0.0108, 0.0180], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0109, 0.0159, 0.0147, 0.0128, 0.0177, 0.0145, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:12:49,512 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:13:02,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0299, 4.7477, 4.9723, 5.2552, 5.4224, 4.7603, 5.3961, 5.3819], device='cuda:5'), covar=tensor([0.1503, 0.1220, 0.1704, 0.0677, 0.0530, 0.0744, 0.0492, 0.0574], device='cuda:5'), in_proj_covar=tensor([0.0519, 0.0633, 0.0770, 0.0640, 0.0486, 0.0487, 0.0510, 0.0573], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:13:03,071 INFO [train.py:904] (5/8) Epoch 10, batch 700, loss[loss=0.166, simple_loss=0.2582, pruned_loss=0.03692, over 17220.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.267, pruned_loss=0.05288, over 3225535.99 frames. ], batch size: 45, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:13:09,996 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7325, 2.7497, 2.4392, 3.9338, 3.3034, 4.0093, 1.4631, 2.8068], device='cuda:5'), covar=tensor([0.1319, 0.0551, 0.1063, 0.0138, 0.0194, 0.0356, 0.1397, 0.0766], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0152, 0.0177, 0.0131, 0.0189, 0.0208, 0.0176, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 04:14:09,230 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7734, 2.8596, 2.3484, 2.6138, 3.2333, 3.0296, 3.7858, 3.4923], device='cuda:5'), covar=tensor([0.0071, 0.0228, 0.0312, 0.0280, 0.0152, 0.0215, 0.0135, 0.0136], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0198, 0.0193, 0.0192, 0.0195, 0.0198, 0.0199, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:14:12,216 INFO [train.py:904] (5/8) Epoch 10, batch 750, loss[loss=0.2058, simple_loss=0.2862, pruned_loss=0.06277, over 16678.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.268, pruned_loss=0.05326, over 3244475.03 frames. ], batch size: 89, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:14:43,239 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5159, 4.4693, 4.4730, 3.3360, 4.4471, 1.6307, 4.1502, 4.1822], device='cuda:5'), covar=tensor([0.0144, 0.0144, 0.0182, 0.0654, 0.0134, 0.2970, 0.0197, 0.0318], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0110, 0.0161, 0.0150, 0.0130, 0.0180, 0.0147, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:14:52,015 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.724e+02 2.681e+02 3.065e+02 4.087e+02 6.685e+02, threshold=6.130e+02, percent-clipped=5.0 2023-04-29 04:15:22,695 INFO [train.py:904] (5/8) Epoch 10, batch 800, loss[loss=0.1991, simple_loss=0.2732, pruned_loss=0.0625, over 12256.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2681, pruned_loss=0.05328, over 3255614.94 frames. ], batch size: 246, lr: 7.01e-03, grad_scale: 4.0 2023-04-29 04:15:48,842 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-29 04:16:22,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8755, 3.8919, 2.9902, 2.3384, 2.6648, 2.3404, 3.9853, 3.6600], device='cuda:5'), covar=tensor([0.2143, 0.0551, 0.1392, 0.2243, 0.2088, 0.1671, 0.0496, 0.1046], device='cuda:5'), in_proj_covar=tensor([0.0295, 0.0253, 0.0277, 0.0268, 0.0267, 0.0214, 0.0262, 0.0286], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:16:27,645 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:16:32,818 INFO [train.py:904] (5/8) Epoch 10, batch 850, loss[loss=0.1766, simple_loss=0.2678, pruned_loss=0.04271, over 17047.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2673, pruned_loss=0.05263, over 3272108.68 frames. ], batch size: 50, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:17:10,142 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.453e+02 2.911e+02 3.766e+02 9.676e+02, threshold=5.821e+02, percent-clipped=2.0 2023-04-29 04:17:33,335 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:17:34,350 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:17:41,162 INFO [train.py:904] (5/8) Epoch 10, batch 900, loss[loss=0.1774, simple_loss=0.2583, pruned_loss=0.04822, over 16574.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2662, pruned_loss=0.05226, over 3278491.89 frames. ], batch size: 68, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:18:15,231 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0905, 4.1795, 2.2509, 4.7149, 2.9890, 4.7198, 2.4391, 3.2153], device='cuda:5'), covar=tensor([0.0197, 0.0286, 0.1561, 0.0164, 0.0752, 0.0346, 0.1432, 0.0642], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0165, 0.0189, 0.0124, 0.0167, 0.0204, 0.0193, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 04:18:50,838 INFO [train.py:904] (5/8) Epoch 10, batch 950, loss[loss=0.1827, simple_loss=0.2529, pruned_loss=0.05624, over 16741.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2662, pruned_loss=0.05272, over 3291289.03 frames. ], batch size: 134, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:19:04,207 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:19:27,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6398, 2.6563, 1.7942, 2.7437, 2.1891, 2.8116, 1.9921, 2.3479], device='cuda:5'), covar=tensor([0.0236, 0.0338, 0.1274, 0.0205, 0.0637, 0.0549, 0.1205, 0.0576], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0164, 0.0188, 0.0124, 0.0166, 0.0203, 0.0192, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 04:19:29,766 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.335e+02 2.688e+02 3.318e+02 6.424e+02, threshold=5.375e+02, percent-clipped=3.0 2023-04-29 04:19:39,563 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:19:58,792 INFO [train.py:904] (5/8) Epoch 10, batch 1000, loss[loss=0.185, simple_loss=0.2616, pruned_loss=0.05419, over 15536.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2644, pruned_loss=0.0523, over 3295712.07 frames. ], batch size: 190, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:20:47,707 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 04:20:59,964 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2937, 4.2458, 4.2436, 3.7208, 4.2819, 1.7700, 4.0063, 3.9305], device='cuda:5'), covar=tensor([0.0096, 0.0081, 0.0128, 0.0258, 0.0079, 0.2117, 0.0109, 0.0155], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0111, 0.0162, 0.0151, 0.0131, 0.0179, 0.0149, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:21:03,432 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8270, 3.8661, 2.1414, 4.2040, 2.7449, 4.1856, 2.1677, 2.9396], device='cuda:5'), covar=tensor([0.0181, 0.0330, 0.1554, 0.0210, 0.0779, 0.0492, 0.1534, 0.0648], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0163, 0.0186, 0.0123, 0.0164, 0.0202, 0.0191, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 04:21:09,130 INFO [train.py:904] (5/8) Epoch 10, batch 1050, loss[loss=0.1728, simple_loss=0.2477, pruned_loss=0.04893, over 16489.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2635, pruned_loss=0.05185, over 3292475.93 frames. ], batch size: 146, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:21:48,359 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.612e+02 3.041e+02 3.664e+02 7.677e+02, threshold=6.083e+02, percent-clipped=4.0 2023-04-29 04:22:18,526 INFO [train.py:904] (5/8) Epoch 10, batch 1100, loss[loss=0.1979, simple_loss=0.2697, pruned_loss=0.06301, over 16879.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2633, pruned_loss=0.05138, over 3301441.08 frames. ], batch size: 109, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:23:28,213 INFO [train.py:904] (5/8) Epoch 10, batch 1150, loss[loss=0.2133, simple_loss=0.2808, pruned_loss=0.07289, over 16725.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2627, pruned_loss=0.05038, over 3308014.37 frames. ], batch size: 124, lr: 6.99e-03, grad_scale: 4.0 2023-04-29 04:24:08,396 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.186e+02 2.590e+02 3.516e+02 6.121e+02, threshold=5.179e+02, percent-clipped=1.0 2023-04-29 04:24:32,775 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92547.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:24:38,835 INFO [train.py:904] (5/8) Epoch 10, batch 1200, loss[loss=0.1876, simple_loss=0.263, pruned_loss=0.05614, over 16202.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.262, pruned_loss=0.05014, over 3304759.05 frames. ], batch size: 165, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:25:06,522 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8549, 4.8979, 5.4760, 5.4607, 5.4285, 5.0222, 4.9657, 4.6937], device='cuda:5'), covar=tensor([0.0282, 0.0395, 0.0313, 0.0366, 0.0360, 0.0309, 0.0827, 0.0404], device='cuda:5'), in_proj_covar=tensor([0.0335, 0.0339, 0.0338, 0.0326, 0.0380, 0.0356, 0.0463, 0.0286], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 04:25:11,132 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9004, 4.3043, 4.4689, 3.2989, 3.8893, 4.2935, 4.0726, 2.7586], device='cuda:5'), covar=tensor([0.0318, 0.0040, 0.0021, 0.0223, 0.0065, 0.0057, 0.0047, 0.0310], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0070, 0.0068, 0.0125, 0.0077, 0.0084, 0.0075, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 04:25:39,103 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:25:47,777 INFO [train.py:904] (5/8) Epoch 10, batch 1250, loss[loss=0.1844, simple_loss=0.2668, pruned_loss=0.051, over 16568.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2625, pruned_loss=0.05045, over 3315377.95 frames. ], batch size: 62, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:26:01,769 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:21,274 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:27,275 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.554e+02 2.910e+02 3.549e+02 5.849e+02, threshold=5.821e+02, percent-clipped=4.0 2023-04-29 04:26:37,526 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:26:58,806 INFO [train.py:904] (5/8) Epoch 10, batch 1300, loss[loss=0.1828, simple_loss=0.269, pruned_loss=0.0483, over 17103.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2623, pruned_loss=0.0502, over 3311106.82 frames. ], batch size: 48, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:27:07,467 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:27:12,220 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3279, 5.0587, 5.2093, 5.5109, 5.6710, 4.9039, 5.5702, 5.5843], device='cuda:5'), covar=tensor([0.1161, 0.0872, 0.1455, 0.0558, 0.0385, 0.0649, 0.0487, 0.0515], device='cuda:5'), in_proj_covar=tensor([0.0533, 0.0650, 0.0799, 0.0664, 0.0498, 0.0503, 0.0524, 0.0593], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:27:44,071 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:27:45,927 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:07,569 INFO [train.py:904] (5/8) Epoch 10, batch 1350, loss[loss=0.1957, simple_loss=0.2641, pruned_loss=0.06364, over 12412.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2629, pruned_loss=0.05014, over 3310104.75 frames. ], batch size: 246, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:28:32,784 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5593, 4.5093, 4.6795, 4.5989, 4.6101, 5.1736, 4.7917, 4.4516], device='cuda:5'), covar=tensor([0.1451, 0.1946, 0.2039, 0.2246, 0.2667, 0.1133, 0.1470, 0.2438], device='cuda:5'), in_proj_covar=tensor([0.0343, 0.0497, 0.0524, 0.0430, 0.0570, 0.0548, 0.0423, 0.0573], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 04:28:34,025 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:47,047 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.580e+02 3.033e+02 3.471e+02 5.594e+02, threshold=6.066e+02, percent-clipped=0.0 2023-04-29 04:28:50,689 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8520, 3.0677, 2.7223, 4.6294, 3.9298, 4.4490, 1.5626, 3.2350], device='cuda:5'), covar=tensor([0.1290, 0.0562, 0.1040, 0.0148, 0.0262, 0.0332, 0.1458, 0.0701], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0154, 0.0177, 0.0135, 0.0195, 0.0211, 0.0177, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 04:29:18,974 INFO [train.py:904] (5/8) Epoch 10, batch 1400, loss[loss=0.1585, simple_loss=0.2398, pruned_loss=0.03859, over 16859.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2626, pruned_loss=0.04987, over 3313677.09 frames. ], batch size: 42, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:30:00,680 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:30:26,943 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4002, 5.7326, 5.4706, 5.5487, 5.1349, 5.1024, 5.2030, 5.8751], device='cuda:5'), covar=tensor([0.1187, 0.0902, 0.1062, 0.0668, 0.0842, 0.0649, 0.1023, 0.0911], device='cuda:5'), in_proj_covar=tensor([0.0534, 0.0682, 0.0562, 0.0468, 0.0428, 0.0438, 0.0569, 0.0516], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:30:28,857 INFO [train.py:904] (5/8) Epoch 10, batch 1450, loss[loss=0.1812, simple_loss=0.2589, pruned_loss=0.05172, over 17209.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2623, pruned_loss=0.04989, over 3320136.98 frames. ], batch size: 44, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:30:32,496 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 04:31:02,367 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9540, 4.7232, 4.8940, 5.2031, 5.3853, 4.6267, 5.3474, 5.2946], device='cuda:5'), covar=tensor([0.1435, 0.1044, 0.1725, 0.0660, 0.0475, 0.0838, 0.0476, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0538, 0.0660, 0.0809, 0.0674, 0.0503, 0.0507, 0.0528, 0.0597], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:31:08,004 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 2.441e+02 3.023e+02 3.567e+02 6.383e+02, threshold=6.046e+02, percent-clipped=1.0 2023-04-29 04:31:09,673 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:31:38,225 INFO [train.py:904] (5/8) Epoch 10, batch 1500, loss[loss=0.2055, simple_loss=0.2671, pruned_loss=0.07196, over 16830.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2624, pruned_loss=0.05045, over 3311649.90 frames. ], batch size: 116, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:32:08,014 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4163, 3.9947, 3.9922, 2.1569, 3.1091, 2.5655, 3.8814, 3.8553], device='cuda:5'), covar=tensor([0.0252, 0.0596, 0.0453, 0.1633, 0.0754, 0.0925, 0.0626, 0.1003], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0142, 0.0158, 0.0142, 0.0135, 0.0125, 0.0136, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 04:32:34,061 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:45,546 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:48,551 INFO [train.py:904] (5/8) Epoch 10, batch 1550, loss[loss=0.2258, simple_loss=0.2979, pruned_loss=0.0768, over 16383.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2637, pruned_loss=0.05229, over 3310550.75 frames. ], batch size: 145, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:33:26,144 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.944e+02 3.490e+02 4.097e+02 8.318e+02, threshold=6.980e+02, percent-clipped=5.0 2023-04-29 04:33:28,361 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:33:56,449 INFO [train.py:904] (5/8) Epoch 10, batch 1600, loss[loss=0.2074, simple_loss=0.2968, pruned_loss=0.059, over 16567.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2657, pruned_loss=0.05369, over 3315587.82 frames. ], batch size: 62, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:34:07,796 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:29,029 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:37,301 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:52,637 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:01,409 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5296, 1.5299, 2.0837, 2.4451, 2.5394, 2.5007, 1.6013, 2.7707], device='cuda:5'), covar=tensor([0.0126, 0.0336, 0.0217, 0.0176, 0.0166, 0.0160, 0.0353, 0.0068], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0167, 0.0153, 0.0156, 0.0164, 0.0120, 0.0169, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 04:35:06,245 INFO [train.py:904] (5/8) Epoch 10, batch 1650, loss[loss=0.1885, simple_loss=0.2776, pruned_loss=0.04969, over 17035.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.267, pruned_loss=0.05393, over 3317646.80 frames. ], batch size: 55, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:35:18,222 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:45,202 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.314e+02 2.824e+02 3.373e+02 5.658e+02, threshold=5.648e+02, percent-clipped=0.0 2023-04-29 04:35:53,067 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:53,114 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:14,149 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 04:36:15,641 INFO [train.py:904] (5/8) Epoch 10, batch 1700, loss[loss=0.1489, simple_loss=0.2285, pruned_loss=0.0346, over 16978.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2678, pruned_loss=0.05395, over 3322561.55 frames. ], batch size: 41, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:36:42,654 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:48,265 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:57,382 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5959, 2.2418, 2.3862, 4.4163, 2.0362, 2.6942, 2.3209, 2.4728], device='cuda:5'), covar=tensor([0.0827, 0.3000, 0.2050, 0.0334, 0.3728, 0.2053, 0.2764, 0.3037], device='cuda:5'), in_proj_covar=tensor([0.0358, 0.0382, 0.0324, 0.0322, 0.0408, 0.0435, 0.0344, 0.0451], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:37:17,842 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:37:23,104 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7380, 4.0667, 4.2581, 2.3027, 3.4938, 2.8513, 4.0481, 4.0247], device='cuda:5'), covar=tensor([0.0221, 0.0571, 0.0401, 0.1576, 0.0599, 0.0798, 0.0520, 0.0931], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0142, 0.0157, 0.0142, 0.0135, 0.0125, 0.0136, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 04:37:23,694 INFO [train.py:904] (5/8) Epoch 10, batch 1750, loss[loss=0.1822, simple_loss=0.2665, pruned_loss=0.04895, over 15914.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.269, pruned_loss=0.05369, over 3330996.46 frames. ], batch size: 35, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:37:42,827 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:38:01,414 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.469e+02 2.889e+02 3.646e+02 7.131e+02, threshold=5.778e+02, percent-clipped=4.0 2023-04-29 04:38:32,569 INFO [train.py:904] (5/8) Epoch 10, batch 1800, loss[loss=0.1612, simple_loss=0.2488, pruned_loss=0.0368, over 16825.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2703, pruned_loss=0.05378, over 3323582.67 frames. ], batch size: 42, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:39:06,696 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:39:10,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8833, 4.0687, 2.3453, 4.5498, 2.9847, 4.4991, 2.3328, 3.2821], device='cuda:5'), covar=tensor([0.0197, 0.0261, 0.1372, 0.0141, 0.0771, 0.0377, 0.1409, 0.0597], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0167, 0.0186, 0.0126, 0.0166, 0.0206, 0.0193, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 04:39:20,493 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93187.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:39:33,087 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 04:39:42,538 INFO [train.py:904] (5/8) Epoch 10, batch 1850, loss[loss=0.1927, simple_loss=0.282, pruned_loss=0.05169, over 16760.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2717, pruned_loss=0.05423, over 3329263.17 frames. ], batch size: 62, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:39:56,102 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5453, 3.7566, 3.9456, 1.9541, 4.1167, 4.1050, 3.1678, 2.9544], device='cuda:5'), covar=tensor([0.0744, 0.0145, 0.0137, 0.1091, 0.0055, 0.0148, 0.0362, 0.0389], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0099, 0.0087, 0.0141, 0.0070, 0.0101, 0.0122, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 04:40:21,087 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.500e+02 2.909e+02 3.446e+02 8.007e+02, threshold=5.817e+02, percent-clipped=2.0 2023-04-29 04:40:39,273 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3735, 4.3538, 4.2603, 3.8336, 4.3602, 1.7020, 4.1267, 4.0867], device='cuda:5'), covar=tensor([0.0090, 0.0077, 0.0138, 0.0266, 0.0073, 0.2216, 0.0111, 0.0153], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0116, 0.0168, 0.0158, 0.0136, 0.0181, 0.0154, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:40:52,070 INFO [train.py:904] (5/8) Epoch 10, batch 1900, loss[loss=0.2071, simple_loss=0.2877, pruned_loss=0.06328, over 16518.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2703, pruned_loss=0.05322, over 3329254.78 frames. ], batch size: 75, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:40:56,819 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:33,367 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:41,659 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:48,707 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6536, 4.8059, 4.9425, 4.8622, 4.7408, 5.4298, 5.0215, 4.7207], device='cuda:5'), covar=tensor([0.1331, 0.1823, 0.1924, 0.2128, 0.2993, 0.1224, 0.1421, 0.2621], device='cuda:5'), in_proj_covar=tensor([0.0345, 0.0495, 0.0519, 0.0429, 0.0570, 0.0548, 0.0420, 0.0572], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 04:42:02,736 INFO [train.py:904] (5/8) Epoch 10, batch 1950, loss[loss=0.178, simple_loss=0.2673, pruned_loss=0.04437, over 17118.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2707, pruned_loss=0.05277, over 3331144.40 frames. ], batch size: 47, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:42:22,821 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1018, 4.8045, 4.9887, 5.2673, 5.4907, 4.7673, 5.4670, 5.4281], device='cuda:5'), covar=tensor([0.1158, 0.0952, 0.1664, 0.0598, 0.0412, 0.0723, 0.0372, 0.0434], device='cuda:5'), in_proj_covar=tensor([0.0542, 0.0661, 0.0817, 0.0675, 0.0506, 0.0513, 0.0530, 0.0601], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:42:40,591 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:42:41,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.458e+02 3.015e+02 3.561e+02 8.313e+02, threshold=6.031e+02, percent-clipped=4.0 2023-04-29 04:42:43,603 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:12,566 INFO [train.py:904] (5/8) Epoch 10, batch 2000, loss[loss=0.1869, simple_loss=0.2629, pruned_loss=0.05547, over 16873.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2705, pruned_loss=0.05279, over 3330966.76 frames. ], batch size: 102, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:43:31,503 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:45,284 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:44:08,432 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 04:44:21,506 INFO [train.py:904] (5/8) Epoch 10, batch 2050, loss[loss=0.1861, simple_loss=0.2757, pruned_loss=0.04828, over 17225.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2703, pruned_loss=0.05322, over 3332661.51 frames. ], batch size: 46, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:44:23,774 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 04:44:51,438 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:45:00,362 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7958, 2.8842, 2.5240, 4.3005, 3.6226, 4.2512, 1.5432, 2.8912], device='cuda:5'), covar=tensor([0.1282, 0.0558, 0.1029, 0.0143, 0.0183, 0.0324, 0.1346, 0.0757], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0138, 0.0198, 0.0213, 0.0178, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 04:45:00,991 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.531e+02 2.855e+02 3.295e+02 5.942e+02, threshold=5.709e+02, percent-clipped=0.0 2023-04-29 04:45:29,946 INFO [train.py:904] (5/8) Epoch 10, batch 2100, loss[loss=0.1755, simple_loss=0.2732, pruned_loss=0.03894, over 17125.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2718, pruned_loss=0.05404, over 3328346.75 frames. ], batch size: 49, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:45:30,356 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8508, 2.8538, 2.2858, 2.5793, 3.0684, 2.8647, 3.6557, 3.4084], device='cuda:5'), covar=tensor([0.0061, 0.0225, 0.0324, 0.0290, 0.0183, 0.0244, 0.0163, 0.0161], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0201, 0.0197, 0.0196, 0.0199, 0.0200, 0.0209, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:45:56,833 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:08,414 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2784, 4.1231, 4.3041, 4.4842, 4.5824, 4.1187, 4.2945, 4.5516], device='cuda:5'), covar=tensor([0.1329, 0.0920, 0.1291, 0.0636, 0.0621, 0.1153, 0.1888, 0.0640], device='cuda:5'), in_proj_covar=tensor([0.0538, 0.0654, 0.0806, 0.0672, 0.0502, 0.0510, 0.0528, 0.0597], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:46:18,674 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:20,203 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 04:46:40,125 INFO [train.py:904] (5/8) Epoch 10, batch 2150, loss[loss=0.1716, simple_loss=0.2707, pruned_loss=0.03621, over 17046.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2731, pruned_loss=0.05442, over 3322095.32 frames. ], batch size: 50, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:47:14,161 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3863, 5.7391, 5.4749, 5.5839, 5.1363, 5.0053, 5.1866, 5.8678], device='cuda:5'), covar=tensor([0.0958, 0.0863, 0.0915, 0.0574, 0.0811, 0.0622, 0.0913, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0530, 0.0676, 0.0556, 0.0461, 0.0424, 0.0430, 0.0562, 0.0511], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:47:18,313 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.473e+02 3.061e+02 3.473e+02 5.653e+02, threshold=6.122e+02, percent-clipped=0.0 2023-04-29 04:47:24,633 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:47:46,706 INFO [train.py:904] (5/8) Epoch 10, batch 2200, loss[loss=0.1721, simple_loss=0.2606, pruned_loss=0.04184, over 17013.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2736, pruned_loss=0.05486, over 3326319.81 frames. ], batch size: 41, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:47:52,026 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:00,037 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 04:48:35,405 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:50,658 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9860, 2.5119, 1.9129, 2.3322, 2.9130, 2.7287, 3.1626, 3.1006], device='cuda:5'), covar=tensor([0.0107, 0.0234, 0.0358, 0.0292, 0.0158, 0.0222, 0.0159, 0.0148], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0202, 0.0198, 0.0199, 0.0200, 0.0202, 0.0211, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:48:54,028 INFO [train.py:904] (5/8) Epoch 10, batch 2250, loss[loss=0.2023, simple_loss=0.2756, pruned_loss=0.06448, over 16887.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2738, pruned_loss=0.05501, over 3328548.74 frames. ], batch size: 96, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:48:55,441 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:49:06,533 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 04:49:06,599 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 04:49:10,932 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3946, 2.4779, 1.9785, 2.3614, 2.8202, 2.6401, 3.3375, 3.0779], device='cuda:5'), covar=tensor([0.0088, 0.0268, 0.0369, 0.0294, 0.0177, 0.0251, 0.0172, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0202, 0.0198, 0.0199, 0.0200, 0.0201, 0.0210, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 04:49:33,832 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.560e+02 3.200e+02 3.958e+02 7.230e+02, threshold=6.400e+02, percent-clipped=2.0 2023-04-29 04:49:34,823 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93631.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:49:39,250 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:00,684 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 04:50:04,002 INFO [train.py:904] (5/8) Epoch 10, batch 2300, loss[loss=0.1678, simple_loss=0.2545, pruned_loss=0.04052, over 16829.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2743, pruned_loss=0.05476, over 3332426.77 frames. ], batch size: 42, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:50:22,493 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:39,724 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:49,492 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8439, 3.0600, 2.4613, 4.2891, 3.5989, 4.1986, 1.5137, 2.9806], device='cuda:5'), covar=tensor([0.1253, 0.0506, 0.1061, 0.0142, 0.0170, 0.0327, 0.1369, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0156, 0.0178, 0.0137, 0.0198, 0.0212, 0.0177, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 04:50:57,944 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:04,556 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 04:51:09,779 INFO [train.py:904] (5/8) Epoch 10, batch 2350, loss[loss=0.2224, simple_loss=0.2967, pruned_loss=0.07408, over 16810.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2746, pruned_loss=0.05522, over 3332375.09 frames. ], batch size: 102, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:51:27,740 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:49,911 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.363e+02 2.805e+02 3.305e+02 9.718e+02, threshold=5.610e+02, percent-clipped=1.0 2023-04-29 04:52:03,179 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:17,496 INFO [train.py:904] (5/8) Epoch 10, batch 2400, loss[loss=0.2188, simple_loss=0.286, pruned_loss=0.07584, over 16882.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2753, pruned_loss=0.05591, over 3336803.66 frames. ], batch size: 96, lr: 6.95e-03, grad_scale: 8.0 2023-04-29 04:52:41,474 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:43,113 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:49,842 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4654, 3.4531, 3.7124, 2.6042, 3.3471, 3.7734, 3.5941, 2.1592], device='cuda:5'), covar=tensor([0.0341, 0.0116, 0.0037, 0.0253, 0.0081, 0.0062, 0.0055, 0.0349], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0070, 0.0069, 0.0125, 0.0077, 0.0085, 0.0076, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 04:53:17,248 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 04:53:26,815 INFO [train.py:904] (5/8) Epoch 10, batch 2450, loss[loss=0.1944, simple_loss=0.2756, pruned_loss=0.0566, over 16786.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2769, pruned_loss=0.05646, over 3319804.26 frames. ], batch size: 102, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:53:49,967 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:54:05,074 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.505e+02 3.029e+02 3.825e+02 7.236e+02, threshold=6.058e+02, percent-clipped=4.0 2023-04-29 04:54:05,505 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:54:16,377 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1227, 3.0450, 3.1703, 1.7097, 3.3155, 3.3385, 2.6831, 2.4931], device='cuda:5'), covar=tensor([0.0743, 0.0179, 0.0171, 0.1057, 0.0078, 0.0161, 0.0415, 0.0441], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0098, 0.0087, 0.0138, 0.0070, 0.0100, 0.0121, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 04:54:34,543 INFO [train.py:904] (5/8) Epoch 10, batch 2500, loss[loss=0.1938, simple_loss=0.2721, pruned_loss=0.05769, over 16559.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2764, pruned_loss=0.05592, over 3326589.50 frames. ], batch size: 75, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:55:16,362 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 04:55:43,670 INFO [train.py:904] (5/8) Epoch 10, batch 2550, loss[loss=0.1993, simple_loss=0.284, pruned_loss=0.05732, over 16564.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2764, pruned_loss=0.05574, over 3330752.02 frames. ], batch size: 75, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:56:23,964 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.408e+02 2.927e+02 3.582e+02 7.180e+02, threshold=5.854e+02, percent-clipped=2.0 2023-04-29 04:56:52,860 INFO [train.py:904] (5/8) Epoch 10, batch 2600, loss[loss=0.1756, simple_loss=0.276, pruned_loss=0.03758, over 17265.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.276, pruned_loss=0.05558, over 3324412.93 frames. ], batch size: 52, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:57:32,347 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:58:03,853 INFO [train.py:904] (5/8) Epoch 10, batch 2650, loss[loss=0.2035, simple_loss=0.2931, pruned_loss=0.05691, over 16806.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2764, pruned_loss=0.05519, over 3325355.09 frames. ], batch size: 102, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:58:43,803 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.229e+02 2.744e+02 3.271e+02 8.724e+02, threshold=5.488e+02, percent-clipped=1.0 2023-04-29 04:59:00,269 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94042.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:59:13,565 INFO [train.py:904] (5/8) Epoch 10, batch 2700, loss[loss=0.1829, simple_loss=0.2702, pruned_loss=0.04782, over 17115.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2763, pruned_loss=0.05488, over 3321566.77 frames. ], batch size: 47, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:59:14,645 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 04:59:16,282 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 04:59:47,309 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 04:59:58,734 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8747, 4.6485, 4.9130, 5.1273, 5.3035, 4.6071, 5.2547, 5.2587], device='cuda:5'), covar=tensor([0.1354, 0.1062, 0.1430, 0.0552, 0.0437, 0.0838, 0.0549, 0.0458], device='cuda:5'), in_proj_covar=tensor([0.0540, 0.0661, 0.0812, 0.0670, 0.0499, 0.0519, 0.0530, 0.0599], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:00:01,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9977, 4.0860, 2.3930, 4.9137, 3.0074, 4.8044, 2.6993, 3.3820], device='cuda:5'), covar=tensor([0.0207, 0.0345, 0.1438, 0.0181, 0.0765, 0.0305, 0.1257, 0.0567], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0167, 0.0187, 0.0130, 0.0168, 0.0210, 0.0196, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 05:00:23,321 INFO [train.py:904] (5/8) Epoch 10, batch 2750, loss[loss=0.1773, simple_loss=0.2686, pruned_loss=0.04298, over 17192.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2759, pruned_loss=0.05407, over 3321213.25 frames. ], batch size: 45, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:00:55,375 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:01:00,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8502, 5.1725, 5.3269, 5.1201, 5.1382, 5.7458, 5.2566, 5.0225], device='cuda:5'), covar=tensor([0.0992, 0.1905, 0.1563, 0.1924, 0.2659, 0.0907, 0.1322, 0.2246], device='cuda:5'), in_proj_covar=tensor([0.0345, 0.0490, 0.0515, 0.0426, 0.0566, 0.0546, 0.0417, 0.0570], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 05:01:01,090 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.410e+02 2.958e+02 3.419e+02 5.641e+02, threshold=5.917e+02, percent-clipped=1.0 2023-04-29 05:01:12,752 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 2023-04-29 05:01:18,089 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0322, 3.3680, 3.1605, 2.1263, 2.8549, 2.3543, 3.5138, 3.5359], device='cuda:5'), covar=tensor([0.0216, 0.0677, 0.0582, 0.1455, 0.0697, 0.0861, 0.0477, 0.0785], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0144, 0.0157, 0.0142, 0.0135, 0.0124, 0.0136, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 05:01:29,688 INFO [train.py:904] (5/8) Epoch 10, batch 2800, loss[loss=0.2005, simple_loss=0.2796, pruned_loss=0.06071, over 16505.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2753, pruned_loss=0.05353, over 3327049.24 frames. ], batch size: 75, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:02:39,411 INFO [train.py:904] (5/8) Epoch 10, batch 2850, loss[loss=0.1828, simple_loss=0.2745, pruned_loss=0.04558, over 16719.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2748, pruned_loss=0.05326, over 3325705.20 frames. ], batch size: 57, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:03:09,782 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:03:20,118 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.429e+02 2.851e+02 3.333e+02 6.061e+02, threshold=5.703e+02, percent-clipped=1.0 2023-04-29 05:03:49,049 INFO [train.py:904] (5/8) Epoch 10, batch 2900, loss[loss=0.2205, simple_loss=0.2863, pruned_loss=0.07733, over 16187.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2738, pruned_loss=0.05345, over 3331874.97 frames. ], batch size: 165, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:03:56,170 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 05:04:33,877 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:04:36,104 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7181, 4.7088, 4.6411, 4.3571, 4.1608, 4.7330, 4.5405, 4.4029], device='cuda:5'), covar=tensor([0.0687, 0.0614, 0.0288, 0.0311, 0.1022, 0.0436, 0.0446, 0.0661], device='cuda:5'), in_proj_covar=tensor([0.0255, 0.0315, 0.0297, 0.0276, 0.0327, 0.0312, 0.0205, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 05:04:58,070 INFO [train.py:904] (5/8) Epoch 10, batch 2950, loss[loss=0.2715, simple_loss=0.3334, pruned_loss=0.1048, over 11948.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.273, pruned_loss=0.05379, over 3330706.51 frames. ], batch size: 246, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:05:39,548 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 2.713e+02 3.214e+02 3.938e+02 7.856e+02, threshold=6.427e+02, percent-clipped=3.0 2023-04-29 05:05:49,319 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:03,112 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:08,015 INFO [train.py:904] (5/8) Epoch 10, batch 3000, loss[loss=0.1777, simple_loss=0.2622, pruned_loss=0.04657, over 17011.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2727, pruned_loss=0.05372, over 3335078.61 frames. ], batch size: 41, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:06:08,015 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 05:06:17,140 INFO [train.py:938] (5/8) Epoch 10, validation: loss=0.1426, simple_loss=0.2488, pruned_loss=0.01818, over 944034.00 frames. 2023-04-29 05:06:17,141 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 05:06:34,630 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-29 05:07:26,699 INFO [train.py:904] (5/8) Epoch 10, batch 3050, loss[loss=0.2088, simple_loss=0.2946, pruned_loss=0.06156, over 16655.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2731, pruned_loss=0.05427, over 3336640.07 frames. ], batch size: 57, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:07:36,862 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:07:57,722 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94426.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:08:05,350 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.667e+02 3.252e+02 4.083e+02 7.974e+02, threshold=6.505e+02, percent-clipped=3.0 2023-04-29 05:08:12,080 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5521, 3.0365, 2.5577, 4.9118, 3.9926, 4.5145, 1.6934, 3.2415], device='cuda:5'), covar=tensor([0.1540, 0.0709, 0.1273, 0.0179, 0.0324, 0.0370, 0.1587, 0.0750], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0157, 0.0179, 0.0139, 0.0200, 0.0214, 0.0176, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 05:08:33,249 INFO [train.py:904] (5/8) Epoch 10, batch 3100, loss[loss=0.1865, simple_loss=0.257, pruned_loss=0.05799, over 16451.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2726, pruned_loss=0.0546, over 3334265.62 frames. ], batch size: 146, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:09:04,484 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:09:13,158 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6306, 4.6174, 4.5338, 3.9979, 4.5694, 1.6852, 4.3574, 4.3518], device='cuda:5'), covar=tensor([0.0087, 0.0060, 0.0125, 0.0295, 0.0073, 0.2276, 0.0112, 0.0158], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0118, 0.0166, 0.0160, 0.0137, 0.0178, 0.0155, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:09:43,508 INFO [train.py:904] (5/8) Epoch 10, batch 3150, loss[loss=0.1843, simple_loss=0.2729, pruned_loss=0.04787, over 17077.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2714, pruned_loss=0.05384, over 3334235.24 frames. ], batch size: 53, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:10:02,398 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9050, 4.3907, 3.2216, 2.2961, 2.9814, 2.5010, 4.6824, 4.0323], device='cuda:5'), covar=tensor([0.2416, 0.0550, 0.1407, 0.2138, 0.2261, 0.1662, 0.0314, 0.0841], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0256, 0.0279, 0.0273, 0.0281, 0.0219, 0.0266, 0.0298], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:10:15,375 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1748, 3.9658, 4.0891, 4.3800, 4.5213, 4.1243, 4.2938, 4.4712], device='cuda:5'), covar=tensor([0.1290, 0.1177, 0.1784, 0.0818, 0.0673, 0.1081, 0.1615, 0.0833], device='cuda:5'), in_proj_covar=tensor([0.0544, 0.0670, 0.0828, 0.0678, 0.0508, 0.0526, 0.0536, 0.0598], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:10:23,686 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.494e+02 3.047e+02 3.554e+02 8.509e+02, threshold=6.093e+02, percent-clipped=1.0 2023-04-29 05:10:52,178 INFO [train.py:904] (5/8) Epoch 10, batch 3200, loss[loss=0.1924, simple_loss=0.2816, pruned_loss=0.05157, over 17247.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2713, pruned_loss=0.05346, over 3332925.95 frames. ], batch size: 52, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:11:32,220 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:11:34,593 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4821, 4.4428, 4.9335, 4.9238, 4.9445, 4.6419, 4.6299, 4.4141], device='cuda:5'), covar=tensor([0.0343, 0.0610, 0.0426, 0.0446, 0.0444, 0.0346, 0.0800, 0.0483], device='cuda:5'), in_proj_covar=tensor([0.0345, 0.0352, 0.0357, 0.0336, 0.0396, 0.0371, 0.0479, 0.0296], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 05:12:04,577 INFO [train.py:904] (5/8) Epoch 10, batch 3250, loss[loss=0.222, simple_loss=0.2906, pruned_loss=0.07674, over 16416.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2715, pruned_loss=0.05396, over 3333030.63 frames. ], batch size: 146, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:12:44,903 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.346e+02 2.940e+02 3.507e+02 9.203e+02, threshold=5.881e+02, percent-clipped=1.0 2023-04-29 05:12:49,995 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 05:12:53,055 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:13:14,625 INFO [train.py:904] (5/8) Epoch 10, batch 3300, loss[loss=0.2272, simple_loss=0.3007, pruned_loss=0.07684, over 15541.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2728, pruned_loss=0.05445, over 3334613.95 frames. ], batch size: 190, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:02,319 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:14:19,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4185, 2.0559, 2.2525, 4.0084, 2.0462, 2.5431, 2.1198, 2.2910], device='cuda:5'), covar=tensor([0.0874, 0.3170, 0.1993, 0.0413, 0.3345, 0.2143, 0.3148, 0.2588], device='cuda:5'), in_proj_covar=tensor([0.0364, 0.0387, 0.0325, 0.0325, 0.0410, 0.0443, 0.0348, 0.0458], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:14:24,569 INFO [train.py:904] (5/8) Epoch 10, batch 3350, loss[loss=0.1846, simple_loss=0.2703, pruned_loss=0.04945, over 16559.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2735, pruned_loss=0.0542, over 3332320.55 frames. ], batch size: 75, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:28,543 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:15:05,007 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.488e+02 2.932e+02 3.912e+02 8.438e+02, threshold=5.863e+02, percent-clipped=4.0 2023-04-29 05:15:35,793 INFO [train.py:904] (5/8) Epoch 10, batch 3400, loss[loss=0.2178, simple_loss=0.2855, pruned_loss=0.0751, over 16904.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2746, pruned_loss=0.0546, over 3315989.65 frames. ], batch size: 109, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:16:26,725 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6597, 4.6725, 4.5671, 4.1361, 4.6329, 1.8499, 4.3780, 4.3944], device='cuda:5'), covar=tensor([0.0107, 0.0067, 0.0127, 0.0265, 0.0070, 0.2199, 0.0132, 0.0160], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0120, 0.0169, 0.0162, 0.0138, 0.0179, 0.0158, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:16:43,139 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0312, 1.8646, 2.3771, 2.8568, 2.7060, 3.2940, 2.1430, 3.1638], device='cuda:5'), covar=tensor([0.0160, 0.0334, 0.0228, 0.0202, 0.0203, 0.0128, 0.0307, 0.0130], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0172, 0.0158, 0.0163, 0.0167, 0.0124, 0.0171, 0.0115], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 05:16:44,912 INFO [train.py:904] (5/8) Epoch 10, batch 3450, loss[loss=0.2043, simple_loss=0.2926, pruned_loss=0.05805, over 17054.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2726, pruned_loss=0.05315, over 3327303.50 frames. ], batch size: 53, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:17:15,334 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8983, 5.3032, 4.9936, 5.0322, 4.7487, 4.6547, 4.7961, 5.3667], device='cuda:5'), covar=tensor([0.0972, 0.0840, 0.1008, 0.0641, 0.0772, 0.0911, 0.0958, 0.0911], device='cuda:5'), in_proj_covar=tensor([0.0539, 0.0689, 0.0563, 0.0469, 0.0432, 0.0437, 0.0568, 0.0523], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:17:26,308 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.564e+02 2.957e+02 3.532e+02 6.981e+02, threshold=5.915e+02, percent-clipped=3.0 2023-04-29 05:17:56,544 INFO [train.py:904] (5/8) Epoch 10, batch 3500, loss[loss=0.1402, simple_loss=0.2266, pruned_loss=0.02688, over 16943.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2713, pruned_loss=0.05256, over 3330715.98 frames. ], batch size: 41, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:18:23,165 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4207, 5.2873, 5.2587, 4.8982, 4.7426, 5.3019, 5.1639, 4.8576], device='cuda:5'), covar=tensor([0.0530, 0.0499, 0.0258, 0.0255, 0.1144, 0.0407, 0.0265, 0.0671], device='cuda:5'), in_proj_covar=tensor([0.0260, 0.0322, 0.0305, 0.0283, 0.0334, 0.0321, 0.0211, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 05:18:35,876 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:19:06,952 INFO [train.py:904] (5/8) Epoch 10, batch 3550, loss[loss=0.1942, simple_loss=0.2747, pruned_loss=0.05686, over 16409.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2697, pruned_loss=0.05242, over 3326813.94 frames. ], batch size: 68, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:19:42,074 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:19:47,570 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.230e+02 2.675e+02 3.251e+02 5.912e+02, threshold=5.350e+02, percent-clipped=0.0 2023-04-29 05:20:10,555 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 05:20:17,542 INFO [train.py:904] (5/8) Epoch 10, batch 3600, loss[loss=0.1586, simple_loss=0.2513, pruned_loss=0.033, over 17056.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2688, pruned_loss=0.05209, over 3328981.17 frames. ], batch size: 50, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:20:59,146 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 05:21:28,702 INFO [train.py:904] (5/8) Epoch 10, batch 3650, loss[loss=0.162, simple_loss=0.2362, pruned_loss=0.04387, over 16864.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2682, pruned_loss=0.05317, over 3309917.95 frames. ], batch size: 96, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:21:33,051 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:21:44,426 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6365, 1.6317, 2.2180, 2.5211, 2.5709, 2.4715, 1.6718, 2.6103], device='cuda:5'), covar=tensor([0.0113, 0.0308, 0.0214, 0.0149, 0.0161, 0.0182, 0.0326, 0.0113], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0172, 0.0158, 0.0162, 0.0169, 0.0125, 0.0172, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 05:22:10,216 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.260e+02 2.711e+02 3.366e+02 9.321e+02, threshold=5.423e+02, percent-clipped=5.0 2023-04-29 05:22:20,346 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7216, 2.7551, 2.4220, 4.0905, 3.4930, 4.2002, 1.6230, 2.7922], device='cuda:5'), covar=tensor([0.1343, 0.0651, 0.1113, 0.0179, 0.0235, 0.0327, 0.1396, 0.0786], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0157, 0.0177, 0.0141, 0.0201, 0.0212, 0.0175, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 05:22:23,355 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8304, 1.6472, 2.2428, 2.7204, 2.6250, 2.6235, 1.7499, 2.8518], device='cuda:5'), covar=tensor([0.0128, 0.0339, 0.0262, 0.0180, 0.0219, 0.0202, 0.0349, 0.0109], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0173, 0.0159, 0.0163, 0.0169, 0.0125, 0.0172, 0.0115], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 05:22:36,803 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5283, 3.5249, 3.7524, 2.0261, 3.8988, 3.9090, 3.0646, 2.9274], device='cuda:5'), covar=tensor([0.0693, 0.0174, 0.0147, 0.0963, 0.0066, 0.0111, 0.0359, 0.0373], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0100, 0.0088, 0.0138, 0.0070, 0.0102, 0.0121, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 05:22:43,032 INFO [train.py:904] (5/8) Epoch 10, batch 3700, loss[loss=0.2012, simple_loss=0.274, pruned_loss=0.06421, over 16765.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2666, pruned_loss=0.05503, over 3283915.65 frames. ], batch size: 134, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:22:43,416 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:23:56,108 INFO [train.py:904] (5/8) Epoch 10, batch 3750, loss[loss=0.2096, simple_loss=0.2653, pruned_loss=0.07696, over 16893.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2672, pruned_loss=0.05632, over 3284327.86 frames. ], batch size: 116, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:24:00,764 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:24:05,089 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 05:24:38,277 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.465e+02 2.786e+02 3.363e+02 5.307e+02, threshold=5.572e+02, percent-clipped=0.0 2023-04-29 05:25:07,907 INFO [train.py:904] (5/8) Epoch 10, batch 3800, loss[loss=0.1944, simple_loss=0.2643, pruned_loss=0.06229, over 16429.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2683, pruned_loss=0.05739, over 3276249.12 frames. ], batch size: 68, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:25:14,903 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-29 05:25:28,724 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:25:53,716 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8397, 4.7402, 4.7345, 4.5297, 4.4181, 4.7759, 4.6269, 4.5088], device='cuda:5'), covar=tensor([0.0527, 0.0574, 0.0237, 0.0245, 0.0853, 0.0393, 0.0328, 0.0532], device='cuda:5'), in_proj_covar=tensor([0.0249, 0.0310, 0.0294, 0.0273, 0.0320, 0.0309, 0.0201, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 05:26:02,038 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5270, 3.4907, 3.4338, 2.8357, 3.4321, 2.0142, 3.2475, 2.9998], device='cuda:5'), covar=tensor([0.0115, 0.0100, 0.0151, 0.0223, 0.0081, 0.2117, 0.0122, 0.0207], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0118, 0.0166, 0.0159, 0.0136, 0.0176, 0.0155, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:26:20,915 INFO [train.py:904] (5/8) Epoch 10, batch 3850, loss[loss=0.1836, simple_loss=0.2524, pruned_loss=0.05744, over 16535.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2688, pruned_loss=0.0581, over 3265425.57 frames. ], batch size: 75, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:26:22,348 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7525, 4.5447, 4.4215, 4.9640, 5.1378, 4.6045, 4.9670, 5.0488], device='cuda:5'), covar=tensor([0.1233, 0.1011, 0.2292, 0.0786, 0.0659, 0.0841, 0.0919, 0.0853], device='cuda:5'), in_proj_covar=tensor([0.0534, 0.0657, 0.0804, 0.0669, 0.0500, 0.0515, 0.0526, 0.0594], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:27:00,984 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 2.452e+02 2.920e+02 3.414e+02 5.310e+02, threshold=5.839e+02, percent-clipped=0.0 2023-04-29 05:27:15,669 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4417, 3.5123, 2.0245, 3.6617, 2.6965, 3.7076, 2.0112, 2.8554], device='cuda:5'), covar=tensor([0.0200, 0.0416, 0.1414, 0.0178, 0.0644, 0.0531, 0.1362, 0.0570], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0166, 0.0186, 0.0130, 0.0167, 0.0211, 0.0194, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 05:27:31,962 INFO [train.py:904] (5/8) Epoch 10, batch 3900, loss[loss=0.1951, simple_loss=0.2675, pruned_loss=0.06132, over 16461.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2689, pruned_loss=0.05878, over 3261913.37 frames. ], batch size: 146, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:27:43,955 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:27:56,187 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:28:07,430 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8677, 3.2063, 3.1423, 1.8765, 2.6772, 2.2311, 3.3920, 3.4316], device='cuda:5'), covar=tensor([0.0233, 0.0657, 0.0561, 0.1729, 0.0831, 0.0925, 0.0533, 0.0707], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0144, 0.0154, 0.0141, 0.0134, 0.0123, 0.0135, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 05:28:16,448 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 05:28:21,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9288, 4.0258, 4.3767, 4.3267, 4.3602, 4.0796, 4.0914, 3.9497], device='cuda:5'), covar=tensor([0.0406, 0.0652, 0.0363, 0.0434, 0.0480, 0.0407, 0.0729, 0.0548], device='cuda:5'), in_proj_covar=tensor([0.0339, 0.0345, 0.0346, 0.0330, 0.0392, 0.0368, 0.0473, 0.0290], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 05:28:42,073 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6460, 1.7582, 2.1843, 2.5414, 2.6437, 2.4758, 1.7023, 2.7393], device='cuda:5'), covar=tensor([0.0124, 0.0305, 0.0226, 0.0182, 0.0175, 0.0184, 0.0316, 0.0084], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0169, 0.0155, 0.0160, 0.0166, 0.0123, 0.0169, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 05:28:44,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0107, 1.8786, 2.4148, 2.9483, 2.9560, 3.3677, 1.9988, 3.1989], device='cuda:5'), covar=tensor([0.0134, 0.0338, 0.0226, 0.0202, 0.0179, 0.0098, 0.0325, 0.0074], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0169, 0.0155, 0.0160, 0.0166, 0.0123, 0.0169, 0.0112], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 05:28:45,299 INFO [train.py:904] (5/8) Epoch 10, batch 3950, loss[loss=0.1772, simple_loss=0.2571, pruned_loss=0.04867, over 16721.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2681, pruned_loss=0.05933, over 3270848.79 frames. ], batch size: 57, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:29:02,594 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7019, 4.5499, 4.7186, 4.8987, 5.0152, 4.4849, 4.8951, 4.9821], device='cuda:5'), covar=tensor([0.1413, 0.0982, 0.1475, 0.0571, 0.0553, 0.0864, 0.0831, 0.0589], device='cuda:5'), in_proj_covar=tensor([0.0527, 0.0649, 0.0795, 0.0658, 0.0492, 0.0507, 0.0516, 0.0587], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:29:12,996 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:22,316 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-04-29 05:29:23,121 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:25,662 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.393e+02 2.876e+02 3.494e+02 7.568e+02, threshold=5.751e+02, percent-clipped=4.0 2023-04-29 05:29:29,799 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 05:29:38,139 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95339.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:56,313 INFO [train.py:904] (5/8) Epoch 10, batch 4000, loss[loss=0.1977, simple_loss=0.275, pruned_loss=0.06024, over 16575.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2678, pruned_loss=0.05952, over 3266784.65 frames. ], batch size: 68, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:31:05,651 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95400.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:31:07,596 INFO [train.py:904] (5/8) Epoch 10, batch 4050, loss[loss=0.1776, simple_loss=0.2555, pruned_loss=0.0498, over 16641.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2678, pruned_loss=0.05824, over 3264767.33 frames. ], batch size: 57, lr: 6.89e-03, grad_scale: 16.0 2023-04-29 05:31:49,146 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 1.964e+02 2.335e+02 2.702e+02 4.319e+02, threshold=4.671e+02, percent-clipped=0.0 2023-04-29 05:32:20,031 INFO [train.py:904] (5/8) Epoch 10, batch 4100, loss[loss=0.1955, simple_loss=0.2814, pruned_loss=0.05475, over 16604.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2691, pruned_loss=0.05773, over 3247638.27 frames. ], batch size: 62, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:32:34,842 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 05:32:54,963 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-29 05:33:10,561 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 05:33:33,945 INFO [train.py:904] (5/8) Epoch 10, batch 4150, loss[loss=0.2449, simple_loss=0.3244, pruned_loss=0.0827, over 15486.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2773, pruned_loss=0.06103, over 3220839.94 frames. ], batch size: 191, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:34:17,114 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.709e+02 3.150e+02 3.936e+02 7.135e+02, threshold=6.300e+02, percent-clipped=10.0 2023-04-29 05:34:32,392 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5133, 2.2247, 2.2490, 4.2209, 1.9843, 2.6585, 2.3369, 2.4336], device='cuda:5'), covar=tensor([0.0879, 0.3236, 0.2072, 0.0417, 0.3899, 0.2129, 0.2721, 0.3171], device='cuda:5'), in_proj_covar=tensor([0.0362, 0.0391, 0.0324, 0.0324, 0.0406, 0.0447, 0.0350, 0.0461], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:34:49,632 INFO [train.py:904] (5/8) Epoch 10, batch 4200, loss[loss=0.2277, simple_loss=0.3202, pruned_loss=0.06761, over 16462.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2844, pruned_loss=0.06253, over 3202819.94 frames. ], batch size: 146, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:35:30,564 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4924, 4.5231, 4.9352, 4.8973, 4.9285, 4.5658, 4.5155, 4.4159], device='cuda:5'), covar=tensor([0.0269, 0.0352, 0.0328, 0.0334, 0.0304, 0.0308, 0.0872, 0.0392], device='cuda:5'), in_proj_covar=tensor([0.0323, 0.0329, 0.0332, 0.0316, 0.0376, 0.0353, 0.0454, 0.0280], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 05:36:04,071 INFO [train.py:904] (5/8) Epoch 10, batch 4250, loss[loss=0.1886, simple_loss=0.2763, pruned_loss=0.05043, over 17255.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2878, pruned_loss=0.06256, over 3192501.70 frames. ], batch size: 45, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:36:24,757 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:26,055 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:38,763 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:49,135 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.325e+02 2.809e+02 3.314e+02 5.860e+02, threshold=5.619e+02, percent-clipped=0.0 2023-04-29 05:37:19,473 INFO [train.py:904] (5/8) Epoch 10, batch 4300, loss[loss=0.2245, simple_loss=0.3227, pruned_loss=0.06316, over 16406.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2884, pruned_loss=0.06098, over 3197216.74 frames. ], batch size: 146, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:37:36,420 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1706, 3.3797, 3.5652, 3.5206, 3.5150, 3.3337, 3.3485, 3.4235], device='cuda:5'), covar=tensor([0.0385, 0.0556, 0.0408, 0.0476, 0.0543, 0.0444, 0.0816, 0.0467], device='cuda:5'), in_proj_covar=tensor([0.0325, 0.0330, 0.0332, 0.0317, 0.0377, 0.0352, 0.0456, 0.0280], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 05:37:59,690 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:24,132 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:33,672 INFO [train.py:904] (5/8) Epoch 10, batch 4350, loss[loss=0.2223, simple_loss=0.299, pruned_loss=0.07279, over 11694.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.292, pruned_loss=0.06225, over 3190167.60 frames. ], batch size: 247, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:39:18,686 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.523e+02 3.025e+02 3.679e+02 8.417e+02, threshold=6.050e+02, percent-clipped=3.0 2023-04-29 05:39:49,474 INFO [train.py:904] (5/8) Epoch 10, batch 4400, loss[loss=0.232, simple_loss=0.3173, pruned_loss=0.07334, over 16369.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2934, pruned_loss=0.06286, over 3214758.09 frames. ], batch size: 35, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:40:02,640 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:41:01,606 INFO [train.py:904] (5/8) Epoch 10, batch 4450, loss[loss=0.2394, simple_loss=0.3095, pruned_loss=0.08463, over 11888.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2965, pruned_loss=0.06337, over 3228248.41 frames. ], batch size: 247, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:41:07,092 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1583, 5.1674, 4.9712, 4.6606, 4.6462, 5.0524, 4.9116, 4.6612], device='cuda:5'), covar=tensor([0.0412, 0.0178, 0.0192, 0.0222, 0.0761, 0.0227, 0.0248, 0.0547], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0292, 0.0280, 0.0256, 0.0302, 0.0289, 0.0192, 0.0326], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:41:08,380 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4694, 4.5304, 4.2874, 4.0409, 3.9986, 4.4205, 4.1349, 4.0468], device='cuda:5'), covar=tensor([0.0415, 0.0218, 0.0230, 0.0251, 0.0750, 0.0259, 0.0454, 0.0577], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0292, 0.0280, 0.0256, 0.0302, 0.0289, 0.0191, 0.0326], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:41:12,080 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2757, 3.6776, 3.3634, 1.7836, 2.9698, 2.5281, 3.4228, 3.7490], device='cuda:5'), covar=tensor([0.0205, 0.0510, 0.0546, 0.1856, 0.0737, 0.0817, 0.0633, 0.0674], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0143, 0.0154, 0.0141, 0.0134, 0.0123, 0.0134, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 05:41:13,573 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:41:17,826 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:41:40,025 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9748, 3.2896, 3.1437, 2.0690, 2.9438, 3.2069, 3.0842, 1.8376], device='cuda:5'), covar=tensor([0.0404, 0.0027, 0.0039, 0.0336, 0.0065, 0.0068, 0.0057, 0.0344], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0066, 0.0066, 0.0121, 0.0076, 0.0083, 0.0073, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 05:41:46,153 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.254e+02 2.672e+02 3.297e+02 5.015e+02, threshold=5.344e+02, percent-clipped=0.0 2023-04-29 05:42:08,796 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:42:15,001 INFO [train.py:904] (5/8) Epoch 10, batch 4500, loss[loss=0.2173, simple_loss=0.3012, pruned_loss=0.06668, over 17251.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2963, pruned_loss=0.06362, over 3224635.41 frames. ], batch size: 45, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:42:44,170 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9952, 3.3712, 3.3725, 1.7485, 2.8242, 2.2072, 3.3712, 3.4493], device='cuda:5'), covar=tensor([0.0241, 0.0653, 0.0499, 0.1904, 0.0787, 0.0930, 0.0665, 0.0818], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0142, 0.0154, 0.0141, 0.0133, 0.0122, 0.0134, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 05:42:46,015 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:43:27,209 INFO [train.py:904] (5/8) Epoch 10, batch 4550, loss[loss=0.2138, simple_loss=0.3034, pruned_loss=0.0621, over 16734.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2968, pruned_loss=0.06435, over 3217862.43 frames. ], batch size: 124, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:43:35,571 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:43:47,799 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:43:59,545 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:44:10,341 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.125e+02 2.355e+02 2.832e+02 4.769e+02, threshold=4.710e+02, percent-clipped=0.0 2023-04-29 05:44:39,213 INFO [train.py:904] (5/8) Epoch 10, batch 4600, loss[loss=0.2051, simple_loss=0.2963, pruned_loss=0.05699, over 16867.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2971, pruned_loss=0.06389, over 3234624.41 frames. ], batch size: 96, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:44:57,860 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:09,575 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:12,091 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:20,966 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:43,331 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:55,187 INFO [train.py:904] (5/8) Epoch 10, batch 4650, loss[loss=0.2117, simple_loss=0.2925, pruned_loss=0.06545, over 16875.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2961, pruned_loss=0.06384, over 3228093.53 frames. ], batch size: 116, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:46:40,629 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 2.110e+02 2.403e+02 2.925e+02 6.538e+02, threshold=4.805e+02, percent-clipped=1.0 2023-04-29 05:46:52,344 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7275, 2.6246, 2.0301, 2.5605, 3.0274, 2.7487, 3.4076, 3.2867], device='cuda:5'), covar=tensor([0.0042, 0.0264, 0.0384, 0.0296, 0.0165, 0.0272, 0.0127, 0.0164], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0195, 0.0193, 0.0190, 0.0194, 0.0197, 0.0198, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:46:55,345 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96041.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:46:58,683 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:46:58,980 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5636, 2.1491, 2.3059, 4.3475, 2.0907, 2.5912, 2.2118, 2.4136], device='cuda:5'), covar=tensor([0.0850, 0.2881, 0.2045, 0.0342, 0.3702, 0.2168, 0.2670, 0.2936], device='cuda:5'), in_proj_covar=tensor([0.0358, 0.0385, 0.0318, 0.0318, 0.0407, 0.0441, 0.0346, 0.0452], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:47:10,307 INFO [train.py:904] (5/8) Epoch 10, batch 4700, loss[loss=0.2069, simple_loss=0.2837, pruned_loss=0.06502, over 16780.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2928, pruned_loss=0.06277, over 3212684.43 frames. ], batch size: 39, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:47:13,922 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4367, 3.6803, 3.5512, 1.9472, 3.0117, 2.3714, 3.7287, 3.7171], device='cuda:5'), covar=tensor([0.0212, 0.0604, 0.0531, 0.1817, 0.0783, 0.0882, 0.0621, 0.0900], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0143, 0.0155, 0.0142, 0.0134, 0.0123, 0.0134, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 05:47:52,297 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:48:19,070 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5444, 4.5486, 4.4392, 4.1276, 3.9997, 4.4649, 4.3000, 4.1549], device='cuda:5'), covar=tensor([0.0523, 0.0387, 0.0223, 0.0236, 0.0900, 0.0404, 0.0415, 0.0531], device='cuda:5'), in_proj_covar=tensor([0.0230, 0.0287, 0.0274, 0.0252, 0.0298, 0.0285, 0.0187, 0.0319], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:48:24,008 INFO [train.py:904] (5/8) Epoch 10, batch 4750, loss[loss=0.1822, simple_loss=0.2669, pruned_loss=0.04879, over 16706.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2893, pruned_loss=0.06107, over 3210139.53 frames. ], batch size: 57, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:08,953 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.162e+02 2.549e+02 3.399e+02 6.347e+02, threshold=5.097e+02, percent-clipped=5.0 2023-04-29 05:49:23,234 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96141.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:49:38,026 INFO [train.py:904] (5/8) Epoch 10, batch 4800, loss[loss=0.1828, simple_loss=0.2747, pruned_loss=0.04547, over 16465.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2861, pruned_loss=0.05942, over 3202156.69 frames. ], batch size: 75, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:58,728 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:50:03,020 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:50:29,800 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 05:50:54,668 INFO [train.py:904] (5/8) Epoch 10, batch 4850, loss[loss=0.2555, simple_loss=0.3223, pruned_loss=0.09432, over 12204.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2873, pruned_loss=0.05918, over 3175119.01 frames. ], batch size: 247, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:50:56,361 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:51:32,234 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96226.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:51:40,159 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.214e+02 2.693e+02 3.150e+02 7.967e+02, threshold=5.386e+02, percent-clipped=5.0 2023-04-29 05:52:10,086 INFO [train.py:904] (5/8) Epoch 10, batch 4900, loss[loss=0.1587, simple_loss=0.2592, pruned_loss=0.0291, over 16821.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2858, pruned_loss=0.05737, over 3182241.82 frames. ], batch size: 102, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:52:22,435 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7708, 3.7492, 4.1942, 1.9173, 4.4065, 4.4276, 3.0548, 3.1376], device='cuda:5'), covar=tensor([0.0696, 0.0175, 0.0098, 0.1108, 0.0034, 0.0052, 0.0334, 0.0423], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0098, 0.0084, 0.0136, 0.0068, 0.0096, 0.0118, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 05:52:42,909 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:24,562 INFO [train.py:904] (5/8) Epoch 10, batch 4950, loss[loss=0.2094, simple_loss=0.2884, pruned_loss=0.0652, over 16969.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2855, pruned_loss=0.05699, over 3189317.07 frames. ], batch size: 41, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:53:47,693 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6560, 3.6848, 2.8350, 2.1561, 2.6509, 2.2841, 3.8245, 3.4182], device='cuda:5'), covar=tensor([0.2505, 0.0749, 0.1596, 0.2194, 0.2160, 0.1642, 0.0567, 0.0933], device='cuda:5'), in_proj_covar=tensor([0.0300, 0.0256, 0.0278, 0.0273, 0.0281, 0.0216, 0.0266, 0.0291], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:53:52,126 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:56,333 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:05,870 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.230e+02 2.596e+02 3.100e+02 4.863e+02, threshold=5.193e+02, percent-clipped=0.0 2023-04-29 05:54:12,446 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:33,273 INFO [train.py:904] (5/8) Epoch 10, batch 5000, loss[loss=0.222, simple_loss=0.3031, pruned_loss=0.07049, over 11933.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2872, pruned_loss=0.05753, over 3177586.99 frames. ], batch size: 247, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:55:17,105 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0690, 1.8409, 1.8879, 3.6512, 1.7685, 2.2914, 1.9778, 1.9864], device='cuda:5'), covar=tensor([0.1052, 0.3390, 0.2277, 0.0450, 0.4095, 0.2225, 0.3054, 0.3060], device='cuda:5'), in_proj_covar=tensor([0.0356, 0.0385, 0.0319, 0.0319, 0.0405, 0.0438, 0.0346, 0.0451], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:55:21,035 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:27,988 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:44,810 INFO [train.py:904] (5/8) Epoch 10, batch 5050, loss[loss=0.2292, simple_loss=0.314, pruned_loss=0.0722, over 15474.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2878, pruned_loss=0.05738, over 3195672.19 frames. ], batch size: 191, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:56:22,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7588, 4.6551, 5.0822, 5.0531, 5.0799, 4.7903, 4.7590, 4.5709], device='cuda:5'), covar=tensor([0.0250, 0.0453, 0.0304, 0.0367, 0.0318, 0.0256, 0.0669, 0.0338], device='cuda:5'), in_proj_covar=tensor([0.0321, 0.0324, 0.0328, 0.0312, 0.0374, 0.0349, 0.0449, 0.0278], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 05:56:27,894 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.386e+02 2.799e+02 3.376e+02 6.474e+02, threshold=5.598e+02, percent-clipped=3.0 2023-04-29 05:56:33,458 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:55,164 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:55,810 INFO [train.py:904] (5/8) Epoch 10, batch 5100, loss[loss=0.1885, simple_loss=0.2706, pruned_loss=0.05326, over 17109.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2861, pruned_loss=0.05666, over 3195984.66 frames. ], batch size: 49, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:57:20,210 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:58:08,613 INFO [train.py:904] (5/8) Epoch 10, batch 5150, loss[loss=0.2089, simple_loss=0.3052, pruned_loss=0.05629, over 16746.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2859, pruned_loss=0.05549, over 3210276.65 frames. ], batch size: 83, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:58:11,206 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:58:30,252 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:58:37,351 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96521.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:58:52,023 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.088e+02 2.479e+02 2.937e+02 7.130e+02, threshold=4.958e+02, percent-clipped=1.0 2023-04-29 05:58:54,132 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1592, 4.0611, 4.0954, 3.4132, 4.1031, 1.6107, 3.8193, 3.8149], device='cuda:5'), covar=tensor([0.0088, 0.0085, 0.0116, 0.0362, 0.0072, 0.2460, 0.0125, 0.0189], device='cuda:5'), in_proj_covar=tensor([0.0121, 0.0110, 0.0155, 0.0152, 0.0127, 0.0170, 0.0144, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 05:59:16,085 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 05:59:21,948 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:59:22,746 INFO [train.py:904] (5/8) Epoch 10, batch 5200, loss[loss=0.196, simple_loss=0.2863, pruned_loss=0.05286, over 16314.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2851, pruned_loss=0.0555, over 3212021.01 frames. ], batch size: 165, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 06:00:35,342 INFO [train.py:904] (5/8) Epoch 10, batch 5250, loss[loss=0.1779, simple_loss=0.2691, pruned_loss=0.0434, over 16719.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2821, pruned_loss=0.0551, over 3203623.96 frames. ], batch size: 89, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:00:40,750 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2936, 3.7663, 3.5251, 2.0637, 3.0899, 2.5658, 3.6878, 3.7714], device='cuda:5'), covar=tensor([0.0205, 0.0546, 0.0538, 0.1611, 0.0699, 0.0786, 0.0521, 0.0714], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0142, 0.0157, 0.0141, 0.0134, 0.0123, 0.0135, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 06:01:17,130 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7528, 1.7910, 2.2627, 2.7440, 2.7799, 3.1337, 1.9928, 2.9363], device='cuda:5'), covar=tensor([0.0119, 0.0328, 0.0213, 0.0180, 0.0169, 0.0100, 0.0335, 0.0078], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0165, 0.0151, 0.0156, 0.0162, 0.0119, 0.0169, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 06:01:21,023 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.301e+02 2.656e+02 3.144e+02 5.435e+02, threshold=5.311e+02, percent-clipped=2.0 2023-04-29 06:01:26,437 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:01:48,995 INFO [train.py:904] (5/8) Epoch 10, batch 5300, loss[loss=0.1684, simple_loss=0.2498, pruned_loss=0.04353, over 16697.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2783, pruned_loss=0.0534, over 3213391.20 frames. ], batch size: 134, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:02:01,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2202, 3.3249, 1.7479, 3.5184, 2.4833, 3.5186, 2.0568, 2.6634], device='cuda:5'), covar=tensor([0.0203, 0.0279, 0.1668, 0.0113, 0.0763, 0.0393, 0.1510, 0.0684], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0162, 0.0187, 0.0119, 0.0166, 0.0201, 0.0191, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 06:02:30,168 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:02:36,967 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:03:03,552 INFO [train.py:904] (5/8) Epoch 10, batch 5350, loss[loss=0.2124, simple_loss=0.3058, pruned_loss=0.0595, over 16919.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2769, pruned_loss=0.05257, over 3223603.00 frames. ], batch size: 96, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:03:48,673 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.305e+02 2.858e+02 3.473e+02 6.088e+02, threshold=5.716e+02, percent-clipped=3.0 2023-04-29 06:03:53,376 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:04:03,726 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3151, 2.3951, 1.9309, 2.1800, 2.7766, 2.5304, 3.0142, 3.0462], device='cuda:5'), covar=tensor([0.0049, 0.0260, 0.0376, 0.0300, 0.0181, 0.0250, 0.0142, 0.0152], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0193, 0.0190, 0.0189, 0.0193, 0.0195, 0.0193, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:04:08,476 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96746.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:04:15,627 INFO [train.py:904] (5/8) Epoch 10, batch 5400, loss[loss=0.235, simple_loss=0.3134, pruned_loss=0.07829, over 12090.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2797, pruned_loss=0.05351, over 3213594.35 frames. ], batch size: 246, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:04:52,097 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:02,545 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:07,276 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3689, 2.9149, 2.6569, 2.2025, 2.2457, 2.1906, 2.8997, 2.8420], device='cuda:5'), covar=tensor([0.2152, 0.0683, 0.1198, 0.1868, 0.1873, 0.1643, 0.0498, 0.0932], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0255, 0.0277, 0.0272, 0.0278, 0.0215, 0.0266, 0.0290], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:05:31,413 INFO [train.py:904] (5/8) Epoch 10, batch 5450, loss[loss=0.2914, simple_loss=0.3671, pruned_loss=0.1078, over 15210.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2841, pruned_loss=0.05622, over 3193065.93 frames. ], batch size: 190, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:06:02,239 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:06:20,033 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.545e+02 3.377e+02 4.195e+02 1.255e+03, threshold=6.754e+02, percent-clipped=10.0 2023-04-29 06:06:27,124 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:06:49,550 INFO [train.py:904] (5/8) Epoch 10, batch 5500, loss[loss=0.2584, simple_loss=0.3301, pruned_loss=0.09337, over 15392.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2925, pruned_loss=0.06212, over 3165617.73 frames. ], batch size: 190, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:06:50,590 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3056, 4.6571, 4.3700, 4.4110, 4.1362, 4.1363, 4.2111, 4.6591], device='cuda:5'), covar=tensor([0.0892, 0.0761, 0.0990, 0.0672, 0.0744, 0.1224, 0.0914, 0.0787], device='cuda:5'), in_proj_covar=tensor([0.0500, 0.0635, 0.0528, 0.0438, 0.0400, 0.0410, 0.0527, 0.0486], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:07:17,928 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:08:08,503 INFO [train.py:904] (5/8) Epoch 10, batch 5550, loss[loss=0.2209, simple_loss=0.3087, pruned_loss=0.06652, over 16455.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3005, pruned_loss=0.06819, over 3130699.35 frames. ], batch size: 68, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:09:01,355 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.332e+02 3.697e+02 4.404e+02 5.267e+02 9.227e+02, threshold=8.809e+02, percent-clipped=8.0 2023-04-29 06:09:12,338 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2928, 1.5069, 1.9751, 2.1208, 2.3375, 2.5514, 1.6225, 2.4002], device='cuda:5'), covar=tensor([0.0163, 0.0326, 0.0202, 0.0220, 0.0184, 0.0105, 0.0332, 0.0077], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0167, 0.0153, 0.0157, 0.0164, 0.0119, 0.0170, 0.0110], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 06:09:12,722 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-29 06:09:17,936 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8967, 5.2150, 4.9295, 4.9652, 4.6243, 4.5666, 4.6490, 5.3081], device='cuda:5'), covar=tensor([0.1008, 0.0734, 0.1011, 0.0736, 0.0823, 0.0903, 0.0998, 0.0819], device='cuda:5'), in_proj_covar=tensor([0.0503, 0.0637, 0.0531, 0.0440, 0.0401, 0.0410, 0.0528, 0.0488], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:09:28,007 INFO [train.py:904] (5/8) Epoch 10, batch 5600, loss[loss=0.247, simple_loss=0.3163, pruned_loss=0.0889, over 16695.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.306, pruned_loss=0.07274, over 3094709.94 frames. ], batch size: 134, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:10:15,923 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:51,168 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:53,959 INFO [train.py:904] (5/8) Epoch 10, batch 5650, loss[loss=0.2819, simple_loss=0.3383, pruned_loss=0.1128, over 11506.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3109, pruned_loss=0.07661, over 3066681.65 frames. ], batch size: 246, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:11:15,938 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0990, 3.3398, 3.5241, 3.4746, 3.5000, 3.2952, 3.3258, 3.4057], device='cuda:5'), covar=tensor([0.0449, 0.0632, 0.0429, 0.0476, 0.0533, 0.0538, 0.0902, 0.0530], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0322, 0.0325, 0.0310, 0.0372, 0.0345, 0.0443, 0.0277], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 06:11:34,647 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97028.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:11:43,933 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.521e+02 3.909e+02 4.849e+02 5.778e+02 1.289e+03, threshold=9.698e+02, percent-clipped=3.0 2023-04-29 06:12:02,790 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:12:11,120 INFO [train.py:904] (5/8) Epoch 10, batch 5700, loss[loss=0.2456, simple_loss=0.3306, pruned_loss=0.08025, over 16838.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3134, pruned_loss=0.07941, over 3038960.48 frames. ], batch size: 116, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:12:25,398 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:13:15,320 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7323, 2.5717, 2.2869, 3.3388, 2.5549, 3.6278, 1.4746, 2.7365], device='cuda:5'), covar=tensor([0.1237, 0.0581, 0.1114, 0.0133, 0.0192, 0.0379, 0.1443, 0.0724], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0156, 0.0178, 0.0136, 0.0199, 0.0205, 0.0178, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 06:13:16,960 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:13:29,503 INFO [train.py:904] (5/8) Epoch 10, batch 5750, loss[loss=0.2425, simple_loss=0.3216, pruned_loss=0.08166, over 16839.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.316, pruned_loss=0.08119, over 3018759.22 frames. ], batch size: 42, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:13:54,864 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1515, 3.3341, 3.5345, 3.5037, 3.5112, 3.3022, 3.3527, 3.4009], device='cuda:5'), covar=tensor([0.0399, 0.0670, 0.0449, 0.0478, 0.0550, 0.0543, 0.0945, 0.0550], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0320, 0.0324, 0.0308, 0.0372, 0.0345, 0.0442, 0.0277], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 06:14:05,450 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 06:14:17,792 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:14:22,111 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.407e+02 3.370e+02 4.173e+02 4.986e+02 1.197e+03, threshold=8.346e+02, percent-clipped=2.0 2023-04-29 06:14:49,735 INFO [train.py:904] (5/8) Epoch 10, batch 5800, loss[loss=0.215, simple_loss=0.2909, pruned_loss=0.06954, over 16384.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3157, pruned_loss=0.08041, over 3004265.76 frames. ], batch size: 35, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:14:56,664 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-29 06:15:21,444 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2267, 3.6274, 3.5147, 1.9708, 3.1457, 2.4246, 3.6620, 3.7501], device='cuda:5'), covar=tensor([0.0206, 0.0578, 0.0521, 0.1754, 0.0640, 0.0830, 0.0499, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0141, 0.0156, 0.0141, 0.0133, 0.0124, 0.0134, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 06:16:07,878 INFO [train.py:904] (5/8) Epoch 10, batch 5850, loss[loss=0.211, simple_loss=0.292, pruned_loss=0.06498, over 16467.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3135, pruned_loss=0.07821, over 3015591.03 frames. ], batch size: 75, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:16:56,940 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-04-29 06:17:00,838 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.062e+02 2.965e+02 3.707e+02 4.622e+02 9.015e+02, threshold=7.415e+02, percent-clipped=1.0 2023-04-29 06:17:08,395 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6912, 4.4528, 4.3846, 4.9296, 4.9904, 4.5069, 4.9856, 5.0456], device='cuda:5'), covar=tensor([0.1383, 0.1031, 0.2306, 0.0670, 0.0826, 0.0954, 0.0799, 0.0733], device='cuda:5'), in_proj_covar=tensor([0.0499, 0.0620, 0.0751, 0.0629, 0.0480, 0.0483, 0.0497, 0.0561], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:17:28,944 INFO [train.py:904] (5/8) Epoch 10, batch 5900, loss[loss=0.2148, simple_loss=0.2963, pruned_loss=0.06668, over 16740.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3127, pruned_loss=0.07758, over 3032870.67 frames. ], batch size: 124, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:17:45,187 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:17:55,452 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3704, 4.6686, 4.4522, 4.4783, 4.1381, 4.1686, 4.2329, 4.6794], device='cuda:5'), covar=tensor([0.0850, 0.0802, 0.0834, 0.0651, 0.0748, 0.1302, 0.0923, 0.0880], device='cuda:5'), in_proj_covar=tensor([0.0503, 0.0635, 0.0530, 0.0441, 0.0399, 0.0411, 0.0525, 0.0484], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:17:58,277 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7992, 5.2049, 5.4279, 5.1220, 5.1755, 5.7924, 5.3163, 5.1379], device='cuda:5'), covar=tensor([0.1030, 0.1728, 0.1771, 0.1664, 0.2438, 0.0877, 0.1300, 0.2207], device='cuda:5'), in_proj_covar=tensor([0.0344, 0.0475, 0.0501, 0.0413, 0.0551, 0.0538, 0.0410, 0.0565], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 06:18:10,667 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7150, 4.7092, 4.5814, 4.2843, 4.1626, 4.6164, 4.4543, 4.2437], device='cuda:5'), covar=tensor([0.0608, 0.0548, 0.0270, 0.0280, 0.0983, 0.0461, 0.0478, 0.0759], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0293, 0.0277, 0.0253, 0.0299, 0.0290, 0.0187, 0.0321], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:18:49,538 INFO [train.py:904] (5/8) Epoch 10, batch 5950, loss[loss=0.1914, simple_loss=0.2838, pruned_loss=0.04952, over 16766.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3122, pruned_loss=0.07542, over 3045448.81 frames. ], batch size: 102, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:19:20,379 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:19:32,760 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7384, 4.0854, 3.0205, 2.3216, 3.1767, 2.4417, 4.1940, 4.0145], device='cuda:5'), covar=tensor([0.2832, 0.0747, 0.1659, 0.2059, 0.2151, 0.1726, 0.0553, 0.0842], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0257, 0.0281, 0.0277, 0.0281, 0.0218, 0.0267, 0.0291], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:19:41,584 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.897e+02 3.264e+02 4.239e+02 7.603e+02, threshold=6.529e+02, percent-clipped=1.0 2023-04-29 06:19:43,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1913, 3.3107, 3.5717, 3.5468, 3.5519, 3.3278, 3.3971, 3.4512], device='cuda:5'), covar=tensor([0.0386, 0.0624, 0.0411, 0.0444, 0.0492, 0.0487, 0.0813, 0.0468], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0324, 0.0327, 0.0314, 0.0377, 0.0350, 0.0449, 0.0282], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 06:19:52,934 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:01,456 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97347.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:09,122 INFO [train.py:904] (5/8) Epoch 10, batch 6000, loss[loss=0.2567, simple_loss=0.3117, pruned_loss=0.1009, over 11188.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3109, pruned_loss=0.07466, over 3040387.31 frames. ], batch size: 246, lr: 6.82e-03, grad_scale: 4.0 2023-04-29 06:20:09,123 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 06:20:23,729 INFO [train.py:938] (5/8) Epoch 10, validation: loss=0.165, simple_loss=0.2783, pruned_loss=0.02583, over 944034.00 frames. 2023-04-29 06:20:23,730 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 06:20:30,461 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:36,744 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2362, 3.3303, 3.6269, 1.5712, 3.8536, 3.9066, 2.8107, 2.7879], device='cuda:5'), covar=tensor([0.0845, 0.0210, 0.0144, 0.1289, 0.0056, 0.0100, 0.0385, 0.0440], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0097, 0.0085, 0.0138, 0.0068, 0.0096, 0.0120, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 06:20:53,627 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5434, 4.5182, 4.3834, 4.1624, 4.0169, 4.4395, 4.3291, 4.1036], device='cuda:5'), covar=tensor([0.0593, 0.0476, 0.0271, 0.0239, 0.0958, 0.0435, 0.0431, 0.0641], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0292, 0.0278, 0.0253, 0.0299, 0.0291, 0.0187, 0.0320], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:20:58,056 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:21:42,486 INFO [train.py:904] (5/8) Epoch 10, batch 6050, loss[loss=0.214, simple_loss=0.309, pruned_loss=0.05954, over 16635.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3093, pruned_loss=0.07341, over 3058665.12 frames. ], batch size: 134, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:21:45,511 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 06:21:52,782 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:16,337 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-04-29 06:22:30,487 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:35,036 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:35,675 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 3.012e+02 3.614e+02 4.338e+02 6.849e+02, threshold=7.229e+02, percent-clipped=1.0 2023-04-29 06:22:37,262 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 06:23:02,115 INFO [train.py:904] (5/8) Epoch 10, batch 6100, loss[loss=0.2118, simple_loss=0.3059, pruned_loss=0.05883, over 16898.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.308, pruned_loss=0.07166, over 3082019.00 frames. ], batch size: 96, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:23:49,436 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:24:23,714 INFO [train.py:904] (5/8) Epoch 10, batch 6150, loss[loss=0.2027, simple_loss=0.285, pruned_loss=0.06019, over 16613.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3062, pruned_loss=0.07109, over 3104712.31 frames. ], batch size: 134, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:25:17,523 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 3.229e+02 3.939e+02 5.020e+02 8.476e+02, threshold=7.879e+02, percent-clipped=2.0 2023-04-29 06:25:41,184 INFO [train.py:904] (5/8) Epoch 10, batch 6200, loss[loss=0.2016, simple_loss=0.2891, pruned_loss=0.0571, over 16653.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3056, pruned_loss=0.07189, over 3082620.06 frames. ], batch size: 89, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:25:46,702 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:26:57,944 INFO [train.py:904] (5/8) Epoch 10, batch 6250, loss[loss=0.2152, simple_loss=0.3111, pruned_loss=0.05967, over 16837.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3047, pruned_loss=0.0711, over 3100762.50 frames. ], batch size: 116, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:27:18,731 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:18,874 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:47,813 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 3.287e+02 3.846e+02 4.993e+02 1.247e+03, threshold=7.692e+02, percent-clipped=7.0 2023-04-29 06:28:02,025 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2960, 2.0441, 2.5674, 3.1894, 3.1264, 3.5876, 1.9963, 3.5229], device='cuda:5'), covar=tensor([0.0108, 0.0310, 0.0199, 0.0143, 0.0158, 0.0085, 0.0351, 0.0071], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0164, 0.0148, 0.0153, 0.0160, 0.0117, 0.0168, 0.0109], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 06:28:11,802 INFO [train.py:904] (5/8) Epoch 10, batch 6300, loss[loss=0.2092, simple_loss=0.2949, pruned_loss=0.06177, over 16932.00 frames. ], tot_loss[loss=0.223, simple_loss=0.305, pruned_loss=0.07053, over 3102038.77 frames. ], batch size: 109, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:28:18,640 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:28:43,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4218, 2.1643, 1.7558, 1.8801, 2.4772, 2.1813, 2.5064, 2.6440], device='cuda:5'), covar=tensor([0.0115, 0.0251, 0.0359, 0.0321, 0.0146, 0.0243, 0.0176, 0.0166], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0196, 0.0193, 0.0194, 0.0194, 0.0199, 0.0200, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:28:49,001 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:29:24,493 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:29:30,249 INFO [train.py:904] (5/8) Epoch 10, batch 6350, loss[loss=0.2715, simple_loss=0.3301, pruned_loss=0.1065, over 11463.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3055, pruned_loss=0.07165, over 3086388.64 frames. ], batch size: 246, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:29:31,958 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:29:33,084 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:29:47,738 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2725, 4.0365, 4.1368, 4.4694, 4.5351, 4.2196, 4.5469, 4.5758], device='cuda:5'), covar=tensor([0.1493, 0.1134, 0.1754, 0.0722, 0.0802, 0.1147, 0.0827, 0.0729], device='cuda:5'), in_proj_covar=tensor([0.0504, 0.0622, 0.0756, 0.0635, 0.0485, 0.0482, 0.0504, 0.0564], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:30:13,655 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:30:22,193 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 06:30:22,850 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 3.422e+02 4.156e+02 5.083e+02 8.318e+02, threshold=8.312e+02, percent-clipped=1.0 2023-04-29 06:30:46,252 INFO [train.py:904] (5/8) Epoch 10, batch 6400, loss[loss=0.3128, simple_loss=0.362, pruned_loss=0.1318, over 11296.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3055, pruned_loss=0.07288, over 3071835.36 frames. ], batch size: 248, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:32:00,797 INFO [train.py:904] (5/8) Epoch 10, batch 6450, loss[loss=0.2247, simple_loss=0.3038, pruned_loss=0.07275, over 16897.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3047, pruned_loss=0.07155, over 3081588.54 frames. ], batch size: 116, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:32:57,259 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.110e+02 2.867e+02 3.528e+02 4.342e+02 1.041e+03, threshold=7.056e+02, percent-clipped=1.0 2023-04-29 06:33:18,171 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6659, 4.6466, 4.5306, 4.2390, 4.1501, 4.5646, 4.4645, 4.2261], device='cuda:5'), covar=tensor([0.0553, 0.0518, 0.0243, 0.0266, 0.0891, 0.0436, 0.0404, 0.0644], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0294, 0.0275, 0.0254, 0.0298, 0.0287, 0.0187, 0.0320], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:33:21,811 INFO [train.py:904] (5/8) Epoch 10, batch 6500, loss[loss=0.225, simple_loss=0.3072, pruned_loss=0.07142, over 16872.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3031, pruned_loss=0.07101, over 3081740.42 frames. ], batch size: 116, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:33:44,410 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 06:34:17,550 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:34:41,160 INFO [train.py:904] (5/8) Epoch 10, batch 6550, loss[loss=0.2873, simple_loss=0.3429, pruned_loss=0.1158, over 11450.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.306, pruned_loss=0.07178, over 3088336.09 frames. ], batch size: 247, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:34:55,679 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:04,961 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:35,380 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 3.236e+02 3.952e+02 5.112e+02 9.277e+02, threshold=7.905e+02, percent-clipped=2.0 2023-04-29 06:35:56,912 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:36:00,724 INFO [train.py:904] (5/8) Epoch 10, batch 6600, loss[loss=0.228, simple_loss=0.3076, pruned_loss=0.07419, over 15294.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3083, pruned_loss=0.07238, over 3090489.18 frames. ], batch size: 190, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:36:19,455 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:37:14,853 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:37:22,483 INFO [train.py:904] (5/8) Epoch 10, batch 6650, loss[loss=0.28, simple_loss=0.3406, pruned_loss=0.1097, over 11454.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3096, pruned_loss=0.07404, over 3068223.75 frames. ], batch size: 248, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:37:24,770 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:05,410 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:38:05,431 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98030.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:14,895 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 3.281e+02 3.856e+02 4.909e+02 8.809e+02, threshold=7.713e+02, percent-clipped=1.0 2023-04-29 06:38:31,003 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:39,002 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:39,801 INFO [train.py:904] (5/8) Epoch 10, batch 6700, loss[loss=0.2818, simple_loss=0.3359, pruned_loss=0.1139, over 11312.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3072, pruned_loss=0.07265, over 3095801.71 frames. ], batch size: 247, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:39:14,417 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1810, 5.1295, 5.0010, 4.7228, 4.6306, 5.0363, 5.0546, 4.6784], device='cuda:5'), covar=tensor([0.0502, 0.0291, 0.0212, 0.0219, 0.0874, 0.0318, 0.0232, 0.0644], device='cuda:5'), in_proj_covar=tensor([0.0233, 0.0292, 0.0274, 0.0252, 0.0298, 0.0285, 0.0187, 0.0318], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:39:20,942 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:39:57,938 INFO [train.py:904] (5/8) Epoch 10, batch 6750, loss[loss=0.1985, simple_loss=0.2862, pruned_loss=0.05542, over 16455.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3058, pruned_loss=0.07219, over 3104285.94 frames. ], batch size: 68, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:40:04,377 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:40:04,626 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-29 06:40:49,812 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.169e+02 3.584e+02 4.478e+02 5.468e+02 7.454e+02, threshold=8.956e+02, percent-clipped=0.0 2023-04-29 06:41:15,042 INFO [train.py:904] (5/8) Epoch 10, batch 6800, loss[loss=0.226, simple_loss=0.3063, pruned_loss=0.07284, over 17122.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3057, pruned_loss=0.0723, over 3104604.65 frames. ], batch size: 48, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:41:39,116 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98167.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:42:33,970 INFO [train.py:904] (5/8) Epoch 10, batch 6850, loss[loss=0.2191, simple_loss=0.325, pruned_loss=0.05654, over 16732.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3068, pruned_loss=0.07297, over 3100036.42 frames. ], batch size: 76, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:42:39,933 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7918, 3.8533, 4.2925, 2.0874, 4.5251, 4.5040, 3.4564, 3.4037], device='cuda:5'), covar=tensor([0.0635, 0.0166, 0.0125, 0.1050, 0.0034, 0.0068, 0.0243, 0.0385], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0099, 0.0087, 0.0139, 0.0068, 0.0097, 0.0121, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 06:42:47,592 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:42:48,264 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 06:43:24,608 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.168e+02 2.932e+02 3.586e+02 4.673e+02 8.284e+02, threshold=7.173e+02, percent-clipped=0.0 2023-04-29 06:43:37,315 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:43:49,046 INFO [train.py:904] (5/8) Epoch 10, batch 6900, loss[loss=0.3041, simple_loss=0.3495, pruned_loss=0.1294, over 11834.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3094, pruned_loss=0.07277, over 3106749.36 frames. ], batch size: 246, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:44:01,235 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:44:33,051 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 06:45:07,306 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 06:45:09,157 INFO [train.py:904] (5/8) Epoch 10, batch 6950, loss[loss=0.2699, simple_loss=0.3305, pruned_loss=0.1046, over 11511.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3113, pruned_loss=0.07464, over 3074367.55 frames. ], batch size: 248, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:45:54,217 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 06:46:01,762 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 3.204e+02 3.949e+02 4.845e+02 8.099e+02, threshold=7.898e+02, percent-clipped=3.0 2023-04-29 06:46:12,656 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8413, 4.1407, 3.9306, 3.9915, 3.6254, 3.7828, 3.8555, 4.1256], device='cuda:5'), covar=tensor([0.1066, 0.0892, 0.0959, 0.0701, 0.0759, 0.1576, 0.0897, 0.0923], device='cuda:5'), in_proj_covar=tensor([0.0512, 0.0644, 0.0536, 0.0447, 0.0402, 0.0419, 0.0539, 0.0487], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:46:23,209 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5246, 3.5311, 3.4683, 2.8894, 3.4415, 2.0720, 3.2444, 2.8423], device='cuda:5'), covar=tensor([0.0108, 0.0082, 0.0131, 0.0218, 0.0080, 0.1901, 0.0100, 0.0161], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0109, 0.0156, 0.0151, 0.0127, 0.0173, 0.0143, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:46:27,413 INFO [train.py:904] (5/8) Epoch 10, batch 7000, loss[loss=0.2449, simple_loss=0.3085, pruned_loss=0.0907, over 11709.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3118, pruned_loss=0.07445, over 3057242.49 frames. ], batch size: 246, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:46:40,906 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 06:47:08,045 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:47:43,145 INFO [train.py:904] (5/8) Epoch 10, batch 7050, loss[loss=0.2134, simple_loss=0.3004, pruned_loss=0.06321, over 16401.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3119, pruned_loss=0.07352, over 3075767.97 frames. ], batch size: 146, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:47:57,278 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:48:34,411 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 3.125e+02 3.922e+02 4.883e+02 1.063e+03, threshold=7.844e+02, percent-clipped=4.0 2023-04-29 06:48:59,562 INFO [train.py:904] (5/8) Epoch 10, batch 7100, loss[loss=0.2036, simple_loss=0.2936, pruned_loss=0.05684, over 17196.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3099, pruned_loss=0.07336, over 3067344.69 frames. ], batch size: 44, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:49:14,098 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:49:28,001 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:50:12,821 INFO [train.py:904] (5/8) Epoch 10, batch 7150, loss[loss=0.2374, simple_loss=0.3158, pruned_loss=0.07944, over 16447.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3085, pruned_loss=0.07302, over 3079208.13 frames. ], batch size: 146, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:51:03,787 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.197e+02 3.117e+02 3.720e+02 4.761e+02 8.896e+02, threshold=7.439e+02, percent-clipped=3.0 2023-04-29 06:51:15,664 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:16,733 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:29,370 INFO [train.py:904] (5/8) Epoch 10, batch 7200, loss[loss=0.2031, simple_loss=0.2814, pruned_loss=0.06238, over 11559.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3066, pruned_loss=0.07185, over 3054545.12 frames. ], batch size: 248, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:52:01,032 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 06:52:04,572 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 06:52:32,648 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:52:49,207 INFO [train.py:904] (5/8) Epoch 10, batch 7250, loss[loss=0.1926, simple_loss=0.2791, pruned_loss=0.05301, over 16809.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3041, pruned_loss=0.07029, over 3074133.56 frames. ], batch size: 102, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:52:50,536 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2840, 2.4490, 1.6738, 1.8696, 2.8969, 2.4411, 3.2355, 3.1745], device='cuda:5'), covar=tensor([0.0067, 0.0294, 0.0478, 0.0412, 0.0170, 0.0312, 0.0146, 0.0148], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0194, 0.0193, 0.0192, 0.0193, 0.0197, 0.0197, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:52:53,627 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:52:59,361 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2118, 1.9405, 1.5762, 1.6570, 2.2664, 1.9233, 2.1600, 2.3681], device='cuda:5'), covar=tensor([0.0103, 0.0250, 0.0348, 0.0343, 0.0162, 0.0267, 0.0147, 0.0159], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0194, 0.0193, 0.0192, 0.0193, 0.0197, 0.0197, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:53:45,151 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 3.052e+02 3.513e+02 4.472e+02 1.147e+03, threshold=7.026e+02, percent-clipped=3.0 2023-04-29 06:54:05,201 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8687, 4.8931, 4.6072, 4.4349, 4.3225, 4.7469, 4.6501, 4.4137], device='cuda:5'), covar=tensor([0.0518, 0.0286, 0.0266, 0.0229, 0.0915, 0.0344, 0.0333, 0.0560], device='cuda:5'), in_proj_covar=tensor([0.0232, 0.0290, 0.0270, 0.0248, 0.0294, 0.0285, 0.0186, 0.0314], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:54:05,891 INFO [train.py:904] (5/8) Epoch 10, batch 7300, loss[loss=0.2087, simple_loss=0.2976, pruned_loss=0.05994, over 16690.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3034, pruned_loss=0.06975, over 3083621.60 frames. ], batch size: 76, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:54:25,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8984, 2.0571, 2.3493, 3.0853, 2.1469, 2.2678, 2.2762, 2.1174], device='cuda:5'), covar=tensor([0.0815, 0.2660, 0.1638, 0.0538, 0.3179, 0.1842, 0.2306, 0.2883], device='cuda:5'), in_proj_covar=tensor([0.0355, 0.0382, 0.0320, 0.0316, 0.0408, 0.0434, 0.0344, 0.0449], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 06:54:42,383 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 06:54:58,477 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7018, 3.5876, 3.8349, 3.6391, 3.7381, 4.1607, 3.8231, 3.5765], device='cuda:5'), covar=tensor([0.2104, 0.2283, 0.1932, 0.2248, 0.2602, 0.1506, 0.1480, 0.2547], device='cuda:5'), in_proj_covar=tensor([0.0340, 0.0473, 0.0506, 0.0409, 0.0542, 0.0536, 0.0415, 0.0562], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 06:55:23,883 INFO [train.py:904] (5/8) Epoch 10, batch 7350, loss[loss=0.2364, simple_loss=0.2982, pruned_loss=0.08726, over 11423.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3039, pruned_loss=0.07036, over 3074206.50 frames. ], batch size: 249, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:19,127 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.271e+02 3.294e+02 3.832e+02 4.622e+02 6.263e+02, threshold=7.664e+02, percent-clipped=0.0 2023-04-29 06:56:23,866 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6682, 3.6917, 4.0905, 4.0693, 4.0380, 3.7761, 3.8039, 3.7606], device='cuda:5'), covar=tensor([0.0350, 0.0637, 0.0392, 0.0414, 0.0497, 0.0426, 0.0875, 0.0537], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0319, 0.0321, 0.0307, 0.0369, 0.0341, 0.0442, 0.0276], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 06:56:41,011 INFO [train.py:904] (5/8) Epoch 10, batch 7400, loss[loss=0.2535, simple_loss=0.3267, pruned_loss=0.09013, over 16718.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3051, pruned_loss=0.071, over 3089760.48 frames. ], batch size: 124, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:57,610 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:57:03,110 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:57:59,100 INFO [train.py:904] (5/8) Epoch 10, batch 7450, loss[loss=0.221, simple_loss=0.2946, pruned_loss=0.07373, over 17034.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3061, pruned_loss=0.07234, over 3069673.04 frames. ], batch size: 55, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:58:14,257 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:58:33,584 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 06:58:57,292 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 3.303e+02 3.942e+02 4.934e+02 1.157e+03, threshold=7.883e+02, percent-clipped=3.0 2023-04-29 06:59:03,848 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 06:59:15,899 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-04-29 06:59:16,034 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-29 06:59:19,908 INFO [train.py:904] (5/8) Epoch 10, batch 7500, loss[loss=0.2053, simple_loss=0.2951, pruned_loss=0.05773, over 16681.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3066, pruned_loss=0.07238, over 3049631.53 frames. ], batch size: 89, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:59:26,722 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-29 07:00:20,573 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:27,952 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:34,240 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:38,286 INFO [train.py:904] (5/8) Epoch 10, batch 7550, loss[loss=0.2263, simple_loss=0.3024, pruned_loss=0.07511, over 15230.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3054, pruned_loss=0.07223, over 3038325.37 frames. ], batch size: 190, lr: 6.76e-03, grad_scale: 2.0 2023-04-29 07:01:32,292 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.425e+02 3.280e+02 3.927e+02 5.001e+02 7.830e+02, threshold=7.854e+02, percent-clipped=0.0 2023-04-29 07:01:41,666 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8221, 1.7358, 1.5358, 1.5043, 1.8288, 1.6022, 1.7153, 1.8906], device='cuda:5'), covar=tensor([0.0095, 0.0206, 0.0305, 0.0268, 0.0152, 0.0193, 0.0145, 0.0143], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0194, 0.0192, 0.0191, 0.0193, 0.0196, 0.0196, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:01:53,314 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:01:53,982 INFO [train.py:904] (5/8) Epoch 10, batch 7600, loss[loss=0.2325, simple_loss=0.3082, pruned_loss=0.07843, over 15245.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3049, pruned_loss=0.07255, over 3028385.11 frames. ], batch size: 191, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:02:01,607 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:02:24,352 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98971.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:03:11,990 INFO [train.py:904] (5/8) Epoch 10, batch 7650, loss[loss=0.2766, simple_loss=0.3331, pruned_loss=0.11, over 11534.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.305, pruned_loss=0.07311, over 3024126.21 frames. ], batch size: 246, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:03:12,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2633, 5.5381, 5.2300, 5.3020, 4.9462, 4.8492, 5.0793, 5.6145], device='cuda:5'), covar=tensor([0.0966, 0.0739, 0.0944, 0.0669, 0.0706, 0.0688, 0.0883, 0.0740], device='cuda:5'), in_proj_covar=tensor([0.0513, 0.0638, 0.0535, 0.0443, 0.0399, 0.0423, 0.0534, 0.0485], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:03:26,763 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9816, 4.0775, 4.4447, 2.1068, 4.8136, 4.7834, 3.1644, 3.7087], device='cuda:5'), covar=tensor([0.0699, 0.0173, 0.0188, 0.1104, 0.0032, 0.0073, 0.0380, 0.0325], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0099, 0.0085, 0.0137, 0.0068, 0.0096, 0.0120, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 07:03:44,994 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7392, 2.6774, 2.4526, 4.1799, 3.1057, 4.0948, 1.4990, 2.9327], device='cuda:5'), covar=tensor([0.1231, 0.0669, 0.1081, 0.0123, 0.0226, 0.0325, 0.1442, 0.0706], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0136, 0.0202, 0.0207, 0.0179, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 07:03:59,255 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:04:08,575 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.981e+02 3.374e+02 4.235e+02 5.349e+02 9.496e+02, threshold=8.471e+02, percent-clipped=6.0 2023-04-29 07:04:28,266 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:04:29,013 INFO [train.py:904] (5/8) Epoch 10, batch 7700, loss[loss=0.2818, simple_loss=0.3323, pruned_loss=0.1157, over 11546.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3051, pruned_loss=0.07334, over 3041698.17 frames. ], batch size: 248, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:04:50,190 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:05:32,288 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2878, 4.2844, 4.1768, 3.4852, 4.2630, 1.5817, 3.9548, 3.8720], device='cuda:5'), covar=tensor([0.0087, 0.0074, 0.0138, 0.0346, 0.0074, 0.2583, 0.0127, 0.0202], device='cuda:5'), in_proj_covar=tensor([0.0122, 0.0107, 0.0155, 0.0150, 0.0126, 0.0173, 0.0142, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:05:43,697 INFO [train.py:904] (5/8) Epoch 10, batch 7750, loss[loss=0.2299, simple_loss=0.3187, pruned_loss=0.07056, over 16776.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3057, pruned_loss=0.07276, over 3065674.56 frames. ], batch size: 83, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:05:51,266 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5516, 2.2375, 2.2595, 4.1507, 2.0890, 2.6305, 2.3246, 2.4304], device='cuda:5'), covar=tensor([0.0878, 0.2902, 0.2137, 0.0384, 0.3730, 0.2023, 0.2713, 0.2777], device='cuda:5'), in_proj_covar=tensor([0.0355, 0.0385, 0.0321, 0.0319, 0.0413, 0.0436, 0.0346, 0.0450], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:05:52,475 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7645, 2.2817, 2.2866, 4.4711, 2.1520, 2.7213, 2.3866, 2.5558], device='cuda:5'), covar=tensor([0.0800, 0.2936, 0.2115, 0.0335, 0.3552, 0.2027, 0.2728, 0.2726], device='cuda:5'), in_proj_covar=tensor([0.0355, 0.0385, 0.0321, 0.0319, 0.0413, 0.0436, 0.0346, 0.0450], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:06:00,271 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99112.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:06:02,532 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:06:14,924 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4851, 2.1020, 2.3467, 4.2364, 2.0587, 2.5706, 2.2517, 2.3978], device='cuda:5'), covar=tensor([0.0876, 0.3219, 0.1997, 0.0348, 0.3683, 0.2052, 0.2842, 0.2810], device='cuda:5'), in_proj_covar=tensor([0.0353, 0.0385, 0.0320, 0.0318, 0.0411, 0.0435, 0.0345, 0.0448], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:06:38,059 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.181e+02 3.321e+02 3.885e+02 5.264e+02 1.269e+03, threshold=7.770e+02, percent-clipped=1.0 2023-04-29 07:06:59,345 INFO [train.py:904] (5/8) Epoch 10, batch 7800, loss[loss=0.2846, simple_loss=0.3395, pruned_loss=0.1149, over 11448.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3064, pruned_loss=0.07333, over 3083718.71 frames. ], batch size: 247, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:08:06,905 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8689, 2.6193, 2.5446, 1.9660, 2.4336, 2.5163, 2.5863, 1.8268], device='cuda:5'), covar=tensor([0.0326, 0.0052, 0.0058, 0.0257, 0.0080, 0.0079, 0.0067, 0.0320], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0065, 0.0068, 0.0125, 0.0076, 0.0087, 0.0074, 0.0118], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 07:08:12,878 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:08:16,573 INFO [train.py:904] (5/8) Epoch 10, batch 7850, loss[loss=0.25, simple_loss=0.316, pruned_loss=0.09203, over 11497.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3069, pruned_loss=0.07301, over 3079316.54 frames. ], batch size: 248, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:08:57,400 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:10,226 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.079e+02 2.953e+02 3.751e+02 4.670e+02 9.934e+02, threshold=7.502e+02, percent-clipped=3.0 2023-04-29 07:09:20,695 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8571, 3.1305, 3.1112, 2.0457, 2.8541, 3.1035, 3.0225, 1.8456], device='cuda:5'), covar=tensor([0.0445, 0.0042, 0.0046, 0.0337, 0.0090, 0.0089, 0.0070, 0.0360], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0065, 0.0069, 0.0125, 0.0076, 0.0087, 0.0074, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 07:09:22,401 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:24,278 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:29,741 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:30,448 INFO [train.py:904] (5/8) Epoch 10, batch 7900, loss[loss=0.213, simple_loss=0.3034, pruned_loss=0.06133, over 16465.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3063, pruned_loss=0.07257, over 3078365.18 frames. ], batch size: 146, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:09:55,617 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:10:31,507 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99290.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:10:49,031 INFO [train.py:904] (5/8) Epoch 10, batch 7950, loss[loss=0.216, simple_loss=0.296, pruned_loss=0.068, over 16327.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3075, pruned_loss=0.07337, over 3077636.78 frames. ], batch size: 165, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:11:22,531 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:27,243 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:30,535 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:43,789 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 3.138e+02 3.909e+02 4.608e+02 7.916e+02, threshold=7.818e+02, percent-clipped=1.0 2023-04-29 07:11:52,060 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5301, 5.8553, 5.5734, 5.6361, 5.1153, 5.1008, 5.3093, 6.0019], device='cuda:5'), covar=tensor([0.0923, 0.0709, 0.1020, 0.0662, 0.0899, 0.0698, 0.0953, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0512, 0.0641, 0.0538, 0.0447, 0.0401, 0.0425, 0.0539, 0.0488], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:12:06,034 INFO [train.py:904] (5/8) Epoch 10, batch 8000, loss[loss=0.2695, simple_loss=0.3316, pruned_loss=0.1037, over 11613.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3073, pruned_loss=0.07337, over 3079951.20 frames. ], batch size: 247, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:12:32,996 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5191, 4.5788, 5.0210, 4.9738, 4.9815, 4.6507, 4.6385, 4.3712], device='cuda:5'), covar=tensor([0.0291, 0.0404, 0.0286, 0.0386, 0.0399, 0.0319, 0.0847, 0.0457], device='cuda:5'), in_proj_covar=tensor([0.0316, 0.0323, 0.0325, 0.0311, 0.0373, 0.0343, 0.0443, 0.0281], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 07:12:57,225 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99385.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:13:21,646 INFO [train.py:904] (5/8) Epoch 10, batch 8050, loss[loss=0.2266, simple_loss=0.3081, pruned_loss=0.07251, over 15382.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.308, pruned_loss=0.07369, over 3071609.13 frames. ], batch size: 190, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:13:29,908 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:13:40,889 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 07:13:50,474 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:14:18,157 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.046e+02 3.114e+02 3.682e+02 4.495e+02 9.204e+02, threshold=7.364e+02, percent-clipped=1.0 2023-04-29 07:14:39,583 INFO [train.py:904] (5/8) Epoch 10, batch 8100, loss[loss=0.213, simple_loss=0.2991, pruned_loss=0.0635, over 16367.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3074, pruned_loss=0.073, over 3075064.04 frames. ], batch size: 146, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:15:03,937 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:15:23,077 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99481.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:15:54,078 INFO [train.py:904] (5/8) Epoch 10, batch 8150, loss[loss=0.2183, simple_loss=0.2943, pruned_loss=0.07116, over 15279.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3045, pruned_loss=0.07178, over 3069416.02 frames. ], batch size: 190, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:15:59,901 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1443, 3.5737, 3.6712, 2.3253, 3.3978, 3.6236, 3.5206, 1.9511], device='cuda:5'), covar=tensor([0.0420, 0.0042, 0.0036, 0.0334, 0.0070, 0.0093, 0.0059, 0.0367], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0065, 0.0068, 0.0124, 0.0075, 0.0087, 0.0074, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 07:16:35,943 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:16:49,695 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.193e+02 2.958e+02 3.724e+02 4.411e+02 7.985e+02, threshold=7.447e+02, percent-clipped=3.0 2023-04-29 07:17:01,424 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:08,958 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:09,699 INFO [train.py:904] (5/8) Epoch 10, batch 8200, loss[loss=0.2017, simple_loss=0.2876, pruned_loss=0.05789, over 16389.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3021, pruned_loss=0.07159, over 3057850.09 frames. ], batch size: 146, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:17:36,146 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:04,739 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:11,892 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99589.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:20,196 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:28,354 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99599.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:32,982 INFO [train.py:904] (5/8) Epoch 10, batch 8250, loss[loss=0.179, simple_loss=0.2643, pruned_loss=0.04691, over 12150.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.301, pruned_loss=0.06924, over 3036820.77 frames. ], batch size: 248, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:19:10,203 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:16,595 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:19,882 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:36,139 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.844e+02 3.459e+02 4.111e+02 7.773e+02, threshold=6.919e+02, percent-clipped=2.0 2023-04-29 07:19:55,483 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:57,956 INFO [train.py:904] (5/8) Epoch 10, batch 8300, loss[loss=0.2064, simple_loss=0.3024, pruned_loss=0.05523, over 16862.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2982, pruned_loss=0.06614, over 3021212.20 frames. ], batch size: 116, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:20:37,233 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,718 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,874 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:21:21,641 INFO [train.py:904] (5/8) Epoch 10, batch 8350, loss[loss=0.2209, simple_loss=0.3028, pruned_loss=0.06952, over 16794.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2968, pruned_loss=0.06359, over 3045395.43 frames. ], batch size: 116, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:21:30,484 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:22:18,498 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7862, 3.7418, 4.1131, 4.1054, 4.0820, 3.8520, 3.8344, 3.8044], device='cuda:5'), covar=tensor([0.0310, 0.0656, 0.0382, 0.0395, 0.0430, 0.0395, 0.0906, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0312, 0.0321, 0.0319, 0.0306, 0.0369, 0.0340, 0.0438, 0.0277], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 07:22:21,851 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.332e+02 2.884e+02 3.614e+02 8.033e+02, threshold=5.769e+02, percent-clipped=2.0 2023-04-29 07:22:26,797 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:22:43,915 INFO [train.py:904] (5/8) Epoch 10, batch 8400, loss[loss=0.1935, simple_loss=0.2876, pruned_loss=0.04967, over 16698.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2935, pruned_loss=0.06061, over 3054433.22 frames. ], batch size: 124, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:22:49,991 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99755.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:23:25,435 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:24:06,261 INFO [train.py:904] (5/8) Epoch 10, batch 8450, loss[loss=0.1969, simple_loss=0.2735, pruned_loss=0.06012, over 12154.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2917, pruned_loss=0.05907, over 3053125.79 frames. ], batch size: 248, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:24:28,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5681, 3.5066, 3.6645, 1.9425, 3.8746, 3.9276, 3.0113, 3.1424], device='cuda:5'), covar=tensor([0.0568, 0.0173, 0.0189, 0.0944, 0.0040, 0.0082, 0.0341, 0.0285], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0096, 0.0081, 0.0132, 0.0065, 0.0093, 0.0115, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-29 07:24:42,769 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:25:06,073 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.551e+02 3.005e+02 3.693e+02 6.066e+02, threshold=6.011e+02, percent-clipped=2.0 2023-04-29 07:25:25,948 INFO [train.py:904] (5/8) Epoch 10, batch 8500, loss[loss=0.1797, simple_loss=0.2571, pruned_loss=0.05115, over 12055.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2875, pruned_loss=0.05615, over 3055374.50 frames. ], batch size: 246, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:26:18,968 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 07:26:21,611 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99885.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:26:50,587 INFO [train.py:904] (5/8) Epoch 10, batch 8550, loss[loss=0.1782, simple_loss=0.2604, pruned_loss=0.04797, over 11839.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2846, pruned_loss=0.05498, over 3040779.97 frames. ], batch size: 246, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:27:20,475 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3930, 2.9692, 3.0984, 1.7596, 2.6564, 2.2070, 2.9568, 3.1040], device='cuda:5'), covar=tensor([0.0313, 0.0702, 0.0461, 0.1813, 0.0778, 0.0899, 0.0734, 0.0830], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0135, 0.0150, 0.0137, 0.0130, 0.0121, 0.0131, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 07:27:34,896 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:35,031 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:50,549 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:02,246 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.540e+02 3.329e+02 3.854e+02 7.309e+02, threshold=6.657e+02, percent-clipped=4.0 2023-04-29 07:28:16,660 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:22,293 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8015, 1.3096, 1.5382, 1.7544, 1.7941, 1.8037, 1.5690, 1.8069], device='cuda:5'), covar=tensor([0.0147, 0.0276, 0.0161, 0.0182, 0.0183, 0.0153, 0.0270, 0.0083], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0166, 0.0148, 0.0151, 0.0161, 0.0115, 0.0166, 0.0105], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 07:28:29,561 INFO [train.py:904] (5/8) Epoch 10, batch 8600, loss[loss=0.1995, simple_loss=0.295, pruned_loss=0.052, over 16497.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.285, pruned_loss=0.05406, over 3027862.59 frames. ], batch size: 147, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:28:45,705 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6147, 2.8110, 2.3210, 3.7027, 2.3335, 3.8963, 1.4065, 2.6585], device='cuda:5'), covar=tensor([0.1401, 0.0553, 0.1123, 0.0107, 0.0100, 0.0332, 0.1570, 0.0852], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0155, 0.0177, 0.0133, 0.0194, 0.0203, 0.0178, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 07:29:07,017 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2203, 3.5933, 3.7982, 2.0966, 3.0884, 2.4906, 3.5584, 3.7952], device='cuda:5'), covar=tensor([0.0232, 0.0622, 0.0407, 0.1605, 0.0668, 0.0831, 0.0611, 0.0798], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0135, 0.0150, 0.0136, 0.0130, 0.0121, 0.0131, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 07:29:10,991 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:29:26,473 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:29:35,466 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6283, 1.6128, 1.9707, 2.6378, 2.4358, 2.7781, 1.8927, 2.8606], device='cuda:5'), covar=tensor([0.0129, 0.0393, 0.0270, 0.0189, 0.0216, 0.0127, 0.0330, 0.0093], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0166, 0.0149, 0.0151, 0.0161, 0.0115, 0.0167, 0.0106], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 07:30:11,027 INFO [train.py:904] (5/8) Epoch 10, batch 8650, loss[loss=0.1824, simple_loss=0.2794, pruned_loss=0.04274, over 16644.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2828, pruned_loss=0.05207, over 3039183.64 frames. ], batch size: 134, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:30:24,831 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:30:46,526 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1350, 3.9558, 4.2038, 4.3617, 4.4457, 4.0414, 4.4036, 4.4403], device='cuda:5'), covar=tensor([0.1253, 0.0913, 0.1309, 0.0577, 0.0493, 0.1129, 0.0570, 0.0486], device='cuda:5'), in_proj_covar=tensor([0.0480, 0.0598, 0.0722, 0.0616, 0.0466, 0.0467, 0.0486, 0.0549], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:31:10,594 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:31:27,937 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:31:31,867 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.482e+02 2.900e+02 4.130e+02 1.018e+03, threshold=5.800e+02, percent-clipped=4.0 2023-04-29 07:31:55,799 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0485, 4.0555, 3.9349, 3.4032, 3.9603, 1.7863, 3.7956, 3.6195], device='cuda:5'), covar=tensor([0.0083, 0.0072, 0.0120, 0.0199, 0.0069, 0.2221, 0.0093, 0.0160], device='cuda:5'), in_proj_covar=tensor([0.0119, 0.0106, 0.0150, 0.0142, 0.0123, 0.0170, 0.0138, 0.0139], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:31:56,447 INFO [train.py:904] (5/8) Epoch 10, batch 8700, loss[loss=0.1776, simple_loss=0.2701, pruned_loss=0.04255, over 15352.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2794, pruned_loss=0.05052, over 3030162.77 frames. ], batch size: 191, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:32:19,390 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.42 vs. limit=5.0 2023-04-29 07:32:28,596 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:32:43,351 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:33:35,361 INFO [train.py:904] (5/8) Epoch 10, batch 8750, loss[loss=0.1686, simple_loss=0.2546, pruned_loss=0.04128, over 12124.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2786, pruned_loss=0.04976, over 3021861.71 frames. ], batch size: 248, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:34:32,196 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:34:32,244 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:35:02,704 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.530e+02 3.072e+02 4.161e+02 6.852e+02, threshold=6.144e+02, percent-clipped=8.0 2023-04-29 07:35:29,751 INFO [train.py:904] (5/8) Epoch 10, batch 8800, loss[loss=0.2019, simple_loss=0.2865, pruned_loss=0.05866, over 12784.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2775, pruned_loss=0.04896, over 3030652.07 frames. ], batch size: 249, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:36:12,286 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:37:15,891 INFO [train.py:904] (5/8) Epoch 10, batch 8850, loss[loss=0.1898, simple_loss=0.2911, pruned_loss=0.04423, over 16625.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2807, pruned_loss=0.04842, over 3042238.85 frames. ], batch size: 62, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:38:04,286 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:34,814 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:36,937 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 2.435e+02 3.067e+02 3.598e+02 6.365e+02, threshold=6.134e+02, percent-clipped=1.0 2023-04-29 07:38:48,384 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100245.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:50,415 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3210, 4.6330, 4.4605, 4.4375, 4.1158, 4.1146, 4.1486, 4.6755], device='cuda:5'), covar=tensor([0.0883, 0.0861, 0.0875, 0.0581, 0.0735, 0.1294, 0.0902, 0.0815], device='cuda:5'), in_proj_covar=tensor([0.0486, 0.0613, 0.0503, 0.0423, 0.0384, 0.0405, 0.0513, 0.0467], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:39:01,678 INFO [train.py:904] (5/8) Epoch 10, batch 8900, loss[loss=0.2206, simple_loss=0.3094, pruned_loss=0.06592, over 16342.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2813, pruned_loss=0.04829, over 3046198.59 frames. ], batch size: 146, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:39:43,458 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:40:45,779 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:01,145 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:05,939 INFO [train.py:904] (5/8) Epoch 10, batch 8950, loss[loss=0.1744, simple_loss=0.2652, pruned_loss=0.04179, over 17017.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2806, pruned_loss=0.0481, over 3071450.98 frames. ], batch size: 116, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:42:22,251 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:42:29,138 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.569e+02 2.987e+02 3.833e+02 6.749e+02, threshold=5.974e+02, percent-clipped=2.0 2023-04-29 07:42:55,086 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2388, 5.2523, 5.0648, 4.7030, 4.6070, 5.0796, 5.0358, 4.7015], device='cuda:5'), covar=tensor([0.0524, 0.0503, 0.0232, 0.0244, 0.0968, 0.0537, 0.0197, 0.0678], device='cuda:5'), in_proj_covar=tensor([0.0225, 0.0282, 0.0263, 0.0244, 0.0285, 0.0279, 0.0182, 0.0306], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:42:55,812 INFO [train.py:904] (5/8) Epoch 10, batch 9000, loss[loss=0.1577, simple_loss=0.2441, pruned_loss=0.03566, over 17088.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.277, pruned_loss=0.04626, over 3092017.66 frames. ], batch size: 53, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:42:55,812 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 07:43:05,470 INFO [train.py:938] (5/8) Epoch 10, validation: loss=0.1565, simple_loss=0.2604, pruned_loss=0.02634, over 944034.00 frames. 2023-04-29 07:43:05,470 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 07:43:18,848 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0870, 3.2499, 3.5179, 1.6797, 3.6732, 3.7610, 2.8253, 2.6829], device='cuda:5'), covar=tensor([0.0815, 0.0200, 0.0123, 0.1130, 0.0049, 0.0095, 0.0344, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0097, 0.0081, 0.0135, 0.0066, 0.0094, 0.0116, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:5') 2023-04-29 07:43:30,280 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:44:12,990 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:44:50,486 INFO [train.py:904] (5/8) Epoch 10, batch 9050, loss[loss=0.1937, simple_loss=0.2743, pruned_loss=0.05661, over 15391.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2783, pruned_loss=0.04701, over 3097575.25 frames. ], batch size: 193, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:44:52,484 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:45:08,845 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 07:46:08,566 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.813e+02 2.533e+02 2.923e+02 3.957e+02 1.423e+03, threshold=5.846e+02, percent-clipped=5.0 2023-04-29 07:46:37,395 INFO [train.py:904] (5/8) Epoch 10, batch 9100, loss[loss=0.1859, simple_loss=0.2842, pruned_loss=0.04379, over 16463.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2783, pruned_loss=0.04794, over 3080098.76 frames. ], batch size: 147, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:46:58,820 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:48:34,516 INFO [train.py:904] (5/8) Epoch 10, batch 9150, loss[loss=0.1848, simple_loss=0.2775, pruned_loss=0.04606, over 15315.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2784, pruned_loss=0.04763, over 3053803.95 frames. ], batch size: 190, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:49:10,539 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8671, 4.0052, 3.7988, 3.5920, 3.3064, 3.9215, 3.6265, 3.5913], device='cuda:5'), covar=tensor([0.0661, 0.0500, 0.0333, 0.0328, 0.0998, 0.0462, 0.1012, 0.0710], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0278, 0.0260, 0.0241, 0.0282, 0.0276, 0.0182, 0.0305], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:49:54,225 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-29 07:49:54,595 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.281e+02 2.663e+02 3.521e+02 5.779e+02, threshold=5.326e+02, percent-clipped=0.0 2023-04-29 07:50:13,997 INFO [train.py:904] (5/8) Epoch 10, batch 9200, loss[loss=0.2013, simple_loss=0.2943, pruned_loss=0.05415, over 16751.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2744, pruned_loss=0.04683, over 3052365.17 frames. ], batch size: 124, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:51:35,149 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:51:50,959 INFO [train.py:904] (5/8) Epoch 10, batch 9250, loss[loss=0.1731, simple_loss=0.2682, pruned_loss=0.03903, over 16803.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2738, pruned_loss=0.04687, over 3048103.29 frames. ], batch size: 83, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:52:09,971 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4338, 1.9876, 2.1142, 3.9495, 1.9624, 2.4163, 2.1651, 2.1584], device='cuda:5'), covar=tensor([0.0797, 0.3430, 0.2264, 0.0363, 0.3838, 0.2301, 0.2947, 0.3253], device='cuda:5'), in_proj_covar=tensor([0.0340, 0.0373, 0.0314, 0.0307, 0.0402, 0.0416, 0.0335, 0.0432], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 07:53:14,008 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.424e+02 2.858e+02 3.550e+02 5.723e+02, threshold=5.716e+02, percent-clipped=3.0 2023-04-29 07:53:42,543 INFO [train.py:904] (5/8) Epoch 10, batch 9300, loss[loss=0.1558, simple_loss=0.2449, pruned_loss=0.0334, over 16463.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2723, pruned_loss=0.04619, over 3050069.93 frames. ], batch size: 68, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:54:09,419 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:54:21,748 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-29 07:55:29,738 INFO [train.py:904] (5/8) Epoch 10, batch 9350, loss[loss=0.1687, simple_loss=0.2629, pruned_loss=0.03727, over 16776.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2717, pruned_loss=0.04592, over 3050527.69 frames. ], batch size: 83, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:55:50,308 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:55:59,249 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6469, 3.2900, 3.3394, 1.8119, 2.7901, 2.3105, 3.1799, 3.2724], device='cuda:5'), covar=tensor([0.0284, 0.0531, 0.0444, 0.1748, 0.0686, 0.0845, 0.0713, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0133, 0.0153, 0.0140, 0.0133, 0.0123, 0.0133, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 07:56:13,425 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 07:56:29,127 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:56:48,377 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.395e+02 2.806e+02 3.199e+02 7.223e+02, threshold=5.612e+02, percent-clipped=2.0 2023-04-29 07:57:12,458 INFO [train.py:904] (5/8) Epoch 10, batch 9400, loss[loss=0.1541, simple_loss=0.2368, pruned_loss=0.03566, over 12556.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2712, pruned_loss=0.04573, over 3030132.17 frames. ], batch size: 247, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:57:23,158 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.48 vs. limit=5.0 2023-04-29 07:57:25,483 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:58:33,089 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:58:52,911 INFO [train.py:904] (5/8) Epoch 10, batch 9450, loss[loss=0.1829, simple_loss=0.2773, pruned_loss=0.04429, over 15335.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2729, pruned_loss=0.04595, over 3033286.49 frames. ], batch size: 191, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:59:47,500 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 08:00:10,873 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.411e+02 3.092e+02 3.645e+02 7.379e+02, threshold=6.183e+02, percent-clipped=4.0 2023-04-29 08:00:34,630 INFO [train.py:904] (5/8) Epoch 10, batch 9500, loss[loss=0.1644, simple_loss=0.2641, pruned_loss=0.03233, over 16838.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2722, pruned_loss=0.04546, over 3037068.20 frames. ], batch size: 102, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:01:09,406 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 08:01:58,620 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4329, 2.9795, 2.6523, 2.2166, 2.1626, 2.1537, 2.9619, 2.8422], device='cuda:5'), covar=tensor([0.2225, 0.0632, 0.1287, 0.2026, 0.2186, 0.1842, 0.0449, 0.0971], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0245, 0.0272, 0.0265, 0.0255, 0.0212, 0.0254, 0.0274], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:02:03,548 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:02:20,739 INFO [train.py:904] (5/8) Epoch 10, batch 9550, loss[loss=0.1975, simple_loss=0.2934, pruned_loss=0.05082, over 16241.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.273, pruned_loss=0.04609, over 3038131.80 frames. ], batch size: 146, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:03:40,249 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.447e+02 2.728e+02 3.779e+02 7.191e+02, threshold=5.455e+02, percent-clipped=3.0 2023-04-29 08:03:45,444 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:04:02,629 INFO [train.py:904] (5/8) Epoch 10, batch 9600, loss[loss=0.2065, simple_loss=0.3073, pruned_loss=0.05292, over 16308.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2748, pruned_loss=0.04678, over 3047807.82 frames. ], batch size: 165, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:05:25,331 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6773, 2.6315, 2.3666, 4.0866, 2.7668, 4.0095, 1.4246, 2.7787], device='cuda:5'), covar=tensor([0.1402, 0.0637, 0.1177, 0.0129, 0.0164, 0.0380, 0.1521, 0.0796], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0152, 0.0175, 0.0130, 0.0182, 0.0198, 0.0177, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 08:05:52,888 INFO [train.py:904] (5/8) Epoch 10, batch 9650, loss[loss=0.1857, simple_loss=0.2859, pruned_loss=0.04269, over 16992.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2768, pruned_loss=0.04693, over 3051540.34 frames. ], batch size: 109, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:06:30,348 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3101, 4.2568, 4.7089, 4.6986, 4.7000, 4.4386, 4.4084, 4.2178], device='cuda:5'), covar=tensor([0.0250, 0.0468, 0.0338, 0.0358, 0.0366, 0.0276, 0.0692, 0.0354], device='cuda:5'), in_proj_covar=tensor([0.0294, 0.0303, 0.0306, 0.0291, 0.0350, 0.0321, 0.0408, 0.0264], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-29 08:07:15,511 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.568e+02 3.331e+02 4.260e+02 1.013e+03, threshold=6.663e+02, percent-clipped=7.0 2023-04-29 08:07:41,471 INFO [train.py:904] (5/8) Epoch 10, batch 9700, loss[loss=0.181, simple_loss=0.2735, pruned_loss=0.04431, over 16873.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2748, pruned_loss=0.04596, over 3072678.61 frames. ], batch size: 102, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:07:47,672 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7840, 3.4332, 3.5218, 2.0010, 2.8934, 2.2839, 3.2794, 3.3656], device='cuda:5'), covar=tensor([0.0306, 0.0599, 0.0412, 0.1720, 0.0713, 0.0899, 0.0747, 0.0842], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0132, 0.0152, 0.0140, 0.0132, 0.0122, 0.0132, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 08:07:52,626 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:08:54,365 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:09:25,212 INFO [train.py:904] (5/8) Epoch 10, batch 9750, loss[loss=0.1977, simple_loss=0.273, pruned_loss=0.06118, over 12269.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2741, pruned_loss=0.0463, over 3067442.71 frames. ], batch size: 249, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:09:32,961 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:09:35,517 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:09:35,906 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 08:10:45,039 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.458e+02 3.043e+02 3.868e+02 6.520e+02, threshold=6.087e+02, percent-clipped=0.0 2023-04-29 08:11:05,037 INFO [train.py:904] (5/8) Epoch 10, batch 9800, loss[loss=0.1684, simple_loss=0.2531, pruned_loss=0.04184, over 12158.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2739, pruned_loss=0.0451, over 3068195.38 frames. ], batch size: 246, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:11:07,734 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 08:11:20,997 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3712, 4.4169, 4.2425, 3.9876, 3.9145, 4.3060, 4.1461, 3.9946], device='cuda:5'), covar=tensor([0.0555, 0.0568, 0.0255, 0.0279, 0.0870, 0.0472, 0.0430, 0.0691], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0277, 0.0260, 0.0240, 0.0278, 0.0272, 0.0180, 0.0304], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:11:37,155 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:12:49,195 INFO [train.py:904] (5/8) Epoch 10, batch 9850, loss[loss=0.1787, simple_loss=0.2657, pruned_loss=0.04587, over 12520.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2752, pruned_loss=0.04497, over 3065997.42 frames. ], batch size: 247, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:14:17,929 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.461e+02 2.965e+02 3.704e+02 9.099e+02, threshold=5.931e+02, percent-clipped=1.0 2023-04-29 08:14:41,628 INFO [train.py:904] (5/8) Epoch 10, batch 9900, loss[loss=0.1868, simple_loss=0.2802, pruned_loss=0.0467, over 16902.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2754, pruned_loss=0.04481, over 3065521.14 frames. ], batch size: 116, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:16:40,051 INFO [train.py:904] (5/8) Epoch 10, batch 9950, loss[loss=0.2086, simple_loss=0.2994, pruned_loss=0.05893, over 15456.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2777, pruned_loss=0.04536, over 3064328.70 frames. ], batch size: 190, lr: 6.68e-03, grad_scale: 4.0 2023-04-29 08:18:13,671 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 2.620e+02 2.929e+02 3.358e+02 7.407e+02, threshold=5.858e+02, percent-clipped=2.0 2023-04-29 08:18:17,694 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2970, 4.3142, 4.1659, 3.8815, 3.7945, 4.2100, 4.0008, 3.9808], device='cuda:5'), covar=tensor([0.0484, 0.0573, 0.0280, 0.0266, 0.0831, 0.0459, 0.0466, 0.0582], device='cuda:5'), in_proj_covar=tensor([0.0221, 0.0274, 0.0259, 0.0238, 0.0275, 0.0271, 0.0178, 0.0300], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-29 08:18:42,217 INFO [train.py:904] (5/8) Epoch 10, batch 10000, loss[loss=0.1767, simple_loss=0.2729, pruned_loss=0.04027, over 16750.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2759, pruned_loss=0.04462, over 3079448.90 frames. ], batch size: 134, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:18:49,606 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7707, 4.1953, 3.3514, 2.2423, 2.9161, 2.5147, 4.5063, 3.7659], device='cuda:5'), covar=tensor([0.2421, 0.0471, 0.1165, 0.2216, 0.1851, 0.1546, 0.0338, 0.0763], device='cuda:5'), in_proj_covar=tensor([0.0289, 0.0243, 0.0271, 0.0263, 0.0252, 0.0209, 0.0253, 0.0272], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:19:20,667 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8310, 1.2824, 1.5530, 1.7201, 1.7788, 1.9020, 1.5897, 1.7880], device='cuda:5'), covar=tensor([0.0156, 0.0304, 0.0151, 0.0190, 0.0199, 0.0129, 0.0298, 0.0090], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0164, 0.0148, 0.0149, 0.0162, 0.0113, 0.0167, 0.0104], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 08:19:20,753 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9224, 2.0048, 2.2388, 3.2199, 2.0543, 2.1936, 2.1660, 2.0726], device='cuda:5'), covar=tensor([0.0849, 0.3120, 0.1966, 0.0465, 0.3726, 0.2252, 0.2966, 0.3140], device='cuda:5'), in_proj_covar=tensor([0.0341, 0.0368, 0.0312, 0.0303, 0.0396, 0.0411, 0.0334, 0.0429], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:19:54,023 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:20:23,609 INFO [train.py:904] (5/8) Epoch 10, batch 10050, loss[loss=0.1996, simple_loss=0.2941, pruned_loss=0.05251, over 16634.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.276, pruned_loss=0.04466, over 3067065.92 frames. ], batch size: 134, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:20:26,520 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3519, 3.3606, 2.7676, 2.0353, 2.2824, 2.1617, 3.5870, 3.1193], device='cuda:5'), covar=tensor([0.2660, 0.0612, 0.1413, 0.2168, 0.2109, 0.1749, 0.0460, 0.1062], device='cuda:5'), in_proj_covar=tensor([0.0288, 0.0242, 0.0270, 0.0262, 0.0251, 0.0209, 0.0252, 0.0272], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:21:24,005 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9260, 2.7983, 2.5994, 2.0003, 2.5472, 2.7658, 2.6772, 1.8453], device='cuda:5'), covar=tensor([0.0295, 0.0035, 0.0042, 0.0249, 0.0083, 0.0057, 0.0060, 0.0324], device='cuda:5'), in_proj_covar=tensor([0.0120, 0.0061, 0.0064, 0.0118, 0.0073, 0.0080, 0.0071, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 08:21:25,329 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:21:30,295 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9109, 3.8422, 2.2549, 4.4952, 2.8739, 4.3175, 2.5444, 3.1210], device='cuda:5'), covar=tensor([0.0195, 0.0316, 0.1537, 0.0101, 0.0802, 0.0467, 0.1333, 0.0651], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0152, 0.0180, 0.0114, 0.0160, 0.0190, 0.0189, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 08:21:36,406 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.448e+02 2.853e+02 3.607e+02 8.482e+02, threshold=5.706e+02, percent-clipped=4.0 2023-04-29 08:21:56,151 INFO [train.py:904] (5/8) Epoch 10, batch 10100, loss[loss=0.1804, simple_loss=0.2675, pruned_loss=0.04665, over 16711.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2766, pruned_loss=0.04549, over 3057596.03 frames. ], batch size: 134, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:22:09,210 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-04-29 08:22:16,042 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:22:52,257 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5925, 4.3661, 4.6197, 4.7809, 4.9313, 4.4446, 4.9274, 4.9082], device='cuda:5'), covar=tensor([0.1381, 0.0988, 0.1326, 0.0552, 0.0425, 0.0750, 0.0357, 0.0526], device='cuda:5'), in_proj_covar=tensor([0.0468, 0.0585, 0.0706, 0.0600, 0.0454, 0.0455, 0.0471, 0.0532], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:23:11,691 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-29 08:23:38,434 INFO [train.py:904] (5/8) Epoch 11, batch 0, loss[loss=0.2891, simple_loss=0.3515, pruned_loss=0.1133, over 12242.00 frames. ], tot_loss[loss=0.2891, simple_loss=0.3515, pruned_loss=0.1133, over 12242.00 frames. ], batch size: 246, lr: 6.37e-03, grad_scale: 8.0 2023-04-29 08:23:38,434 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 08:23:45,827 INFO [train.py:938] (5/8) Epoch 11, validation: loss=0.1557, simple_loss=0.2595, pruned_loss=0.02591, over 944034.00 frames. 2023-04-29 08:23:45,828 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 08:24:32,825 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6915, 6.0650, 5.7761, 5.7943, 5.3962, 5.4939, 5.5502, 6.1539], device='cuda:5'), covar=tensor([0.1142, 0.0870, 0.1165, 0.0662, 0.0758, 0.0568, 0.0950, 0.0797], device='cuda:5'), in_proj_covar=tensor([0.0495, 0.0619, 0.0514, 0.0430, 0.0392, 0.0411, 0.0523, 0.0471], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:24:43,129 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.573e+02 2.991e+02 3.932e+02 8.479e+02, threshold=5.981e+02, percent-clipped=4.0 2023-04-29 08:24:55,134 INFO [train.py:904] (5/8) Epoch 11, batch 50, loss[loss=0.228, simple_loss=0.3116, pruned_loss=0.07225, over 15484.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2853, pruned_loss=0.06262, over 754071.13 frames. ], batch size: 190, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:25:21,503 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 08:26:05,603 INFO [train.py:904] (5/8) Epoch 11, batch 100, loss[loss=0.2167, simple_loss=0.2904, pruned_loss=0.07152, over 16722.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2827, pruned_loss=0.06057, over 1317641.64 frames. ], batch size: 124, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:26:49,162 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-29 08:27:03,339 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.523e+02 3.127e+02 3.796e+02 9.987e+02, threshold=6.254e+02, percent-clipped=2.0 2023-04-29 08:27:06,084 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9730, 1.8851, 2.3680, 2.8056, 2.8886, 2.7573, 1.9176, 3.0454], device='cuda:5'), covar=tensor([0.0108, 0.0296, 0.0217, 0.0161, 0.0133, 0.0171, 0.0319, 0.0092], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0168, 0.0152, 0.0153, 0.0164, 0.0117, 0.0169, 0.0107], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 08:27:12,717 INFO [train.py:904] (5/8) Epoch 11, batch 150, loss[loss=0.1877, simple_loss=0.2748, pruned_loss=0.05036, over 16713.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2798, pruned_loss=0.05842, over 1768262.41 frames. ], batch size: 76, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:27:39,413 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 08:27:59,453 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:28:23,286 INFO [train.py:904] (5/8) Epoch 11, batch 200, loss[loss=0.1698, simple_loss=0.2582, pruned_loss=0.04067, over 17175.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.278, pruned_loss=0.05673, over 2117206.16 frames. ], batch size: 46, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:21,764 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.429e+02 2.923e+02 3.406e+02 5.474e+02, threshold=5.846e+02, percent-clipped=1.0 2023-04-29 08:29:23,320 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:29:31,806 INFO [train.py:904] (5/8) Epoch 11, batch 250, loss[loss=0.2079, simple_loss=0.278, pruned_loss=0.06885, over 16763.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2765, pruned_loss=0.05694, over 2386551.12 frames. ], batch size: 124, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:35,625 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8995, 4.8526, 4.7407, 4.2839, 4.7452, 1.8560, 4.5195, 4.6049], device='cuda:5'), covar=tensor([0.0081, 0.0070, 0.0134, 0.0273, 0.0079, 0.2291, 0.0111, 0.0177], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0111, 0.0156, 0.0144, 0.0127, 0.0176, 0.0142, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:29:46,232 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:30:37,970 INFO [train.py:904] (5/8) Epoch 11, batch 300, loss[loss=0.208, simple_loss=0.3094, pruned_loss=0.05325, over 17249.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2754, pruned_loss=0.05619, over 2596830.68 frames. ], batch size: 52, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:30:51,065 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:30:58,033 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5230, 4.6305, 4.8051, 4.6850, 4.6557, 5.2951, 4.8767, 4.5764], device='cuda:5'), covar=tensor([0.1407, 0.2045, 0.2163, 0.2157, 0.3012, 0.1172, 0.1565, 0.2649], device='cuda:5'), in_proj_covar=tensor([0.0339, 0.0481, 0.0516, 0.0416, 0.0554, 0.0546, 0.0413, 0.0561], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 08:31:35,546 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.476e+02 3.003e+02 3.668e+02 8.355e+02, threshold=6.006e+02, percent-clipped=2.0 2023-04-29 08:31:47,613 INFO [train.py:904] (5/8) Epoch 11, batch 350, loss[loss=0.1932, simple_loss=0.2703, pruned_loss=0.05804, over 15595.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2729, pruned_loss=0.05509, over 2760718.60 frames. ], batch size: 190, lr: 6.36e-03, grad_scale: 1.0 2023-04-29 08:31:51,323 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8982, 2.8410, 2.2723, 2.7183, 3.1773, 2.9057, 3.6733, 3.4842], device='cuda:5'), covar=tensor([0.0062, 0.0279, 0.0357, 0.0273, 0.0188, 0.0247, 0.0176, 0.0161], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0202, 0.0200, 0.0199, 0.0200, 0.0203, 0.0202, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:32:54,390 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6370, 2.6259, 2.0785, 2.5443, 2.9987, 2.8488, 3.5058, 3.2987], device='cuda:5'), covar=tensor([0.0073, 0.0280, 0.0383, 0.0320, 0.0190, 0.0256, 0.0144, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0201, 0.0199, 0.0197, 0.0200, 0.0202, 0.0201, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:32:56,343 INFO [train.py:904] (5/8) Epoch 11, batch 400, loss[loss=0.1727, simple_loss=0.2515, pruned_loss=0.04697, over 17111.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2706, pruned_loss=0.05408, over 2883916.80 frames. ], batch size: 48, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:33:21,206 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-29 08:33:22,116 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:33:54,606 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.298e+02 2.716e+02 3.213e+02 6.833e+02, threshold=5.433e+02, percent-clipped=1.0 2023-04-29 08:34:06,168 INFO [train.py:904] (5/8) Epoch 11, batch 450, loss[loss=0.1986, simple_loss=0.2693, pruned_loss=0.06397, over 16745.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2681, pruned_loss=0.05313, over 2985731.88 frames. ], batch size: 134, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:34:47,030 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:34:54,609 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:35:19,215 INFO [train.py:904] (5/8) Epoch 11, batch 500, loss[loss=0.169, simple_loss=0.2475, pruned_loss=0.04525, over 16935.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2672, pruned_loss=0.05262, over 3052316.31 frames. ], batch size: 90, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:35:44,053 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9185, 2.2962, 2.3110, 4.6702, 2.2064, 2.8173, 2.4234, 2.5443], device='cuda:5'), covar=tensor([0.0778, 0.3193, 0.2229, 0.0319, 0.3569, 0.2122, 0.2857, 0.3015], device='cuda:5'), in_proj_covar=tensor([0.0357, 0.0383, 0.0326, 0.0318, 0.0409, 0.0434, 0.0348, 0.0452], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:36:13,473 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:19,096 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.707e+02 2.413e+02 2.842e+02 3.807e+02 1.018e+03, threshold=5.684e+02, percent-clipped=5.0 2023-04-29 08:36:24,911 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:29,131 INFO [train.py:904] (5/8) Epoch 11, batch 550, loss[loss=0.1698, simple_loss=0.2622, pruned_loss=0.03871, over 17092.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2665, pruned_loss=0.05243, over 3112469.41 frames. ], batch size: 47, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:36:57,850 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-29 08:37:40,181 INFO [train.py:904] (5/8) Epoch 11, batch 600, loss[loss=0.1815, simple_loss=0.2717, pruned_loss=0.04571, over 17017.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2657, pruned_loss=0.05186, over 3159653.11 frames. ], batch size: 50, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:38:38,910 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.665e+02 3.250e+02 4.151e+02 1.142e+03, threshold=6.501e+02, percent-clipped=8.0 2023-04-29 08:38:49,714 INFO [train.py:904] (5/8) Epoch 11, batch 650, loss[loss=0.1567, simple_loss=0.2466, pruned_loss=0.03345, over 16834.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2652, pruned_loss=0.0518, over 3184002.61 frames. ], batch size: 42, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:39:03,791 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3509, 2.3327, 1.8395, 2.1597, 2.7313, 2.5024, 2.8567, 2.8629], device='cuda:5'), covar=tensor([0.0148, 0.0273, 0.0371, 0.0314, 0.0150, 0.0224, 0.0171, 0.0165], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0204, 0.0201, 0.0199, 0.0203, 0.0203, 0.0205, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:39:58,910 INFO [train.py:904] (5/8) Epoch 11, batch 700, loss[loss=0.1786, simple_loss=0.2506, pruned_loss=0.05326, over 12004.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2655, pruned_loss=0.05155, over 3220236.37 frames. ], batch size: 247, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:40:41,862 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9740, 4.7370, 5.0262, 5.2456, 5.4030, 4.7809, 5.4085, 5.3633], device='cuda:5'), covar=tensor([0.1786, 0.1133, 0.1753, 0.0736, 0.0575, 0.0801, 0.0509, 0.0524], device='cuda:5'), in_proj_covar=tensor([0.0531, 0.0662, 0.0809, 0.0681, 0.0515, 0.0511, 0.0528, 0.0600], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:40:51,670 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5904, 4.7454, 4.9127, 4.7232, 4.6679, 5.3353, 4.9118, 4.6238], device='cuda:5'), covar=tensor([0.1397, 0.1735, 0.1862, 0.2141, 0.3190, 0.1139, 0.1514, 0.2351], device='cuda:5'), in_proj_covar=tensor([0.0344, 0.0485, 0.0525, 0.0420, 0.0561, 0.0552, 0.0418, 0.0566], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 08:40:57,197 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.406e+02 3.042e+02 3.688e+02 8.109e+02, threshold=6.084e+02, percent-clipped=2.0 2023-04-29 08:41:09,231 INFO [train.py:904] (5/8) Epoch 11, batch 750, loss[loss=0.1835, simple_loss=0.2595, pruned_loss=0.05371, over 16893.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2662, pruned_loss=0.0524, over 3235898.18 frames. ], batch size: 96, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:41:42,362 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 08:42:18,058 INFO [train.py:904] (5/8) Epoch 11, batch 800, loss[loss=0.1576, simple_loss=0.2385, pruned_loss=0.03834, over 17228.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2652, pruned_loss=0.05134, over 3259704.78 frames. ], batch size: 43, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:42:57,881 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0766, 3.9117, 4.1007, 4.2614, 4.3533, 3.8781, 4.1207, 4.3549], device='cuda:5'), covar=tensor([0.1256, 0.1030, 0.1218, 0.0632, 0.0610, 0.1447, 0.1845, 0.0616], device='cuda:5'), in_proj_covar=tensor([0.0525, 0.0658, 0.0805, 0.0678, 0.0512, 0.0511, 0.0526, 0.0597], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:43:11,967 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:15,330 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:16,116 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.524e+02 3.019e+02 3.517e+02 6.392e+02, threshold=6.037e+02, percent-clipped=1.0 2023-04-29 08:43:25,166 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6433, 2.6577, 2.2380, 2.5416, 3.0190, 2.8031, 3.4670, 3.2799], device='cuda:5'), covar=tensor([0.0081, 0.0299, 0.0380, 0.0334, 0.0207, 0.0270, 0.0193, 0.0203], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0206, 0.0202, 0.0201, 0.0205, 0.0204, 0.0208, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:43:27,557 INFO [train.py:904] (5/8) Epoch 11, batch 850, loss[loss=0.1682, simple_loss=0.2399, pruned_loss=0.04824, over 16663.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2642, pruned_loss=0.05113, over 3267225.26 frames. ], batch size: 134, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:44:17,573 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:44:31,805 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5487, 3.9497, 4.1072, 2.8742, 3.6670, 4.0883, 3.7193, 2.2291], device='cuda:5'), covar=tensor([0.0424, 0.0090, 0.0046, 0.0323, 0.0078, 0.0092, 0.0078, 0.0400], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0071, 0.0071, 0.0128, 0.0079, 0.0089, 0.0078, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 08:44:37,486 INFO [train.py:904] (5/8) Epoch 11, batch 900, loss[loss=0.2197, simple_loss=0.2922, pruned_loss=0.07358, over 15424.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2638, pruned_loss=0.05103, over 3271502.04 frames. ], batch size: 190, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:44:58,571 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6493, 2.7511, 2.2580, 2.6500, 3.0684, 2.8433, 3.5610, 3.3287], device='cuda:5'), covar=tensor([0.0084, 0.0268, 0.0389, 0.0310, 0.0196, 0.0264, 0.0162, 0.0196], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0206, 0.0203, 0.0201, 0.0205, 0.0204, 0.0208, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:45:35,220 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.314e+02 2.859e+02 3.467e+02 6.346e+02, threshold=5.718e+02, percent-clipped=1.0 2023-04-29 08:45:45,395 INFO [train.py:904] (5/8) Epoch 11, batch 950, loss[loss=0.1549, simple_loss=0.25, pruned_loss=0.02989, over 17230.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2637, pruned_loss=0.05101, over 3291006.46 frames. ], batch size: 45, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:46:07,435 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-29 08:46:54,284 INFO [train.py:904] (5/8) Epoch 11, batch 1000, loss[loss=0.1717, simple_loss=0.2454, pruned_loss=0.04898, over 16854.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2629, pruned_loss=0.05079, over 3294460.60 frames. ], batch size: 90, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:47:00,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6074, 2.9026, 2.7040, 4.6882, 3.5396, 4.3172, 1.5301, 3.1702], device='cuda:5'), covar=tensor([0.1686, 0.0833, 0.1290, 0.0222, 0.0374, 0.0385, 0.1867, 0.0822], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0157, 0.0180, 0.0139, 0.0193, 0.0209, 0.0180, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 08:47:52,033 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.460e+02 2.841e+02 3.268e+02 6.263e+02, threshold=5.683e+02, percent-clipped=1.0 2023-04-29 08:48:02,263 INFO [train.py:904] (5/8) Epoch 11, batch 1050, loss[loss=0.1693, simple_loss=0.2616, pruned_loss=0.03856, over 17187.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2619, pruned_loss=0.05065, over 3304466.70 frames. ], batch size: 46, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:48:36,351 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:48:42,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8926, 2.0653, 2.3718, 3.1349, 2.1162, 2.2064, 2.2875, 2.1536], device='cuda:5'), covar=tensor([0.1049, 0.2873, 0.1918, 0.0639, 0.3370, 0.2226, 0.2608, 0.3180], device='cuda:5'), in_proj_covar=tensor([0.0362, 0.0391, 0.0329, 0.0324, 0.0412, 0.0442, 0.0354, 0.0458], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 08:49:12,623 INFO [train.py:904] (5/8) Epoch 11, batch 1100, loss[loss=0.1876, simple_loss=0.265, pruned_loss=0.05507, over 12555.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2618, pruned_loss=0.05027, over 3305579.80 frames. ], batch size: 247, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:49:43,774 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:49:52,410 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 08:50:07,615 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:50:09,061 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.199e+02 2.657e+02 3.494e+02 5.570e+02, threshold=5.313e+02, percent-clipped=0.0 2023-04-29 08:50:20,436 INFO [train.py:904] (5/8) Epoch 11, batch 1150, loss[loss=0.1978, simple_loss=0.2715, pruned_loss=0.06211, over 16850.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2613, pruned_loss=0.05004, over 3306256.99 frames. ], batch size: 116, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:51:13,766 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-29 08:51:14,267 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:51:21,313 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 08:51:27,910 INFO [train.py:904] (5/8) Epoch 11, batch 1200, loss[loss=0.1849, simple_loss=0.282, pruned_loss=0.04388, over 17108.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2604, pruned_loss=0.04902, over 3317177.24 frames. ], batch size: 48, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:52:27,632 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.248e+02 2.693e+02 3.149e+02 5.178e+02, threshold=5.386e+02, percent-clipped=0.0 2023-04-29 08:52:39,082 INFO [train.py:904] (5/8) Epoch 11, batch 1250, loss[loss=0.1669, simple_loss=0.2405, pruned_loss=0.04663, over 16425.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2608, pruned_loss=0.04924, over 3312390.11 frames. ], batch size: 146, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:53:44,500 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-29 08:53:49,401 INFO [train.py:904] (5/8) Epoch 11, batch 1300, loss[loss=0.1918, simple_loss=0.2689, pruned_loss=0.05737, over 16766.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2603, pruned_loss=0.04962, over 3310471.86 frames. ], batch size: 124, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:54:46,487 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.481e+02 2.988e+02 3.715e+02 8.832e+02, threshold=5.975e+02, percent-clipped=5.0 2023-04-29 08:54:57,998 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102851.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:54:58,772 INFO [train.py:904] (5/8) Epoch 11, batch 1350, loss[loss=0.1823, simple_loss=0.2724, pruned_loss=0.04606, over 16649.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2608, pruned_loss=0.04979, over 3311814.79 frames. ], batch size: 62, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:55:45,085 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:56:05,625 INFO [train.py:904] (5/8) Epoch 11, batch 1400, loss[loss=0.1693, simple_loss=0.2365, pruned_loss=0.05102, over 16401.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2613, pruned_loss=0.04999, over 3316742.19 frames. ], batch size: 146, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:56:20,011 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:57:05,104 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.362e+02 2.982e+02 3.517e+02 5.508e+02, threshold=5.963e+02, percent-clipped=1.0 2023-04-29 08:57:10,183 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:57:15,717 INFO [train.py:904] (5/8) Epoch 11, batch 1450, loss[loss=0.1666, simple_loss=0.2372, pruned_loss=0.04793, over 15555.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2606, pruned_loss=0.04985, over 3306620.91 frames. ], batch size: 190, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:58:05,986 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-04-29 08:58:25,011 INFO [train.py:904] (5/8) Epoch 11, batch 1500, loss[loss=0.1881, simple_loss=0.2757, pruned_loss=0.05026, over 16181.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2607, pruned_loss=0.04964, over 3308435.10 frames. ], batch size: 36, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:59:24,567 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.416e+02 2.850e+02 3.321e+02 6.158e+02, threshold=5.699e+02, percent-clipped=1.0 2023-04-29 08:59:35,071 INFO [train.py:904] (5/8) Epoch 11, batch 1550, loss[loss=0.1968, simple_loss=0.2697, pruned_loss=0.06189, over 16403.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.262, pruned_loss=0.05069, over 3313937.84 frames. ], batch size: 146, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 09:00:40,049 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:00:45,005 INFO [train.py:904] (5/8) Epoch 11, batch 1600, loss[loss=0.1746, simple_loss=0.2582, pruned_loss=0.04546, over 17205.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2636, pruned_loss=0.05146, over 3313835.09 frames. ], batch size: 45, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:01:26,015 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:01:42,626 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.417e+02 3.013e+02 3.629e+02 9.117e+02, threshold=6.027e+02, percent-clipped=4.0 2023-04-29 09:01:53,531 INFO [train.py:904] (5/8) Epoch 11, batch 1650, loss[loss=0.1955, simple_loss=0.2694, pruned_loss=0.06078, over 16725.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2656, pruned_loss=0.05215, over 3300508.34 frames. ], batch size: 89, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:02:03,323 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:02:50,102 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:03:02,615 INFO [train.py:904] (5/8) Epoch 11, batch 1700, loss[loss=0.2108, simple_loss=0.2954, pruned_loss=0.06309, over 16663.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2683, pruned_loss=0.05254, over 3306640.34 frames. ], batch size: 57, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:03:10,678 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:03:14,257 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6068, 6.0159, 5.7528, 5.8030, 5.3766, 5.2683, 5.4532, 6.1595], device='cuda:5'), covar=tensor([0.1184, 0.0887, 0.1143, 0.0721, 0.0869, 0.0644, 0.1015, 0.0773], device='cuda:5'), in_proj_covar=tensor([0.0552, 0.0691, 0.0572, 0.0482, 0.0432, 0.0447, 0.0579, 0.0525], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:04:01,720 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:04:01,811 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2063, 5.1800, 4.9265, 4.3909, 4.9801, 2.1345, 4.7597, 4.9210], device='cuda:5'), covar=tensor([0.0069, 0.0058, 0.0141, 0.0333, 0.0075, 0.2086, 0.0104, 0.0161], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0119, 0.0169, 0.0157, 0.0138, 0.0182, 0.0155, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:04:02,561 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.478e+02 3.058e+02 3.640e+02 7.869e+02, threshold=6.116e+02, percent-clipped=2.0 2023-04-29 09:04:13,837 INFO [train.py:904] (5/8) Epoch 11, batch 1750, loss[loss=0.1929, simple_loss=0.2753, pruned_loss=0.05526, over 16722.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2685, pruned_loss=0.05118, over 3316862.94 frames. ], batch size: 57, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:05:22,484 INFO [train.py:904] (5/8) Epoch 11, batch 1800, loss[loss=0.1819, simple_loss=0.2774, pruned_loss=0.04317, over 17103.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.27, pruned_loss=0.05205, over 3318376.91 frames. ], batch size: 48, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:05:41,567 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0437, 3.9859, 3.9458, 3.3834, 3.9614, 1.7376, 3.7447, 3.5370], device='cuda:5'), covar=tensor([0.0098, 0.0090, 0.0139, 0.0248, 0.0077, 0.2496, 0.0117, 0.0202], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0120, 0.0170, 0.0158, 0.0138, 0.0182, 0.0156, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:06:01,168 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 09:06:21,368 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.500e+02 2.964e+02 3.680e+02 1.127e+03, threshold=5.929e+02, percent-clipped=5.0 2023-04-29 09:06:31,958 INFO [train.py:904] (5/8) Epoch 11, batch 1850, loss[loss=0.1945, simple_loss=0.2872, pruned_loss=0.05094, over 17038.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2703, pruned_loss=0.05155, over 3316084.01 frames. ], batch size: 53, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:47,822 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9648, 4.7516, 4.9666, 5.2243, 5.3988, 4.7347, 5.3798, 5.3339], device='cuda:5'), covar=tensor([0.1609, 0.1193, 0.1824, 0.0711, 0.0511, 0.0807, 0.0481, 0.0505], device='cuda:5'), in_proj_covar=tensor([0.0548, 0.0683, 0.0841, 0.0703, 0.0531, 0.0529, 0.0541, 0.0620], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:06:51,443 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:07:04,726 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6759, 3.0589, 2.5963, 4.8395, 3.9466, 4.3746, 1.5709, 3.2507], device='cuda:5'), covar=tensor([0.1313, 0.0662, 0.1131, 0.0159, 0.0266, 0.0336, 0.1455, 0.0703], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0158, 0.0180, 0.0144, 0.0196, 0.0210, 0.0180, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 09:07:04,918 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 09:07:22,586 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4543, 4.7798, 4.5814, 4.5735, 4.2715, 4.2310, 4.2298, 4.8298], device='cuda:5'), covar=tensor([0.1055, 0.0802, 0.0917, 0.0656, 0.0746, 0.1198, 0.0983, 0.0774], device='cuda:5'), in_proj_covar=tensor([0.0544, 0.0683, 0.0566, 0.0478, 0.0427, 0.0443, 0.0573, 0.0522], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:07:39,141 INFO [train.py:904] (5/8) Epoch 11, batch 1900, loss[loss=0.1765, simple_loss=0.2634, pruned_loss=0.04485, over 17180.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2701, pruned_loss=0.05101, over 3312305.27 frames. ], batch size: 46, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:08:16,174 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:08:40,926 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.233e+02 2.712e+02 3.149e+02 7.257e+02, threshold=5.425e+02, percent-clipped=1.0 2023-04-29 09:08:51,897 INFO [train.py:904] (5/8) Epoch 11, batch 1950, loss[loss=0.1977, simple_loss=0.2719, pruned_loss=0.06174, over 16902.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2699, pruned_loss=0.05016, over 3324648.08 frames. ], batch size: 109, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:08:55,107 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:09:41,448 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:00,150 INFO [train.py:904] (5/8) Epoch 11, batch 2000, loss[loss=0.211, simple_loss=0.278, pruned_loss=0.07197, over 16861.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2695, pruned_loss=0.05002, over 3322053.66 frames. ], batch size: 109, lr: 6.31e-03, grad_scale: 8.0 2023-04-29 09:10:07,703 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:27,137 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 09:10:58,982 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:11:00,881 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.374e+02 2.796e+02 3.607e+02 7.066e+02, threshold=5.592e+02, percent-clipped=4.0 2023-04-29 09:11:10,892 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 09:11:11,332 INFO [train.py:904] (5/8) Epoch 11, batch 2050, loss[loss=0.223, simple_loss=0.2852, pruned_loss=0.08039, over 16916.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2683, pruned_loss=0.05008, over 3320517.38 frames. ], batch size: 116, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:11:16,493 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:12:05,795 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:12:21,608 INFO [train.py:904] (5/8) Epoch 11, batch 2100, loss[loss=0.1956, simple_loss=0.2674, pruned_loss=0.06186, over 16753.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2704, pruned_loss=0.0521, over 3307815.10 frames. ], batch size: 134, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:13:15,891 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 09:13:22,846 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.732e+02 2.751e+02 3.128e+02 3.657e+02 9.004e+02, threshold=6.256e+02, percent-clipped=1.0 2023-04-29 09:13:31,362 INFO [train.py:904] (5/8) Epoch 11, batch 2150, loss[loss=0.1716, simple_loss=0.2643, pruned_loss=0.03943, over 17173.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2715, pruned_loss=0.05291, over 3312171.37 frames. ], batch size: 46, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:13:40,645 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 09:13:52,979 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 09:14:42,108 INFO [train.py:904] (5/8) Epoch 11, batch 2200, loss[loss=0.225, simple_loss=0.2981, pruned_loss=0.07597, over 15396.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2708, pruned_loss=0.05241, over 3314297.62 frames. ], batch size: 191, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:02,879 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8586, 3.0850, 2.6176, 4.5545, 3.7430, 4.3090, 1.6139, 3.2197], device='cuda:5'), covar=tensor([0.1273, 0.0598, 0.1026, 0.0220, 0.0281, 0.0348, 0.1430, 0.0692], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0158, 0.0179, 0.0143, 0.0195, 0.0209, 0.0178, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 09:15:11,431 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:15:35,638 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:15:44,095 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.339e+02 2.759e+02 3.363e+02 7.654e+02, threshold=5.518e+02, percent-clipped=1.0 2023-04-29 09:15:51,100 INFO [train.py:904] (5/8) Epoch 11, batch 2250, loss[loss=0.1975, simple_loss=0.3008, pruned_loss=0.04712, over 17299.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2715, pruned_loss=0.05301, over 3316704.68 frames. ], batch size: 52, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:54,371 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:12,411 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:44,335 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:01,667 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103801.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:02,366 INFO [train.py:904] (5/8) Epoch 11, batch 2300, loss[loss=0.1876, simple_loss=0.2678, pruned_loss=0.05371, over 15492.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2719, pruned_loss=0.05319, over 3324896.01 frames. ], batch size: 190, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:17:02,657 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:35,763 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:42,084 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1762, 2.5479, 2.1151, 2.3379, 2.9334, 2.6558, 3.2208, 3.1491], device='cuda:5'), covar=tensor([0.0133, 0.0245, 0.0365, 0.0297, 0.0159, 0.0248, 0.0179, 0.0160], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0208, 0.0203, 0.0203, 0.0207, 0.0207, 0.0215, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:17:42,285 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 09:17:48,570 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103836.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:18:02,533 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.457e+02 2.837e+02 3.610e+02 7.865e+02, threshold=5.674e+02, percent-clipped=5.0 2023-04-29 09:18:11,653 INFO [train.py:904] (5/8) Epoch 11, batch 2350, loss[loss=0.2013, simple_loss=0.2739, pruned_loss=0.06438, over 16815.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2716, pruned_loss=0.05369, over 3328966.00 frames. ], batch size: 102, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:18:45,235 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:19:19,668 INFO [train.py:904] (5/8) Epoch 11, batch 2400, loss[loss=0.1982, simple_loss=0.2937, pruned_loss=0.05136, over 17046.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2723, pruned_loss=0.0543, over 3328962.26 frames. ], batch size: 53, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:20:09,869 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 09:20:20,791 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.318e+02 2.704e+02 3.589e+02 9.063e+02, threshold=5.409e+02, percent-clipped=3.0 2023-04-29 09:20:28,641 INFO [train.py:904] (5/8) Epoch 11, batch 2450, loss[loss=0.1865, simple_loss=0.2596, pruned_loss=0.05668, over 16749.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2737, pruned_loss=0.05406, over 3315865.74 frames. ], batch size: 83, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:20:33,729 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9767, 5.4610, 5.6170, 5.4066, 5.4432, 6.0112, 5.5888, 5.3304], device='cuda:5'), covar=tensor([0.0890, 0.1738, 0.1881, 0.1962, 0.2544, 0.0840, 0.1159, 0.2230], device='cuda:5'), in_proj_covar=tensor([0.0358, 0.0501, 0.0545, 0.0439, 0.0576, 0.0571, 0.0428, 0.0588], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 09:21:22,271 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:32,660 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:42,826 INFO [train.py:904] (5/8) Epoch 11, batch 2500, loss[loss=0.212, simple_loss=0.2775, pruned_loss=0.07328, over 16746.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2732, pruned_loss=0.05357, over 3315366.41 frames. ], batch size: 124, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:11,708 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:22:43,747 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.442e+02 2.893e+02 3.416e+02 6.492e+02, threshold=5.787e+02, percent-clipped=3.0 2023-04-29 09:22:47,596 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:49,921 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:50,611 INFO [train.py:904] (5/8) Epoch 11, batch 2550, loss[loss=0.2087, simple_loss=0.2795, pruned_loss=0.06895, over 16664.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2732, pruned_loss=0.05367, over 3322868.81 frames. ], batch size: 134, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:23:00,472 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:23:15,999 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:23:38,186 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 09:23:47,230 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7004, 2.5348, 2.1744, 2.3779, 2.8862, 2.6896, 3.5070, 3.1605], device='cuda:5'), covar=tensor([0.0075, 0.0277, 0.0349, 0.0325, 0.0185, 0.0276, 0.0150, 0.0179], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0208, 0.0201, 0.0202, 0.0206, 0.0206, 0.0215, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:23:51,820 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:23:59,583 INFO [train.py:904] (5/8) Epoch 11, batch 2600, loss[loss=0.1958, simple_loss=0.2704, pruned_loss=0.06063, over 16798.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2726, pruned_loss=0.05304, over 3315721.89 frames. ], batch size: 134, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:24:12,075 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:24:26,719 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:25:01,155 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.655e+02 3.049e+02 3.595e+02 6.269e+02, threshold=6.098e+02, percent-clipped=2.0 2023-04-29 09:25:10,386 INFO [train.py:904] (5/8) Epoch 11, batch 2650, loss[loss=0.1929, simple_loss=0.2867, pruned_loss=0.04958, over 16613.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2731, pruned_loss=0.05299, over 3312986.72 frames. ], batch size: 62, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:26:07,002 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7567, 4.0421, 4.1499, 2.8275, 3.6590, 4.1058, 3.8407, 2.3295], device='cuda:5'), covar=tensor([0.0358, 0.0124, 0.0032, 0.0285, 0.0077, 0.0064, 0.0050, 0.0351], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0125, 0.0079, 0.0089, 0.0077, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 09:26:18,862 INFO [train.py:904] (5/8) Epoch 11, batch 2700, loss[loss=0.1873, simple_loss=0.2716, pruned_loss=0.05147, over 16917.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2729, pruned_loss=0.05186, over 3322198.13 frames. ], batch size: 90, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:27:00,703 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:27:19,009 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.334e+02 2.747e+02 3.753e+02 9.915e+02, threshold=5.495e+02, percent-clipped=4.0 2023-04-29 09:27:27,265 INFO [train.py:904] (5/8) Epoch 11, batch 2750, loss[loss=0.1844, simple_loss=0.2678, pruned_loss=0.05048, over 16379.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2728, pruned_loss=0.05158, over 3328005.22 frames. ], batch size: 165, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:28:16,478 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1166, 4.3383, 3.4051, 2.4694, 3.2635, 2.8486, 4.8553, 4.0000], device='cuda:5'), covar=tensor([0.2388, 0.0690, 0.1411, 0.2171, 0.2238, 0.1627, 0.0307, 0.0985], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0257, 0.0282, 0.0277, 0.0285, 0.0222, 0.0268, 0.0301], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:28:36,573 INFO [train.py:904] (5/8) Epoch 11, batch 2800, loss[loss=0.1706, simple_loss=0.267, pruned_loss=0.03708, over 17109.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2728, pruned_loss=0.05149, over 3333135.15 frames. ], batch size: 47, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:37,341 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.406e+02 2.849e+02 3.526e+02 7.229e+02, threshold=5.698e+02, percent-clipped=5.0 2023-04-29 09:29:37,661 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:29:44,362 INFO [train.py:904] (5/8) Epoch 11, batch 2850, loss[loss=0.1731, simple_loss=0.2661, pruned_loss=0.04004, over 17037.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2723, pruned_loss=0.05182, over 3321860.78 frames. ], batch size: 55, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:45,816 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:17,605 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5867, 4.5562, 4.5378, 4.0337, 4.5139, 1.9312, 4.3124, 4.3214], device='cuda:5'), covar=tensor([0.0076, 0.0069, 0.0115, 0.0254, 0.0070, 0.1977, 0.0107, 0.0142], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0121, 0.0170, 0.0160, 0.0139, 0.0180, 0.0157, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:30:41,717 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9685, 5.4182, 5.6285, 5.3559, 5.4266, 5.9859, 5.4698, 5.2333], device='cuda:5'), covar=tensor([0.0918, 0.1971, 0.1869, 0.2048, 0.2558, 0.0967, 0.1242, 0.2271], device='cuda:5'), in_proj_covar=tensor([0.0353, 0.0499, 0.0542, 0.0434, 0.0576, 0.0568, 0.0427, 0.0584], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 09:30:44,702 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:52,879 INFO [train.py:904] (5/8) Epoch 11, batch 2900, loss[loss=0.1731, simple_loss=0.2462, pruned_loss=0.05002, over 16532.00 frames. ], tot_loss[loss=0.187, simple_loss=0.271, pruned_loss=0.05153, over 3329050.50 frames. ], batch size: 75, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:30:57,324 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:20,084 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:52,197 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104444.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:55,139 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.712e+02 2.403e+02 2.933e+02 3.429e+02 6.056e+02, threshold=5.866e+02, percent-clipped=1.0 2023-04-29 09:32:04,106 INFO [train.py:904] (5/8) Epoch 11, batch 2950, loss[loss=0.184, simple_loss=0.2658, pruned_loss=0.05109, over 17226.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2708, pruned_loss=0.05241, over 3323548.50 frames. ], batch size: 44, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:32:21,267 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7756, 4.0300, 3.1410, 2.2739, 2.9023, 2.5380, 4.2850, 3.8757], device='cuda:5'), covar=tensor([0.2412, 0.0690, 0.1427, 0.2244, 0.2237, 0.1639, 0.0429, 0.0942], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0258, 0.0282, 0.0275, 0.0284, 0.0222, 0.0269, 0.0301], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:32:28,532 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:33:12,382 INFO [train.py:904] (5/8) Epoch 11, batch 3000, loss[loss=0.2046, simple_loss=0.2811, pruned_loss=0.06405, over 16828.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2711, pruned_loss=0.05285, over 3321894.52 frames. ], batch size: 96, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:33:12,383 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 09:33:22,064 INFO [train.py:938] (5/8) Epoch 11, validation: loss=0.1413, simple_loss=0.2475, pruned_loss=0.01754, over 944034.00 frames. 2023-04-29 09:33:22,064 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 09:34:04,335 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:34:20,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.544e+02 3.030e+02 3.628e+02 6.112e+02, threshold=6.060e+02, percent-clipped=1.0 2023-04-29 09:34:30,337 INFO [train.py:904] (5/8) Epoch 11, batch 3050, loss[loss=0.1652, simple_loss=0.2504, pruned_loss=0.03997, over 17189.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2709, pruned_loss=0.05274, over 3326899.87 frames. ], batch size: 43, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:35:07,353 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:35:37,119 INFO [train.py:904] (5/8) Epoch 11, batch 3100, loss[loss=0.2223, simple_loss=0.2914, pruned_loss=0.07655, over 16711.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.271, pruned_loss=0.05328, over 3328299.43 frames. ], batch size: 134, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:03,737 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-29 09:36:26,420 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2073, 4.0570, 4.2514, 4.4196, 4.5316, 4.1004, 4.2491, 4.4844], device='cuda:5'), covar=tensor([0.1329, 0.0909, 0.1278, 0.0629, 0.0512, 0.1148, 0.1629, 0.0568], device='cuda:5'), in_proj_covar=tensor([0.0551, 0.0689, 0.0850, 0.0704, 0.0530, 0.0541, 0.0544, 0.0626], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:36:39,260 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.673e+02 2.465e+02 3.037e+02 3.411e+02 8.178e+02, threshold=6.075e+02, percent-clipped=4.0 2023-04-29 09:36:39,627 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:36:44,067 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 09:36:47,525 INFO [train.py:904] (5/8) Epoch 11, batch 3150, loss[loss=0.1871, simple_loss=0.2623, pruned_loss=0.056, over 16888.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2693, pruned_loss=0.05243, over 3327470.17 frames. ], batch size: 116, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:49,897 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:18,667 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7644, 2.5441, 2.2370, 2.4322, 2.9099, 2.8152, 3.6366, 3.2005], device='cuda:5'), covar=tensor([0.0080, 0.0309, 0.0349, 0.0315, 0.0198, 0.0256, 0.0147, 0.0181], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0208, 0.0202, 0.0203, 0.0208, 0.0205, 0.0217, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:37:32,704 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8161, 3.7186, 3.8590, 4.0125, 4.0746, 3.6424, 3.8562, 4.0893], device='cuda:5'), covar=tensor([0.1238, 0.0896, 0.1122, 0.0554, 0.0532, 0.1941, 0.1475, 0.0593], device='cuda:5'), in_proj_covar=tensor([0.0556, 0.0696, 0.0860, 0.0710, 0.0536, 0.0545, 0.0550, 0.0632], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:37:46,969 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:56,591 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:57,456 INFO [train.py:904] (5/8) Epoch 11, batch 3200, loss[loss=0.1556, simple_loss=0.2373, pruned_loss=0.037, over 16773.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2686, pruned_loss=0.05215, over 3318511.75 frames. ], batch size: 39, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:38:01,281 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:38:57,881 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.472e+02 2.965e+02 3.891e+02 6.936e+02, threshold=5.930e+02, percent-clipped=3.0 2023-04-29 09:39:06,602 INFO [train.py:904] (5/8) Epoch 11, batch 3250, loss[loss=0.145, simple_loss=0.2237, pruned_loss=0.03319, over 16986.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2685, pruned_loss=0.05181, over 3323476.19 frames. ], batch size: 41, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:39:08,059 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:39:40,611 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 09:39:46,216 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-29 09:39:56,435 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7875, 4.9070, 5.3208, 5.3039, 5.3488, 5.0352, 4.8608, 4.7292], device='cuda:5'), covar=tensor([0.0461, 0.0558, 0.0464, 0.0571, 0.0587, 0.0460, 0.1215, 0.0443], device='cuda:5'), in_proj_covar=tensor([0.0340, 0.0356, 0.0354, 0.0334, 0.0401, 0.0371, 0.0478, 0.0300], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 09:40:15,748 INFO [train.py:904] (5/8) Epoch 11, batch 3300, loss[loss=0.1805, simple_loss=0.2637, pruned_loss=0.04864, over 16852.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2697, pruned_loss=0.05211, over 3326726.37 frames. ], batch size: 96, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:40:43,088 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8516, 2.8166, 2.4873, 2.6175, 3.0375, 2.9301, 3.6243, 3.3925], device='cuda:5'), covar=tensor([0.0072, 0.0282, 0.0325, 0.0318, 0.0222, 0.0252, 0.0190, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0209, 0.0202, 0.0202, 0.0208, 0.0205, 0.0218, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:41:16,323 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.335e+02 2.858e+02 3.500e+02 6.085e+02, threshold=5.716e+02, percent-clipped=1.0 2023-04-29 09:41:24,651 INFO [train.py:904] (5/8) Epoch 11, batch 3350, loss[loss=0.2321, simple_loss=0.3015, pruned_loss=0.08137, over 16271.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2704, pruned_loss=0.05235, over 3327087.34 frames. ], batch size: 165, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:41:25,113 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6855, 3.0320, 2.8022, 5.0217, 4.1857, 4.5494, 1.4658, 3.3311], device='cuda:5'), covar=tensor([0.1286, 0.0656, 0.1006, 0.0145, 0.0235, 0.0315, 0.1519, 0.0620], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0159, 0.0179, 0.0149, 0.0200, 0.0211, 0.0178, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 09:42:33,963 INFO [train.py:904] (5/8) Epoch 11, batch 3400, loss[loss=0.2218, simple_loss=0.2984, pruned_loss=0.07259, over 16388.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2705, pruned_loss=0.05249, over 3321038.11 frames. ], batch size: 165, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:43:18,790 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6973, 6.0731, 5.7778, 5.9055, 5.4967, 5.2759, 5.4854, 6.1779], device='cuda:5'), covar=tensor([0.1141, 0.0759, 0.0936, 0.0599, 0.0753, 0.0590, 0.0902, 0.0776], device='cuda:5'), in_proj_covar=tensor([0.0567, 0.0703, 0.0584, 0.0494, 0.0442, 0.0452, 0.0589, 0.0539], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:43:33,851 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.318e+02 2.811e+02 3.324e+02 7.497e+02, threshold=5.622e+02, percent-clipped=1.0 2023-04-29 09:43:41,674 INFO [train.py:904] (5/8) Epoch 11, batch 3450, loss[loss=0.1765, simple_loss=0.257, pruned_loss=0.04797, over 16479.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2684, pruned_loss=0.05193, over 3311246.54 frames. ], batch size: 68, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:43:42,205 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6132, 2.6326, 1.8231, 2.6911, 2.1501, 2.7778, 2.0657, 2.3351], device='cuda:5'), covar=tensor([0.0247, 0.0348, 0.1280, 0.0257, 0.0630, 0.0412, 0.1084, 0.0566], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0169, 0.0188, 0.0136, 0.0169, 0.0214, 0.0195, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 09:44:39,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7539, 2.7024, 2.3069, 2.5836, 2.9734, 2.7980, 3.6070, 3.3099], device='cuda:5'), covar=tensor([0.0088, 0.0285, 0.0343, 0.0312, 0.0204, 0.0276, 0.0148, 0.0183], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0207, 0.0201, 0.0201, 0.0207, 0.0204, 0.0217, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:44:52,767 INFO [train.py:904] (5/8) Epoch 11, batch 3500, loss[loss=0.1734, simple_loss=0.2642, pruned_loss=0.0413, over 16512.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2664, pruned_loss=0.05103, over 3315280.17 frames. ], batch size: 68, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:45:55,147 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.324e+02 2.297e+02 2.659e+02 3.238e+02 6.959e+02, threshold=5.317e+02, percent-clipped=2.0 2023-04-29 09:46:03,259 INFO [train.py:904] (5/8) Epoch 11, batch 3550, loss[loss=0.177, simple_loss=0.2557, pruned_loss=0.04912, over 16789.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2651, pruned_loss=0.05016, over 3315411.65 frames. ], batch size: 83, lr: 6.27e-03, grad_scale: 4.0 2023-04-29 09:46:18,341 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8633, 4.9783, 5.4241, 5.3700, 5.4049, 5.0494, 5.0144, 4.8127], device='cuda:5'), covar=tensor([0.0304, 0.0525, 0.0333, 0.0459, 0.0372, 0.0290, 0.0798, 0.0377], device='cuda:5'), in_proj_covar=tensor([0.0342, 0.0358, 0.0359, 0.0337, 0.0404, 0.0374, 0.0479, 0.0303], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 09:47:12,608 INFO [train.py:904] (5/8) Epoch 11, batch 3600, loss[loss=0.1701, simple_loss=0.2646, pruned_loss=0.03784, over 17133.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2644, pruned_loss=0.05041, over 3310694.84 frames. ], batch size: 48, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:48:17,992 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.381e+02 2.999e+02 3.559e+02 5.389e+02, threshold=5.997e+02, percent-clipped=2.0 2023-04-29 09:48:24,358 INFO [train.py:904] (5/8) Epoch 11, batch 3650, loss[loss=0.1532, simple_loss=0.237, pruned_loss=0.03477, over 16957.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2633, pruned_loss=0.05057, over 3313933.50 frames. ], batch size: 41, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:48:25,347 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7883, 3.6183, 3.7607, 3.5777, 3.6467, 4.1210, 3.8326, 3.4534], device='cuda:5'), covar=tensor([0.1855, 0.2299, 0.1651, 0.2625, 0.2860, 0.1749, 0.1374, 0.2813], device='cuda:5'), in_proj_covar=tensor([0.0352, 0.0496, 0.0537, 0.0430, 0.0570, 0.0566, 0.0426, 0.0583], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 09:49:37,407 INFO [train.py:904] (5/8) Epoch 11, batch 3700, loss[loss=0.2076, simple_loss=0.2843, pruned_loss=0.06543, over 11379.00 frames. ], tot_loss[loss=0.184, simple_loss=0.263, pruned_loss=0.05257, over 3290300.57 frames. ], batch size: 246, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:50:41,015 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.627e+02 2.380e+02 2.864e+02 3.354e+02 5.031e+02, threshold=5.728e+02, percent-clipped=0.0 2023-04-29 09:50:48,720 INFO [train.py:904] (5/8) Epoch 11, batch 3750, loss[loss=0.1786, simple_loss=0.2466, pruned_loss=0.05535, over 16926.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2643, pruned_loss=0.05413, over 3276532.48 frames. ], batch size: 109, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:51:57,093 INFO [train.py:904] (5/8) Epoch 11, batch 3800, loss[loss=0.2177, simple_loss=0.2986, pruned_loss=0.06837, over 12407.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2652, pruned_loss=0.05557, over 3274649.46 frames. ], batch size: 248, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:52:43,204 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7034, 2.6476, 2.0896, 2.5435, 3.1303, 2.7417, 3.4661, 3.3750], device='cuda:5'), covar=tensor([0.0042, 0.0260, 0.0383, 0.0311, 0.0152, 0.0282, 0.0105, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0205, 0.0199, 0.0199, 0.0204, 0.0202, 0.0213, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:53:01,743 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8432, 3.0265, 2.5802, 4.3824, 3.6523, 4.2419, 1.4751, 3.0196], device='cuda:5'), covar=tensor([0.1262, 0.0533, 0.1033, 0.0153, 0.0207, 0.0329, 0.1444, 0.0739], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0157, 0.0179, 0.0147, 0.0199, 0.0210, 0.0178, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 09:53:02,334 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.313e+02 2.608e+02 3.308e+02 8.159e+02, threshold=5.217e+02, percent-clipped=3.0 2023-04-29 09:53:08,977 INFO [train.py:904] (5/8) Epoch 11, batch 3850, loss[loss=0.1871, simple_loss=0.2564, pruned_loss=0.05891, over 16750.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2655, pruned_loss=0.05584, over 3282361.34 frames. ], batch size: 83, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:53:36,429 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.77 vs. limit=5.0 2023-04-29 09:54:00,703 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105388.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:54:19,878 INFO [train.py:904] (5/8) Epoch 11, batch 3900, loss[loss=0.1749, simple_loss=0.2466, pruned_loss=0.05161, over 16810.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2652, pruned_loss=0.0562, over 3284543.76 frames. ], batch size: 83, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:54:29,218 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6294, 4.6084, 4.5964, 4.0462, 4.5562, 1.7971, 4.3623, 4.3847], device='cuda:5'), covar=tensor([0.0089, 0.0074, 0.0138, 0.0287, 0.0087, 0.2266, 0.0115, 0.0166], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0121, 0.0170, 0.0161, 0.0140, 0.0181, 0.0158, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:54:32,908 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 09:54:50,674 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 09:55:25,286 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.279e+02 2.712e+02 3.395e+02 5.874e+02, threshold=5.424e+02, percent-clipped=4.0 2023-04-29 09:55:28,910 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 09:55:31,854 INFO [train.py:904] (5/8) Epoch 11, batch 3950, loss[loss=0.1968, simple_loss=0.2787, pruned_loss=0.05746, over 12421.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2646, pruned_loss=0.05653, over 3281300.81 frames. ], batch size: 247, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:56:21,783 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:56:44,052 INFO [train.py:904] (5/8) Epoch 11, batch 4000, loss[loss=0.2166, simple_loss=0.2891, pruned_loss=0.07199, over 16696.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2647, pruned_loss=0.05697, over 3283961.09 frames. ], batch size: 134, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:56:44,567 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:56:56,322 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 09:57:14,773 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0956, 5.0274, 4.9550, 4.3303, 5.0277, 1.9446, 4.7179, 4.7392], device='cuda:5'), covar=tensor([0.0050, 0.0042, 0.0097, 0.0286, 0.0046, 0.2173, 0.0100, 0.0144], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0121, 0.0169, 0.0161, 0.0140, 0.0180, 0.0158, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 09:57:29,921 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6079, 3.8400, 2.1020, 4.1921, 2.8648, 4.2479, 2.3447, 2.8565], device='cuda:5'), covar=tensor([0.0206, 0.0272, 0.1514, 0.0102, 0.0691, 0.0239, 0.1387, 0.0700], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0167, 0.0186, 0.0133, 0.0166, 0.0210, 0.0193, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 09:57:31,044 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9792, 4.0844, 4.3992, 4.3539, 4.3351, 4.0561, 4.1070, 3.9645], device='cuda:5'), covar=tensor([0.0286, 0.0534, 0.0376, 0.0450, 0.0429, 0.0349, 0.0738, 0.0478], device='cuda:5'), in_proj_covar=tensor([0.0333, 0.0348, 0.0348, 0.0329, 0.0395, 0.0363, 0.0468, 0.0293], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 09:57:48,233 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.126e+02 2.510e+02 3.113e+02 5.391e+02, threshold=5.020e+02, percent-clipped=0.0 2023-04-29 09:57:48,826 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:55,347 INFO [train.py:904] (5/8) Epoch 11, batch 4050, loss[loss=0.1859, simple_loss=0.2707, pruned_loss=0.05058, over 15533.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2651, pruned_loss=0.05605, over 3272566.35 frames. ], batch size: 191, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:58:11,844 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:59:08,734 INFO [train.py:904] (5/8) Epoch 11, batch 4100, loss[loss=0.2096, simple_loss=0.2944, pruned_loss=0.0624, over 16871.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2662, pruned_loss=0.05514, over 3275317.12 frames. ], batch size: 116, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:59:28,664 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6001, 4.6475, 4.4341, 4.2069, 4.1344, 4.5610, 4.3100, 4.2283], device='cuda:5'), covar=tensor([0.0519, 0.0342, 0.0225, 0.0230, 0.0820, 0.0367, 0.0473, 0.0547], device='cuda:5'), in_proj_covar=tensor([0.0257, 0.0323, 0.0302, 0.0280, 0.0324, 0.0322, 0.0205, 0.0350], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:00:18,930 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 2.197e+02 2.575e+02 3.230e+02 7.189e+02, threshold=5.150e+02, percent-clipped=4.0 2023-04-29 10:00:26,725 INFO [train.py:904] (5/8) Epoch 11, batch 4150, loss[loss=0.2367, simple_loss=0.3262, pruned_loss=0.07357, over 16249.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2743, pruned_loss=0.0582, over 3240916.74 frames. ], batch size: 165, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:01:13,265 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0613, 3.6817, 3.5665, 2.1566, 3.1996, 3.5172, 3.3301, 2.0788], device='cuda:5'), covar=tensor([0.0468, 0.0027, 0.0043, 0.0374, 0.0076, 0.0100, 0.0063, 0.0357], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0069, 0.0069, 0.0123, 0.0077, 0.0088, 0.0076, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 10:01:44,702 INFO [train.py:904] (5/8) Epoch 11, batch 4200, loss[loss=0.207, simple_loss=0.299, pruned_loss=0.0575, over 16453.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2812, pruned_loss=0.06019, over 3208353.01 frames. ], batch size: 68, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:02:49,765 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:02:53,623 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.308e+02 2.787e+02 3.444e+02 5.705e+02, threshold=5.573e+02, percent-clipped=4.0 2023-04-29 10:02:59,814 INFO [train.py:904] (5/8) Epoch 11, batch 4250, loss[loss=0.2195, simple_loss=0.315, pruned_loss=0.06199, over 16209.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2846, pruned_loss=0.06038, over 3174340.13 frames. ], batch size: 165, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:03:40,530 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 10:03:59,350 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:04:08,520 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:04:12,692 INFO [train.py:904] (5/8) Epoch 11, batch 4300, loss[loss=0.187, simple_loss=0.2869, pruned_loss=0.04354, over 16836.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2849, pruned_loss=0.05859, over 3190713.15 frames. ], batch size: 102, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:04:30,312 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3043, 3.2004, 2.5314, 2.0879, 2.3079, 2.1238, 3.2433, 3.0593], device='cuda:5'), covar=tensor([0.2610, 0.0714, 0.1627, 0.2105, 0.2099, 0.1765, 0.0550, 0.1005], device='cuda:5'), in_proj_covar=tensor([0.0300, 0.0254, 0.0282, 0.0276, 0.0286, 0.0221, 0.0267, 0.0296], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:04:35,310 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8227, 3.3535, 3.1642, 1.8404, 2.7873, 1.9775, 3.2999, 3.3732], device='cuda:5'), covar=tensor([0.0262, 0.0628, 0.0594, 0.1915, 0.0812, 0.0997, 0.0611, 0.0869], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0145, 0.0156, 0.0144, 0.0135, 0.0123, 0.0137, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 10:05:07,440 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8336, 3.2715, 3.1992, 1.9169, 2.8718, 3.1882, 3.0603, 1.8445], device='cuda:5'), covar=tensor([0.0448, 0.0034, 0.0038, 0.0357, 0.0078, 0.0071, 0.0060, 0.0353], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0125, 0.0078, 0.0089, 0.0077, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 10:05:11,681 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:17,309 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.245e+02 2.727e+02 3.248e+02 5.954e+02, threshold=5.453e+02, percent-clipped=2.0 2023-04-29 10:05:25,485 INFO [train.py:904] (5/8) Epoch 11, batch 4350, loss[loss=0.2308, simple_loss=0.3049, pruned_loss=0.07834, over 11572.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2883, pruned_loss=0.05945, over 3185702.78 frames. ], batch size: 246, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:05:27,972 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:30,281 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6288, 4.7186, 4.9528, 4.7594, 4.8203, 5.3645, 4.9850, 4.6261], device='cuda:5'), covar=tensor([0.0937, 0.1517, 0.1388, 0.1540, 0.2009, 0.0829, 0.1103, 0.2178], device='cuda:5'), in_proj_covar=tensor([0.0341, 0.0477, 0.0517, 0.0415, 0.0548, 0.0542, 0.0410, 0.0561], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 10:05:34,728 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:37,383 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:06:01,131 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0609, 3.9007, 3.8647, 2.2239, 3.3112, 3.7789, 3.5453, 2.1441], device='cuda:5'), covar=tensor([0.0471, 0.0025, 0.0028, 0.0384, 0.0078, 0.0064, 0.0056, 0.0345], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0125, 0.0078, 0.0089, 0.0077, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 10:06:15,791 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 10:06:38,346 INFO [train.py:904] (5/8) Epoch 11, batch 4400, loss[loss=0.2192, simple_loss=0.3106, pruned_loss=0.06394, over 16780.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2911, pruned_loss=0.06048, over 3198848.74 frames. ], batch size: 124, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:06:49,493 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6791, 2.6965, 1.8253, 2.8289, 2.2321, 2.8418, 2.0637, 2.3652], device='cuda:5'), covar=tensor([0.0276, 0.0392, 0.1310, 0.0171, 0.0636, 0.0486, 0.1164, 0.0580], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0164, 0.0185, 0.0128, 0.0165, 0.0207, 0.0193, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 10:06:52,517 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0724, 2.5722, 2.6512, 1.8218, 2.8309, 2.8508, 2.4240, 2.3765], device='cuda:5'), covar=tensor([0.0587, 0.0214, 0.0159, 0.0898, 0.0077, 0.0140, 0.0405, 0.0392], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0101, 0.0089, 0.0138, 0.0070, 0.0102, 0.0121, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 10:07:40,715 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.548e+02 2.924e+02 3.370e+02 5.572e+02, threshold=5.847e+02, percent-clipped=1.0 2023-04-29 10:07:48,900 INFO [train.py:904] (5/8) Epoch 11, batch 4450, loss[loss=0.1983, simple_loss=0.2896, pruned_loss=0.05353, over 17025.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2946, pruned_loss=0.06196, over 3181973.23 frames. ], batch size: 50, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:09:04,914 INFO [train.py:904] (5/8) Epoch 11, batch 4500, loss[loss=0.2088, simple_loss=0.29, pruned_loss=0.06377, over 16519.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2947, pruned_loss=0.06215, over 3188765.60 frames. ], batch size: 68, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:09:35,493 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5345, 2.2832, 2.3448, 4.3864, 2.0877, 2.7698, 2.4124, 2.5218], device='cuda:5'), covar=tensor([0.0870, 0.3018, 0.2081, 0.0337, 0.3674, 0.1959, 0.2552, 0.2952], device='cuda:5'), in_proj_covar=tensor([0.0365, 0.0394, 0.0329, 0.0323, 0.0415, 0.0455, 0.0357, 0.0466], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:10:05,568 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:10:09,395 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 1.945e+02 2.292e+02 2.586e+02 4.524e+02, threshold=4.584e+02, percent-clipped=0.0 2023-04-29 10:10:17,387 INFO [train.py:904] (5/8) Epoch 11, batch 4550, loss[loss=0.2108, simple_loss=0.3004, pruned_loss=0.06058, over 16387.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2952, pruned_loss=0.06256, over 3211939.76 frames. ], batch size: 146, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:10:53,658 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:15,422 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:16,082 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 10:11:29,157 INFO [train.py:904] (5/8) Epoch 11, batch 4600, loss[loss=0.193, simple_loss=0.2787, pruned_loss=0.0536, over 16549.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2959, pruned_loss=0.0629, over 3210780.39 frames. ], batch size: 62, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:12:18,630 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8950, 4.8457, 4.6085, 4.0060, 4.7257, 1.6437, 4.4936, 4.3548], device='cuda:5'), covar=tensor([0.0040, 0.0037, 0.0097, 0.0260, 0.0045, 0.2330, 0.0070, 0.0135], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0116, 0.0163, 0.0156, 0.0133, 0.0175, 0.0150, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:12:22,361 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:28,101 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:33,820 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.42 vs. limit=5.0 2023-04-29 10:12:35,553 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.926e+02 2.251e+02 2.695e+02 4.536e+02, threshold=4.502e+02, percent-clipped=0.0 2023-04-29 10:12:37,727 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:41,948 INFO [train.py:904] (5/8) Epoch 11, batch 4650, loss[loss=0.1904, simple_loss=0.2746, pruned_loss=0.05308, over 16433.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2947, pruned_loss=0.06278, over 3189467.18 frames. ], batch size: 68, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:12:47,378 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:51,881 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:36,514 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8954, 3.9799, 4.2775, 4.2563, 4.2368, 3.9791, 3.9710, 3.8612], device='cuda:5'), covar=tensor([0.0276, 0.0437, 0.0315, 0.0359, 0.0384, 0.0336, 0.0780, 0.0448], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0325, 0.0329, 0.0310, 0.0372, 0.0345, 0.0446, 0.0279], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 10:13:37,559 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:55,750 INFO [train.py:904] (5/8) Epoch 11, batch 4700, loss[loss=0.2118, simple_loss=0.2949, pruned_loss=0.06436, over 16215.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2917, pruned_loss=0.06139, over 3189713.02 frames. ], batch size: 165, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:14:01,917 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:14:03,555 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 10:14:25,828 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 10:14:26,859 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3415, 3.6617, 3.6400, 2.0268, 3.0060, 2.5679, 3.6919, 3.8144], device='cuda:5'), covar=tensor([0.0238, 0.0642, 0.0535, 0.1783, 0.0802, 0.0847, 0.0597, 0.0892], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0145, 0.0156, 0.0143, 0.0135, 0.0124, 0.0136, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 10:15:01,680 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.142e+02 2.451e+02 2.986e+02 5.737e+02, threshold=4.902e+02, percent-clipped=3.0 2023-04-29 10:15:05,769 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-29 10:15:09,055 INFO [train.py:904] (5/8) Epoch 11, batch 4750, loss[loss=0.186, simple_loss=0.2709, pruned_loss=0.05054, over 16711.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2886, pruned_loss=0.05983, over 3192384.99 frames. ], batch size: 124, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:15:32,889 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:15:37,232 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7498, 4.7224, 4.7166, 3.8947, 4.6835, 1.7021, 4.4201, 4.5489], device='cuda:5'), covar=tensor([0.0098, 0.0082, 0.0104, 0.0457, 0.0078, 0.2286, 0.0123, 0.0174], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0116, 0.0164, 0.0157, 0.0134, 0.0177, 0.0151, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:16:22,063 INFO [train.py:904] (5/8) Epoch 11, batch 4800, loss[loss=0.2212, simple_loss=0.2989, pruned_loss=0.07172, over 16852.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2854, pruned_loss=0.05857, over 3174763.67 frames. ], batch size: 116, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:17:02,221 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:12,269 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:27,383 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.167e+02 2.529e+02 2.965e+02 6.623e+02, threshold=5.058e+02, percent-clipped=3.0 2023-04-29 10:17:35,197 INFO [train.py:904] (5/8) Epoch 11, batch 4850, loss[loss=0.2429, simple_loss=0.3231, pruned_loss=0.08137, over 11901.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2866, pruned_loss=0.05845, over 3153946.40 frames. ], batch size: 248, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:17:52,657 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4296, 3.2243, 2.6529, 2.1254, 2.2485, 2.1321, 3.2833, 3.0400], device='cuda:5'), covar=tensor([0.2471, 0.0722, 0.1519, 0.2320, 0.2239, 0.1769, 0.0461, 0.0971], device='cuda:5'), in_proj_covar=tensor([0.0300, 0.0252, 0.0280, 0.0275, 0.0282, 0.0219, 0.0263, 0.0294], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:18:01,097 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 10:18:22,295 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 10:18:29,455 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 10:18:38,706 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106397.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:18:45,473 INFO [train.py:904] (5/8) Epoch 11, batch 4900, loss[loss=0.2012, simple_loss=0.2944, pruned_loss=0.05401, over 15490.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2855, pruned_loss=0.05684, over 3137213.42 frames. ], batch size: 190, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:19:29,962 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:19:50,198 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.157e+02 2.578e+02 2.904e+02 4.407e+02, threshold=5.156e+02, percent-clipped=0.0 2023-04-29 10:19:51,826 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106448.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:19:56,934 INFO [train.py:904] (5/8) Epoch 11, batch 4950, loss[loss=0.197, simple_loss=0.2834, pruned_loss=0.05533, over 17049.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2844, pruned_loss=0.05553, over 3161017.18 frames. ], batch size: 55, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:20:01,977 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:02,990 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:59,801 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9410, 2.5228, 2.3918, 4.6432, 2.4013, 3.1381, 2.5698, 2.8551], device='cuda:5'), covar=tensor([0.0781, 0.3007, 0.2087, 0.0312, 0.3209, 0.1878, 0.2642, 0.2518], device='cuda:5'), in_proj_covar=tensor([0.0363, 0.0393, 0.0327, 0.0319, 0.0411, 0.0452, 0.0355, 0.0460], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:21:00,663 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106496.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:08,412 INFO [train.py:904] (5/8) Epoch 11, batch 5000, loss[loss=0.1928, simple_loss=0.2842, pruned_loss=0.05068, over 16739.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.286, pruned_loss=0.05549, over 3176180.79 frames. ], batch size: 124, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:21:10,910 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:28,693 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:22:14,243 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.415e+02 2.751e+02 3.160e+02 5.367e+02, threshold=5.502e+02, percent-clipped=2.0 2023-04-29 10:22:21,435 INFO [train.py:904] (5/8) Epoch 11, batch 5050, loss[loss=0.2055, simple_loss=0.2979, pruned_loss=0.05652, over 16672.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2864, pruned_loss=0.0551, over 3186904.98 frames. ], batch size: 124, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:22:42,510 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3435, 3.2822, 3.3667, 3.4813, 3.5185, 3.2559, 3.5050, 3.5779], device='cuda:5'), covar=tensor([0.1110, 0.0866, 0.1023, 0.0596, 0.0572, 0.2277, 0.0868, 0.0624], device='cuda:5'), in_proj_covar=tensor([0.0511, 0.0645, 0.0786, 0.0659, 0.0501, 0.0505, 0.0511, 0.0585], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:23:32,377 INFO [train.py:904] (5/8) Epoch 11, batch 5100, loss[loss=0.1923, simple_loss=0.2703, pruned_loss=0.05715, over 16985.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2847, pruned_loss=0.05441, over 3193000.09 frames. ], batch size: 55, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:23:37,690 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1331, 1.4475, 1.7849, 2.1056, 2.1483, 2.3602, 1.5576, 2.2621], device='cuda:5'), covar=tensor([0.0160, 0.0355, 0.0196, 0.0235, 0.0201, 0.0123, 0.0360, 0.0087], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0173, 0.0155, 0.0161, 0.0167, 0.0124, 0.0171, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 10:24:03,608 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:24:38,779 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.440e+02 2.185e+02 2.500e+02 2.940e+02 7.724e+02, threshold=5.000e+02, percent-clipped=1.0 2023-04-29 10:24:44,417 INFO [train.py:904] (5/8) Epoch 11, batch 5150, loss[loss=0.192, simple_loss=0.2873, pruned_loss=0.04836, over 16727.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2849, pruned_loss=0.05381, over 3187342.09 frames. ], batch size: 83, lr: 6.22e-03, grad_scale: 4.0 2023-04-29 10:25:43,758 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:25:56,057 INFO [train.py:904] (5/8) Epoch 11, batch 5200, loss[loss=0.1908, simple_loss=0.2804, pruned_loss=0.05059, over 16276.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.283, pruned_loss=0.05311, over 3196678.13 frames. ], batch size: 146, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:26:34,208 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2085, 5.3010, 5.6975, 5.6742, 5.6693, 5.2894, 5.2283, 4.8899], device='cuda:5'), covar=tensor([0.0245, 0.0402, 0.0252, 0.0351, 0.0479, 0.0290, 0.0831, 0.0390], device='cuda:5'), in_proj_covar=tensor([0.0321, 0.0330, 0.0333, 0.0319, 0.0379, 0.0353, 0.0453, 0.0282], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 10:26:42,777 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:27:02,477 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3699, 5.3621, 5.2167, 4.5559, 5.2357, 2.0405, 4.9720, 5.1673], device='cuda:5'), covar=tensor([0.0061, 0.0052, 0.0112, 0.0352, 0.0063, 0.2103, 0.0088, 0.0131], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0116, 0.0164, 0.0158, 0.0135, 0.0178, 0.0151, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:27:04,342 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.187e+02 2.709e+02 3.054e+02 5.980e+02, threshold=5.418e+02, percent-clipped=1.0 2023-04-29 10:27:11,358 INFO [train.py:904] (5/8) Epoch 11, batch 5250, loss[loss=0.1704, simple_loss=0.2468, pruned_loss=0.04698, over 17136.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2796, pruned_loss=0.05217, over 3212601.68 frames. ], batch size: 48, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:27:52,597 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:28:08,262 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 10:28:22,267 INFO [train.py:904] (5/8) Epoch 11, batch 5300, loss[loss=0.175, simple_loss=0.2653, pruned_loss=0.04235, over 16323.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2757, pruned_loss=0.05041, over 3225252.07 frames. ], batch size: 165, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:28:32,944 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:29:27,204 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.160e+02 2.565e+02 2.956e+02 5.223e+02, threshold=5.130e+02, percent-clipped=0.0 2023-04-29 10:29:33,908 INFO [train.py:904] (5/8) Epoch 11, batch 5350, loss[loss=0.1989, simple_loss=0.2965, pruned_loss=0.05063, over 16739.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2739, pruned_loss=0.04982, over 3213775.21 frames. ], batch size: 89, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:30:39,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0144, 4.7833, 4.9688, 5.1924, 5.3804, 4.7019, 5.3948, 5.3126], device='cuda:5'), covar=tensor([0.1347, 0.0986, 0.1520, 0.0639, 0.0426, 0.0713, 0.0355, 0.0520], device='cuda:5'), in_proj_covar=tensor([0.0519, 0.0651, 0.0793, 0.0667, 0.0507, 0.0514, 0.0514, 0.0592], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:30:45,870 INFO [train.py:904] (5/8) Epoch 11, batch 5400, loss[loss=0.2123, simple_loss=0.3104, pruned_loss=0.05715, over 16703.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2764, pruned_loss=0.05054, over 3218113.12 frames. ], batch size: 134, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:31:18,190 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:31:54,549 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.177e+02 2.658e+02 3.216e+02 5.881e+02, threshold=5.316e+02, percent-clipped=3.0 2023-04-29 10:32:02,032 INFO [train.py:904] (5/8) Epoch 11, batch 5450, loss[loss=0.2674, simple_loss=0.3454, pruned_loss=0.09467, over 15304.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2807, pruned_loss=0.05291, over 3213891.79 frames. ], batch size: 190, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:32:16,724 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8585, 3.1346, 3.2942, 1.4367, 3.4712, 3.5922, 2.7556, 2.5387], device='cuda:5'), covar=tensor([0.1127, 0.0207, 0.0178, 0.1360, 0.0091, 0.0107, 0.0424, 0.0588], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0098, 0.0087, 0.0137, 0.0069, 0.0099, 0.0119, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 10:32:34,433 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:32:40,273 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8070, 1.6695, 2.2722, 2.7196, 2.5864, 3.1068, 1.7253, 3.1127], device='cuda:5'), covar=tensor([0.0135, 0.0366, 0.0224, 0.0188, 0.0203, 0.0106, 0.0368, 0.0074], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0172, 0.0154, 0.0161, 0.0167, 0.0124, 0.0169, 0.0115], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 10:33:04,089 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:33:19,271 INFO [train.py:904] (5/8) Epoch 11, batch 5500, loss[loss=0.2271, simple_loss=0.3132, pruned_loss=0.07052, over 16816.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2885, pruned_loss=0.05862, over 3183101.40 frames. ], batch size: 102, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:33:53,254 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0302, 4.1194, 4.4772, 4.4460, 4.4507, 4.1528, 4.1466, 3.9983], device='cuda:5'), covar=tensor([0.0316, 0.0469, 0.0361, 0.0408, 0.0427, 0.0359, 0.0849, 0.0527], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0334, 0.0336, 0.0322, 0.0381, 0.0356, 0.0457, 0.0285], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 10:34:18,907 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:34:25,772 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8982, 4.1301, 3.1465, 2.4545, 3.0085, 2.6019, 4.6417, 3.7919], device='cuda:5'), covar=tensor([0.2461, 0.0598, 0.1543, 0.2003, 0.2401, 0.1617, 0.0346, 0.0923], device='cuda:5'), in_proj_covar=tensor([0.0300, 0.0254, 0.0280, 0.0275, 0.0280, 0.0218, 0.0265, 0.0292], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:34:31,627 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.119e+02 3.959e+02 4.735e+02 9.206e+02, threshold=7.917e+02, percent-clipped=14.0 2023-04-29 10:34:37,912 INFO [train.py:904] (5/8) Epoch 11, batch 5550, loss[loss=0.2437, simple_loss=0.3266, pruned_loss=0.0804, over 16252.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2969, pruned_loss=0.06479, over 3150773.73 frames. ], batch size: 165, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:35:02,173 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9281, 3.8304, 3.9699, 4.1333, 4.1975, 3.7937, 4.1682, 4.2250], device='cuda:5'), covar=tensor([0.1341, 0.0957, 0.1249, 0.0563, 0.0566, 0.1371, 0.0631, 0.0552], device='cuda:5'), in_proj_covar=tensor([0.0509, 0.0638, 0.0775, 0.0651, 0.0498, 0.0502, 0.0507, 0.0580], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:35:57,910 INFO [train.py:904] (5/8) Epoch 11, batch 5600, loss[loss=0.2331, simple_loss=0.3161, pruned_loss=0.0751, over 16674.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3024, pruned_loss=0.06934, over 3132310.41 frames. ], batch size: 89, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:36:12,223 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:36:41,817 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 10:37:02,558 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:37:17,302 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.615e+02 3.503e+02 4.339e+02 5.380e+02 1.040e+03, threshold=8.679e+02, percent-clipped=3.0 2023-04-29 10:37:21,720 INFO [train.py:904] (5/8) Epoch 11, batch 5650, loss[loss=0.2494, simple_loss=0.3251, pruned_loss=0.0868, over 15351.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3084, pruned_loss=0.07451, over 3100134.69 frames. ], batch size: 190, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:37:22,235 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:37:32,837 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:38:42,812 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:38:43,547 INFO [train.py:904] (5/8) Epoch 11, batch 5700, loss[loss=0.2465, simple_loss=0.3362, pruned_loss=0.0784, over 16381.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3091, pruned_loss=0.07565, over 3091720.57 frames. ], batch size: 165, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:39:02,911 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:39:37,955 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3213, 3.2015, 3.4611, 1.6956, 3.6733, 3.6906, 2.7821, 2.7750], device='cuda:5'), covar=tensor([0.0736, 0.0246, 0.0230, 0.1163, 0.0077, 0.0122, 0.0446, 0.0410], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0098, 0.0087, 0.0136, 0.0068, 0.0100, 0.0118, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 10:39:59,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.558e+02 3.709e+02 4.582e+02 6.080e+02 1.249e+03, threshold=9.164e+02, percent-clipped=1.0 2023-04-29 10:40:04,433 INFO [train.py:904] (5/8) Epoch 11, batch 5750, loss[loss=0.223, simple_loss=0.3058, pruned_loss=0.07008, over 16546.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3123, pruned_loss=0.07741, over 3070906.04 frames. ], batch size: 75, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:40:31,055 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 10:41:25,727 INFO [train.py:904] (5/8) Epoch 11, batch 5800, loss[loss=0.2554, simple_loss=0.3263, pruned_loss=0.09224, over 11778.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3128, pruned_loss=0.07714, over 3041174.09 frames. ], batch size: 246, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:42:19,094 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-04-29 10:42:39,035 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.001e+02 2.956e+02 3.560e+02 4.660e+02 9.709e+02, threshold=7.120e+02, percent-clipped=1.0 2023-04-29 10:42:43,690 INFO [train.py:904] (5/8) Epoch 11, batch 5850, loss[loss=0.2169, simple_loss=0.2957, pruned_loss=0.06906, over 16833.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.31, pruned_loss=0.07468, over 3060804.36 frames. ], batch size: 116, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:44:05,222 INFO [train.py:904] (5/8) Epoch 11, batch 5900, loss[loss=0.2021, simple_loss=0.3008, pruned_loss=0.05175, over 16472.00 frames. ], tot_loss[loss=0.229, simple_loss=0.31, pruned_loss=0.07402, over 3069324.57 frames. ], batch size: 68, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:44:33,102 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7238, 3.7573, 3.8509, 3.7123, 3.8475, 4.1672, 3.8582, 3.5791], device='cuda:5'), covar=tensor([0.2122, 0.2052, 0.1825, 0.2224, 0.2481, 0.1663, 0.1419, 0.2534], device='cuda:5'), in_proj_covar=tensor([0.0349, 0.0481, 0.0524, 0.0419, 0.0551, 0.0550, 0.0411, 0.0568], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 10:45:21,996 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.026e+02 2.914e+02 3.759e+02 4.619e+02 1.179e+03, threshold=7.519e+02, percent-clipped=2.0 2023-04-29 10:45:26,035 INFO [train.py:904] (5/8) Epoch 11, batch 5950, loss[loss=0.2207, simple_loss=0.3058, pruned_loss=0.06778, over 17069.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3097, pruned_loss=0.0722, over 3077716.20 frames. ], batch size: 55, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:46:40,642 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:46:48,947 INFO [train.py:904] (5/8) Epoch 11, batch 6000, loss[loss=0.2225, simple_loss=0.2976, pruned_loss=0.07372, over 15513.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3084, pruned_loss=0.07173, over 3083044.23 frames. ], batch size: 191, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:46:48,948 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 10:46:59,888 INFO [train.py:938] (5/8) Epoch 11, validation: loss=0.163, simple_loss=0.2761, pruned_loss=0.02492, over 944034.00 frames. 2023-04-29 10:46:59,889 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 10:47:10,240 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:47:13,615 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:48:12,723 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.272e+02 3.327e+02 3.927e+02 4.902e+02 9.680e+02, threshold=7.854e+02, percent-clipped=6.0 2023-04-29 10:48:18,417 INFO [train.py:904] (5/8) Epoch 11, batch 6050, loss[loss=0.2145, simple_loss=0.31, pruned_loss=0.05954, over 16612.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3069, pruned_loss=0.07112, over 3088905.09 frames. ], batch size: 68, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:48:48,379 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0489, 3.1711, 3.2803, 1.5327, 3.3997, 3.4957, 2.7373, 2.5984], device='cuda:5'), covar=tensor([0.0851, 0.0195, 0.0168, 0.1338, 0.0088, 0.0149, 0.0409, 0.0466], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0097, 0.0087, 0.0137, 0.0069, 0.0101, 0.0118, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 10:48:48,416 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:48:58,631 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0554, 2.3619, 2.3356, 2.8204, 2.2114, 3.2660, 1.7261, 2.7178], device='cuda:5'), covar=tensor([0.0993, 0.0516, 0.0951, 0.0124, 0.0154, 0.0346, 0.1242, 0.0604], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0158, 0.0181, 0.0143, 0.0199, 0.0209, 0.0180, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 10:49:11,664 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7233, 2.2071, 1.6513, 2.0182, 2.5849, 2.1619, 2.6778, 2.7915], device='cuda:5'), covar=tensor([0.0106, 0.0318, 0.0445, 0.0348, 0.0175, 0.0308, 0.0150, 0.0163], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0200, 0.0197, 0.0197, 0.0199, 0.0201, 0.0205, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:49:34,934 INFO [train.py:904] (5/8) Epoch 11, batch 6100, loss[loss=0.247, simple_loss=0.3373, pruned_loss=0.07837, over 16246.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.306, pruned_loss=0.06985, over 3092790.38 frames. ], batch size: 165, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:50:17,284 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8094, 2.4649, 2.2282, 3.4014, 2.5072, 3.6276, 1.5258, 2.6294], device='cuda:5'), covar=tensor([0.1159, 0.0641, 0.1153, 0.0149, 0.0190, 0.0424, 0.1382, 0.0793], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0158, 0.0181, 0.0143, 0.0198, 0.0208, 0.0179, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 10:50:37,919 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2098, 4.2743, 4.6525, 4.6251, 4.6302, 4.3329, 4.3332, 4.1694], device='cuda:5'), covar=tensor([0.0312, 0.0511, 0.0355, 0.0414, 0.0464, 0.0349, 0.0902, 0.0470], device='cuda:5'), in_proj_covar=tensor([0.0326, 0.0341, 0.0343, 0.0326, 0.0388, 0.0362, 0.0463, 0.0294], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 10:50:42,784 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5313, 3.5244, 2.7422, 2.0677, 2.3825, 2.1497, 3.5707, 3.3214], device='cuda:5'), covar=tensor([0.2582, 0.0655, 0.1526, 0.2307, 0.2112, 0.1816, 0.0502, 0.0950], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0257, 0.0282, 0.0277, 0.0281, 0.0219, 0.0267, 0.0293], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:50:51,422 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.980e+02 3.646e+02 4.413e+02 7.935e+02, threshold=7.292e+02, percent-clipped=3.0 2023-04-29 10:50:56,537 INFO [train.py:904] (5/8) Epoch 11, batch 6150, loss[loss=0.2082, simple_loss=0.2961, pruned_loss=0.06018, over 16480.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.304, pruned_loss=0.06916, over 3086949.67 frames. ], batch size: 146, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:51:07,579 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:52:14,177 INFO [train.py:904] (5/8) Epoch 11, batch 6200, loss[loss=0.2339, simple_loss=0.3011, pruned_loss=0.08333, over 11593.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3022, pruned_loss=0.06895, over 3080846.14 frames. ], batch size: 250, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:52:42,614 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4627, 2.0958, 1.7434, 1.8992, 2.3985, 2.1424, 2.3985, 2.6303], device='cuda:5'), covar=tensor([0.0117, 0.0268, 0.0362, 0.0342, 0.0169, 0.0253, 0.0141, 0.0174], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0200, 0.0198, 0.0198, 0.0200, 0.0201, 0.0205, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:52:42,619 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:53:07,978 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:53:27,430 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.232e+02 3.321e+02 3.959e+02 5.175e+02 1.024e+03, threshold=7.919e+02, percent-clipped=5.0 2023-04-29 10:53:29,999 INFO [train.py:904] (5/8) Epoch 11, batch 6250, loss[loss=0.2104, simple_loss=0.3095, pruned_loss=0.05564, over 16594.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3017, pruned_loss=0.06839, over 3087948.69 frames. ], batch size: 68, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:36,373 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:54:38,955 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:54:45,115 INFO [train.py:904] (5/8) Epoch 11, batch 6300, loss[loss=0.1995, simple_loss=0.2836, pruned_loss=0.05766, over 16687.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3014, pruned_loss=0.06746, over 3108335.03 frames. ], batch size: 89, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:54,817 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:55:52,958 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:56:00,520 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 3.066e+02 3.848e+02 4.977e+02 9.936e+02, threshold=7.696e+02, percent-clipped=3.0 2023-04-29 10:56:03,044 INFO [train.py:904] (5/8) Epoch 11, batch 6350, loss[loss=0.2389, simple_loss=0.3133, pruned_loss=0.0823, over 16699.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3021, pruned_loss=0.06842, over 3112759.53 frames. ], batch size: 134, lr: 6.18e-03, grad_scale: 4.0 2023-04-29 10:56:10,572 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:56:25,833 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:56:59,572 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3111, 3.5910, 2.4165, 1.9815, 2.4359, 2.0153, 3.5098, 3.3273], device='cuda:5'), covar=tensor([0.3085, 0.0703, 0.2010, 0.2438, 0.2344, 0.2101, 0.0645, 0.0987], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0254, 0.0279, 0.0274, 0.0277, 0.0217, 0.0263, 0.0288], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 10:57:20,961 INFO [train.py:904] (5/8) Epoch 11, batch 6400, loss[loss=0.1905, simple_loss=0.275, pruned_loss=0.05304, over 16536.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3022, pruned_loss=0.06975, over 3105163.23 frames. ], batch size: 62, lr: 6.18e-03, grad_scale: 8.0 2023-04-29 10:58:35,869 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.893e+02 3.293e+02 4.168e+02 5.156e+02 1.369e+03, threshold=8.335e+02, percent-clipped=6.0 2023-04-29 10:58:35,891 INFO [train.py:904] (5/8) Epoch 11, batch 6450, loss[loss=0.2142, simple_loss=0.2967, pruned_loss=0.06581, over 16356.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3024, pruned_loss=0.06923, over 3106356.97 frames. ], batch size: 146, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 10:59:54,951 INFO [train.py:904] (5/8) Epoch 11, batch 6500, loss[loss=0.1921, simple_loss=0.2786, pruned_loss=0.05283, over 16880.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3001, pruned_loss=0.06815, over 3115565.47 frames. ], batch size: 96, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:00:14,837 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:01:12,926 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 3.175e+02 3.633e+02 4.590e+02 1.124e+03, threshold=7.266e+02, percent-clipped=3.0 2023-04-29 11:01:12,947 INFO [train.py:904] (5/8) Epoch 11, batch 6550, loss[loss=0.238, simple_loss=0.3444, pruned_loss=0.06574, over 16560.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.303, pruned_loss=0.06901, over 3113779.98 frames. ], batch size: 68, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:02:13,653 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:02:15,218 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 11:02:15,231 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 11:02:25,854 INFO [train.py:904] (5/8) Epoch 11, batch 6600, loss[loss=0.2766, simple_loss=0.3395, pruned_loss=0.1068, over 11656.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3061, pruned_loss=0.06999, over 3114479.17 frames. ], batch size: 247, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:03:27,068 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3848, 3.5778, 3.9738, 1.6193, 4.1834, 4.1977, 2.8494, 2.8873], device='cuda:5'), covar=tensor([0.0813, 0.0209, 0.0140, 0.1264, 0.0043, 0.0094, 0.0413, 0.0460], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0099, 0.0088, 0.0137, 0.0069, 0.0102, 0.0119, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 11:03:41,609 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.097e+02 3.251e+02 4.038e+02 4.757e+02 1.371e+03, threshold=8.077e+02, percent-clipped=5.0 2023-04-29 11:03:41,631 INFO [train.py:904] (5/8) Epoch 11, batch 6650, loss[loss=0.3152, simple_loss=0.3575, pruned_loss=0.1364, over 11220.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3062, pruned_loss=0.07071, over 3099498.29 frames. ], batch size: 247, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:04:00,006 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0347, 1.9187, 2.1080, 3.6139, 1.9170, 2.2963, 2.1214, 2.0641], device='cuda:5'), covar=tensor([0.1112, 0.3208, 0.2353, 0.0521, 0.3980, 0.2315, 0.3107, 0.3152], device='cuda:5'), in_proj_covar=tensor([0.0359, 0.0387, 0.0325, 0.0318, 0.0409, 0.0445, 0.0353, 0.0455], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:04:03,254 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:04:09,501 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1853, 5.4855, 5.1707, 5.1736, 4.8827, 4.7107, 4.8607, 5.5812], device='cuda:5'), covar=tensor([0.1073, 0.0774, 0.1026, 0.0746, 0.0894, 0.0862, 0.1055, 0.0843], device='cuda:5'), in_proj_covar=tensor([0.0545, 0.0670, 0.0558, 0.0463, 0.0423, 0.0437, 0.0559, 0.0515], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:04:56,957 INFO [train.py:904] (5/8) Epoch 11, batch 6700, loss[loss=0.2004, simple_loss=0.29, pruned_loss=0.05542, over 17026.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3046, pruned_loss=0.07042, over 3110173.41 frames. ], batch size: 55, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:05:15,642 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:05:35,912 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3022, 4.2914, 4.1742, 3.5035, 4.2039, 1.6856, 3.9958, 3.9367], device='cuda:5'), covar=tensor([0.0078, 0.0065, 0.0141, 0.0305, 0.0077, 0.2331, 0.0117, 0.0158], device='cuda:5'), in_proj_covar=tensor([0.0125, 0.0113, 0.0161, 0.0152, 0.0132, 0.0175, 0.0147, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:05:48,518 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6078, 2.7776, 2.4013, 4.1490, 3.1124, 4.0368, 1.5505, 2.9104], device='cuda:5'), covar=tensor([0.1308, 0.0654, 0.1202, 0.0133, 0.0230, 0.0359, 0.1458, 0.0762], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0159, 0.0183, 0.0144, 0.0201, 0.0209, 0.0181, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 11:06:08,121 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7835, 4.5316, 4.7975, 4.9638, 5.1344, 4.6091, 5.0738, 5.1030], device='cuda:5'), covar=tensor([0.1371, 0.1202, 0.1418, 0.0613, 0.0520, 0.0825, 0.0526, 0.0552], device='cuda:5'), in_proj_covar=tensor([0.0514, 0.0636, 0.0768, 0.0645, 0.0496, 0.0497, 0.0514, 0.0578], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:06:13,522 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 3.397e+02 4.413e+02 5.907e+02 2.679e+03, threshold=8.826e+02, percent-clipped=7.0 2023-04-29 11:06:13,537 INFO [train.py:904] (5/8) Epoch 11, batch 6750, loss[loss=0.1905, simple_loss=0.2832, pruned_loss=0.04895, over 16638.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3041, pruned_loss=0.07092, over 3113446.64 frames. ], batch size: 62, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:07:28,522 INFO [train.py:904] (5/8) Epoch 11, batch 6800, loss[loss=0.2145, simple_loss=0.2991, pruned_loss=0.06499, over 17121.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3049, pruned_loss=0.07138, over 3098352.06 frames. ], batch size: 49, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:07:49,153 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:05,598 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:45,536 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.125e+02 2.977e+02 3.381e+02 4.053e+02 7.016e+02, threshold=6.762e+02, percent-clipped=0.0 2023-04-29 11:08:45,557 INFO [train.py:904] (5/8) Epoch 11, batch 6850, loss[loss=0.2266, simple_loss=0.3249, pruned_loss=0.0642, over 16756.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3056, pruned_loss=0.07183, over 3095742.68 frames. ], batch size: 83, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:09:01,713 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:35,642 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:46,555 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108393.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:59,726 INFO [train.py:904] (5/8) Epoch 11, batch 6900, loss[loss=0.2515, simple_loss=0.3398, pruned_loss=0.08156, over 16830.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3078, pruned_loss=0.07072, over 3127522.31 frames. ], batch size: 83, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:10:37,133 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4329, 3.4630, 3.7497, 1.8206, 3.9136, 3.9800, 2.9277, 2.8725], device='cuda:5'), covar=tensor([0.0763, 0.0207, 0.0139, 0.1126, 0.0060, 0.0112, 0.0379, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0099, 0.0088, 0.0139, 0.0069, 0.0102, 0.0120, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 11:10:45,840 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:10:59,695 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:11:15,803 INFO [train.py:904] (5/8) Epoch 11, batch 6950, loss[loss=0.2608, simple_loss=0.3305, pruned_loss=0.09555, over 15348.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3092, pruned_loss=0.07221, over 3129256.22 frames. ], batch size: 191, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:11:17,889 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 2.933e+02 3.744e+02 4.621e+02 9.342e+02, threshold=7.489e+02, percent-clipped=9.0 2023-04-29 11:12:20,887 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:12:33,554 INFO [train.py:904] (5/8) Epoch 11, batch 7000, loss[loss=0.2009, simple_loss=0.3013, pruned_loss=0.05021, over 17139.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3093, pruned_loss=0.07125, over 3127030.16 frames. ], batch size: 47, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:13:52,279 INFO [train.py:904] (5/8) Epoch 11, batch 7050, loss[loss=0.2853, simple_loss=0.332, pruned_loss=0.1193, over 11171.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.31, pruned_loss=0.07144, over 3108324.48 frames. ], batch size: 247, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:13:53,485 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 3.044e+02 3.927e+02 4.822e+02 1.171e+03, threshold=7.854e+02, percent-clipped=4.0 2023-04-29 11:14:33,524 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:15:11,207 INFO [train.py:904] (5/8) Epoch 11, batch 7100, loss[loss=0.1974, simple_loss=0.2895, pruned_loss=0.0527, over 16735.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3087, pruned_loss=0.07165, over 3077571.31 frames. ], batch size: 83, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:07,291 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:16:27,146 INFO [train.py:904] (5/8) Epoch 11, batch 7150, loss[loss=0.27, simple_loss=0.3262, pruned_loss=0.1068, over 11339.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3077, pruned_loss=0.07251, over 3050938.40 frames. ], batch size: 247, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:28,932 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 3.054e+02 3.516e+02 4.486e+02 8.068e+02, threshold=7.031e+02, percent-clipped=2.0 2023-04-29 11:17:10,563 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:17:41,823 INFO [train.py:904] (5/8) Epoch 11, batch 7200, loss[loss=0.2098, simple_loss=0.2934, pruned_loss=0.06313, over 15392.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3048, pruned_loss=0.07067, over 3038578.84 frames. ], batch size: 191, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:18:05,938 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4301, 5.7640, 5.4515, 5.5474, 5.1816, 5.0010, 5.2816, 5.8716], device='cuda:5'), covar=tensor([0.0948, 0.0668, 0.1008, 0.0569, 0.0757, 0.0704, 0.0845, 0.0762], device='cuda:5'), in_proj_covar=tensor([0.0530, 0.0651, 0.0542, 0.0448, 0.0413, 0.0426, 0.0545, 0.0499], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:18:44,824 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9750, 3.4071, 3.4737, 1.8494, 2.9398, 2.3084, 3.4084, 3.4914], device='cuda:5'), covar=tensor([0.0261, 0.0635, 0.0564, 0.1952, 0.0785, 0.0931, 0.0693, 0.0924], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0146, 0.0160, 0.0146, 0.0139, 0.0127, 0.0139, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 11:19:02,077 INFO [train.py:904] (5/8) Epoch 11, batch 7250, loss[loss=0.2062, simple_loss=0.2863, pruned_loss=0.06309, over 16740.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3019, pruned_loss=0.06872, over 3051156.71 frames. ], batch size: 83, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:19:03,147 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.903e+02 3.497e+02 4.380e+02 7.660e+02, threshold=6.994e+02, percent-clipped=1.0 2023-04-29 11:19:55,787 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:20:00,936 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:20:16,092 INFO [train.py:904] (5/8) Epoch 11, batch 7300, loss[loss=0.2222, simple_loss=0.3102, pruned_loss=0.06712, over 16708.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3015, pruned_loss=0.06833, over 3065128.19 frames. ], batch size: 134, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:21:05,535 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0704, 2.9782, 3.1072, 1.6250, 3.2817, 3.3348, 2.5938, 2.5224], device='cuda:5'), covar=tensor([0.0823, 0.0219, 0.0165, 0.1129, 0.0064, 0.0130, 0.0462, 0.0442], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0097, 0.0086, 0.0135, 0.0067, 0.0099, 0.0117, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 11:21:17,726 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5363, 2.6068, 2.2393, 3.7891, 2.8486, 3.8625, 1.3817, 2.8843], device='cuda:5'), covar=tensor([0.1355, 0.0719, 0.1295, 0.0143, 0.0248, 0.0386, 0.1563, 0.0740], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0159, 0.0182, 0.0144, 0.0201, 0.0209, 0.0182, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 11:21:34,174 INFO [train.py:904] (5/8) Epoch 11, batch 7350, loss[loss=0.1929, simple_loss=0.2753, pruned_loss=0.05523, over 16513.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3013, pruned_loss=0.06878, over 3060990.57 frames. ], batch size: 68, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:21:34,664 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:21:35,292 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 3.104e+02 3.844e+02 4.749e+02 1.066e+03, threshold=7.688e+02, percent-clipped=6.0 2023-04-29 11:22:54,336 INFO [train.py:904] (5/8) Epoch 11, batch 7400, loss[loss=0.2123, simple_loss=0.302, pruned_loss=0.06127, over 16691.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3024, pruned_loss=0.06901, over 3071966.46 frames. ], batch size: 76, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:23:30,455 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4608, 4.5554, 4.7006, 4.5596, 4.6129, 5.1408, 4.6523, 4.4285], device='cuda:5'), covar=tensor([0.1232, 0.1759, 0.1966, 0.1886, 0.2495, 0.1032, 0.1520, 0.2478], device='cuda:5'), in_proj_covar=tensor([0.0346, 0.0480, 0.0525, 0.0416, 0.0553, 0.0550, 0.0414, 0.0568], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 11:23:42,503 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:24:11,401 INFO [train.py:904] (5/8) Epoch 11, batch 7450, loss[loss=0.2195, simple_loss=0.3094, pruned_loss=0.06479, over 15327.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3034, pruned_loss=0.07023, over 3060094.96 frames. ], batch size: 190, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:24:13,641 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.939e+02 3.048e+02 3.705e+02 5.111e+02 8.676e+02, threshold=7.410e+02, percent-clipped=2.0 2023-04-29 11:24:59,152 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:25:31,629 INFO [train.py:904] (5/8) Epoch 11, batch 7500, loss[loss=0.1809, simple_loss=0.2695, pruned_loss=0.04616, over 16522.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3032, pruned_loss=0.06957, over 3063116.09 frames. ], batch size: 35, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:16,355 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:26:28,532 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:26:51,096 INFO [train.py:904] (5/8) Epoch 11, batch 7550, loss[loss=0.2529, simple_loss=0.3116, pruned_loss=0.09709, over 11388.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3027, pruned_loss=0.07045, over 3033796.86 frames. ], batch size: 248, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:52,324 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 3.123e+02 3.909e+02 4.676e+02 1.371e+03, threshold=7.817e+02, percent-clipped=2.0 2023-04-29 11:27:05,305 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0395, 3.5194, 3.3960, 1.9547, 2.7945, 2.2474, 3.5453, 3.6446], device='cuda:5'), covar=tensor([0.0231, 0.0631, 0.0542, 0.1907, 0.0882, 0.0953, 0.0605, 0.0818], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0143, 0.0157, 0.0144, 0.0137, 0.0125, 0.0136, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 11:27:46,166 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:28:01,995 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:28:06,709 INFO [train.py:904] (5/8) Epoch 11, batch 7600, loss[loss=0.2685, simple_loss=0.3269, pruned_loss=0.1051, over 11477.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3024, pruned_loss=0.07082, over 3033933.91 frames. ], batch size: 247, lr: 6.15e-03, grad_scale: 8.0 2023-04-29 11:28:25,683 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 11:28:56,880 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:12,287 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:20,696 INFO [train.py:904] (5/8) Epoch 11, batch 7650, loss[loss=0.2296, simple_loss=0.31, pruned_loss=0.07455, over 16725.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3029, pruned_loss=0.07123, over 3041093.56 frames. ], batch size: 124, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:29:23,631 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 2.787e+02 3.621e+02 4.390e+02 9.045e+02, threshold=7.242e+02, percent-clipped=3.0 2023-04-29 11:30:05,670 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5203, 4.3350, 4.5693, 4.7485, 4.8854, 4.4188, 4.8426, 4.8565], device='cuda:5'), covar=tensor([0.1546, 0.1120, 0.1443, 0.0670, 0.0508, 0.0852, 0.0574, 0.0599], device='cuda:5'), in_proj_covar=tensor([0.0517, 0.0639, 0.0773, 0.0651, 0.0503, 0.0501, 0.0514, 0.0582], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:30:25,807 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:30:36,527 INFO [train.py:904] (5/8) Epoch 11, batch 7700, loss[loss=0.2172, simple_loss=0.297, pruned_loss=0.06868, over 16543.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3029, pruned_loss=0.07121, over 3069031.22 frames. ], batch size: 146, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:31:22,296 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1419, 3.9313, 3.9542, 2.4008, 3.6215, 3.9409, 3.7217, 2.2788], device='cuda:5'), covar=tensor([0.0479, 0.0033, 0.0036, 0.0350, 0.0064, 0.0089, 0.0061, 0.0349], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0068, 0.0069, 0.0127, 0.0078, 0.0090, 0.0078, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 11:31:25,258 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:31:53,779 INFO [train.py:904] (5/8) Epoch 11, batch 7750, loss[loss=0.2696, simple_loss=0.3245, pruned_loss=0.1074, over 11710.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3029, pruned_loss=0.07056, over 3085686.69 frames. ], batch size: 247, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:31:56,734 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.029e+02 3.061e+02 3.488e+02 4.422e+02 1.299e+03, threshold=6.977e+02, percent-clipped=6.0 2023-04-29 11:31:59,798 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:29,057 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 11:32:39,669 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:58,612 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:33:10,727 INFO [train.py:904] (5/8) Epoch 11, batch 7800, loss[loss=0.2671, simple_loss=0.3185, pruned_loss=0.1078, over 11483.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3033, pruned_loss=0.07036, over 3099973.91 frames. ], batch size: 248, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:33:35,386 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:33:36,861 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 11:34:25,579 INFO [train.py:904] (5/8) Epoch 11, batch 7850, loss[loss=0.2332, simple_loss=0.3186, pruned_loss=0.07386, over 16270.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3043, pruned_loss=0.07019, over 3106130.61 frames. ], batch size: 165, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:34:30,498 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.254e+02 2.996e+02 3.807e+02 4.830e+02 8.310e+02, threshold=7.614e+02, percent-clipped=7.0 2023-04-29 11:34:31,704 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:34:51,330 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7857, 5.1766, 5.3804, 5.1308, 5.1475, 5.7447, 5.2119, 4.9864], device='cuda:5'), covar=tensor([0.1006, 0.1682, 0.1683, 0.1704, 0.2187, 0.0855, 0.1400, 0.2410], device='cuda:5'), in_proj_covar=tensor([0.0353, 0.0488, 0.0531, 0.0420, 0.0558, 0.0556, 0.0421, 0.0574], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 11:35:07,103 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:35:15,353 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9303, 4.9035, 4.6980, 4.0837, 4.7711, 1.8896, 4.5155, 4.4952], device='cuda:5'), covar=tensor([0.0062, 0.0047, 0.0117, 0.0278, 0.0061, 0.2112, 0.0086, 0.0163], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0112, 0.0161, 0.0152, 0.0131, 0.0176, 0.0147, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:35:27,315 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:35:41,216 INFO [train.py:904] (5/8) Epoch 11, batch 7900, loss[loss=0.2239, simple_loss=0.2918, pruned_loss=0.07799, over 11715.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3034, pruned_loss=0.06978, over 3101775.93 frames. ], batch size: 247, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:36:52,270 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109447.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:37:00,089 INFO [train.py:904] (5/8) Epoch 11, batch 7950, loss[loss=0.209, simple_loss=0.2825, pruned_loss=0.06775, over 16606.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3042, pruned_loss=0.07024, over 3103810.95 frames. ], batch size: 57, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:37:04,713 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.170e+02 3.029e+02 3.447e+02 4.404e+02 6.785e+02, threshold=6.894e+02, percent-clipped=0.0 2023-04-29 11:37:25,705 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5379, 4.5482, 4.9591, 4.8980, 4.9066, 4.6090, 4.5835, 4.3624], device='cuda:5'), covar=tensor([0.0247, 0.0394, 0.0297, 0.0368, 0.0392, 0.0286, 0.0798, 0.0408], device='cuda:5'), in_proj_covar=tensor([0.0321, 0.0335, 0.0338, 0.0318, 0.0386, 0.0354, 0.0459, 0.0289], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 11:37:39,472 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5587, 4.7070, 4.4755, 4.2168, 3.8686, 4.6295, 4.4459, 4.1458], device='cuda:5'), covar=tensor([0.0719, 0.0512, 0.0351, 0.0349, 0.1371, 0.0514, 0.0378, 0.0778], device='cuda:5'), in_proj_covar=tensor([0.0237, 0.0306, 0.0279, 0.0257, 0.0298, 0.0297, 0.0193, 0.0325], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:38:04,155 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:38:14,167 INFO [train.py:904] (5/8) Epoch 11, batch 8000, loss[loss=0.2374, simple_loss=0.3241, pruned_loss=0.07539, over 16455.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3044, pruned_loss=0.07088, over 3090233.35 frames. ], batch size: 146, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:38:54,463 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0256, 4.7736, 5.0885, 5.2573, 5.4457, 4.7873, 5.3919, 5.3836], device='cuda:5'), covar=tensor([0.1462, 0.1241, 0.1399, 0.0608, 0.0464, 0.0750, 0.0468, 0.0566], device='cuda:5'), in_proj_covar=tensor([0.0513, 0.0638, 0.0768, 0.0648, 0.0502, 0.0495, 0.0511, 0.0581], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:39:24,927 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:39:27,477 INFO [train.py:904] (5/8) Epoch 11, batch 8050, loss[loss=0.2721, simple_loss=0.3304, pruned_loss=0.107, over 11696.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3041, pruned_loss=0.06995, over 3115921.81 frames. ], batch size: 247, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:39:31,008 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.160e+02 2.984e+02 3.790e+02 4.579e+02 1.063e+03, threshold=7.580e+02, percent-clipped=3.0 2023-04-29 11:40:06,491 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:40:40,908 INFO [train.py:904] (5/8) Epoch 11, batch 8100, loss[loss=0.2161, simple_loss=0.2844, pruned_loss=0.07392, over 11640.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3036, pruned_loss=0.06958, over 3105766.35 frames. ], batch size: 247, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:41:32,180 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5134, 2.2057, 1.5899, 1.9877, 2.4328, 2.2051, 2.4654, 2.6737], device='cuda:5'), covar=tensor([0.0130, 0.0285, 0.0423, 0.0329, 0.0182, 0.0288, 0.0167, 0.0185], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0201, 0.0198, 0.0198, 0.0199, 0.0201, 0.0204, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:41:39,287 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:41:54,602 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:41:56,618 INFO [train.py:904] (5/8) Epoch 11, batch 8150, loss[loss=0.2316, simple_loss=0.2934, pruned_loss=0.08488, over 11689.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3009, pruned_loss=0.06875, over 3101997.88 frames. ], batch size: 247, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:42:01,363 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.997e+02 3.110e+02 3.865e+02 4.805e+02 7.636e+02, threshold=7.730e+02, percent-clipped=1.0 2023-04-29 11:42:08,239 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:42:30,928 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:42:59,236 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:43:12,477 INFO [train.py:904] (5/8) Epoch 11, batch 8200, loss[loss=0.204, simple_loss=0.2935, pruned_loss=0.05723, over 16724.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2982, pruned_loss=0.06755, over 3118714.77 frames. ], batch size: 89, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:43:42,567 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:44:16,310 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:44:30,162 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-29 11:44:33,552 INFO [train.py:904] (5/8) Epoch 11, batch 8250, loss[loss=0.1817, simple_loss=0.2782, pruned_loss=0.04261, over 16903.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2974, pruned_loss=0.0662, over 3063678.12 frames. ], batch size: 96, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:44:38,012 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.186e+02 3.150e+02 3.966e+02 4.988e+02 1.179e+03, threshold=7.932e+02, percent-clipped=8.0 2023-04-29 11:45:05,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1584, 4.1467, 4.5165, 4.4862, 4.4741, 4.2509, 4.1589, 4.1080], device='cuda:5'), covar=tensor([0.0294, 0.0453, 0.0348, 0.0359, 0.0429, 0.0316, 0.0992, 0.0446], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0335, 0.0338, 0.0315, 0.0386, 0.0353, 0.0460, 0.0288], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 11:45:52,507 INFO [train.py:904] (5/8) Epoch 11, batch 8300, loss[loss=0.2024, simple_loss=0.2982, pruned_loss=0.05331, over 16393.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2945, pruned_loss=0.06302, over 3060431.40 frames. ], batch size: 146, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:46:39,296 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5359, 3.6554, 3.7341, 2.7177, 3.3982, 3.6943, 3.5886, 2.4376], device='cuda:5'), covar=tensor([0.0307, 0.0030, 0.0027, 0.0225, 0.0056, 0.0056, 0.0045, 0.0299], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0066, 0.0068, 0.0123, 0.0076, 0.0088, 0.0075, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 11:47:09,172 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109850.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:47:11,579 INFO [train.py:904] (5/8) Epoch 11, batch 8350, loss[loss=0.2109, simple_loss=0.3005, pruned_loss=0.06062, over 16653.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2939, pruned_loss=0.06096, over 3059140.14 frames. ], batch size: 134, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:47:16,940 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.495e+02 2.844e+02 3.376e+02 6.294e+02, threshold=5.687e+02, percent-clipped=0.0 2023-04-29 11:48:25,292 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109898.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:48:30,295 INFO [train.py:904] (5/8) Epoch 11, batch 8400, loss[loss=0.2028, simple_loss=0.2935, pruned_loss=0.05608, over 16428.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2904, pruned_loss=0.05862, over 3037288.31 frames. ], batch size: 146, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:48:38,767 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9955, 1.7937, 1.5635, 1.4551, 1.8701, 1.5824, 1.6752, 1.9378], device='cuda:5'), covar=tensor([0.0102, 0.0237, 0.0317, 0.0286, 0.0172, 0.0244, 0.0131, 0.0153], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0196, 0.0193, 0.0192, 0.0195, 0.0196, 0.0198, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:49:05,789 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-04-29 11:49:21,869 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:49:45,404 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:49:47,555 INFO [train.py:904] (5/8) Epoch 11, batch 8450, loss[loss=0.1847, simple_loss=0.2715, pruned_loss=0.04893, over 12535.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2891, pruned_loss=0.05682, over 3047227.23 frames. ], batch size: 248, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:49:52,354 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.195e+02 2.664e+02 3.370e+02 8.341e+02, threshold=5.327e+02, percent-clipped=3.0 2023-04-29 11:50:23,549 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:50:40,858 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 11:50:41,814 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7639, 4.7561, 4.5293, 3.9718, 4.5288, 1.8414, 4.3440, 4.4525], device='cuda:5'), covar=tensor([0.0080, 0.0073, 0.0175, 0.0313, 0.0091, 0.2283, 0.0133, 0.0171], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0112, 0.0161, 0.0150, 0.0131, 0.0178, 0.0147, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:51:00,423 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:07,759 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4336, 1.9932, 2.1203, 3.9005, 2.0547, 2.4480, 2.1107, 2.2185], device='cuda:5'), covar=tensor([0.0818, 0.3553, 0.2462, 0.0389, 0.3971, 0.2434, 0.3367, 0.3285], device='cuda:5'), in_proj_covar=tensor([0.0350, 0.0382, 0.0321, 0.0308, 0.0403, 0.0433, 0.0344, 0.0445], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 11:51:09,685 INFO [train.py:904] (5/8) Epoch 11, batch 8500, loss[loss=0.1811, simple_loss=0.2567, pruned_loss=0.05278, over 12185.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.285, pruned_loss=0.05404, over 3063195.91 frames. ], batch size: 247, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:51:31,238 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:34,024 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:38,744 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-29 11:51:41,431 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:52:17,188 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110044.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:52:31,130 INFO [train.py:904] (5/8) Epoch 11, batch 8550, loss[loss=0.1844, simple_loss=0.2787, pruned_loss=0.04508, over 16459.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2826, pruned_loss=0.05272, over 3060372.60 frames. ], batch size: 68, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:52:37,012 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.340e+02 2.855e+02 3.457e+02 5.505e+02, threshold=5.709e+02, percent-clipped=2.0 2023-04-29 11:53:19,636 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:07,122 INFO [train.py:904] (5/8) Epoch 11, batch 8600, loss[loss=0.1814, simple_loss=0.2799, pruned_loss=0.04143, over 15443.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2822, pruned_loss=0.05214, over 3030481.31 frames. ], batch size: 191, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:54:14,563 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:23,145 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:35,051 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:46,720 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7270, 4.8262, 5.0769, 4.8919, 4.9683, 5.4645, 4.9910, 4.6849], device='cuda:5'), covar=tensor([0.0818, 0.1783, 0.1683, 0.1690, 0.2221, 0.0775, 0.1270, 0.2198], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0464, 0.0507, 0.0399, 0.0528, 0.0534, 0.0401, 0.0544], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 11:55:43,454 INFO [train.py:904] (5/8) Epoch 11, batch 8650, loss[loss=0.1786, simple_loss=0.2757, pruned_loss=0.04074, over 16724.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2804, pruned_loss=0.05053, over 3032373.08 frames. ], batch size: 134, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:55:53,866 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.527e+02 3.196e+02 4.321e+02 7.577e+02, threshold=6.393e+02, percent-clipped=5.0 2023-04-29 11:56:28,778 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:56:40,283 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:56:57,406 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-29 11:57:25,611 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 11:57:30,821 INFO [train.py:904] (5/8) Epoch 11, batch 8700, loss[loss=0.1898, simple_loss=0.2787, pruned_loss=0.05048, over 12536.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2783, pruned_loss=0.04941, over 3052693.58 frames. ], batch size: 247, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:57:47,346 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 11:58:34,099 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:59:05,970 INFO [train.py:904] (5/8) Epoch 11, batch 8750, loss[loss=0.1948, simple_loss=0.289, pruned_loss=0.05032, over 15325.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2785, pruned_loss=0.04888, over 3062008.41 frames. ], batch size: 191, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:59:15,716 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.315e+02 2.719e+02 3.353e+02 7.427e+02, threshold=5.437e+02, percent-clipped=1.0 2023-04-29 11:59:46,820 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4998, 3.4710, 3.7709, 1.7073, 3.9706, 4.0251, 2.9547, 2.8707], device='cuda:5'), covar=tensor([0.0681, 0.0209, 0.0159, 0.1175, 0.0042, 0.0080, 0.0362, 0.0427], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0097, 0.0084, 0.0136, 0.0067, 0.0098, 0.0118, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 12:00:07,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5248, 4.4968, 4.2921, 3.8359, 4.3827, 1.6258, 4.1283, 4.2053], device='cuda:5'), covar=tensor([0.0057, 0.0051, 0.0133, 0.0238, 0.0073, 0.2283, 0.0100, 0.0146], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0110, 0.0158, 0.0146, 0.0129, 0.0175, 0.0145, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:00:22,511 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:00:22,648 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:00,289 INFO [train.py:904] (5/8) Epoch 11, batch 8800, loss[loss=0.1685, simple_loss=0.2646, pruned_loss=0.03625, over 16178.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2763, pruned_loss=0.04714, over 3090962.50 frames. ], batch size: 165, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 12:01:03,818 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:27,888 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:34,228 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:56,342 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 12:02:29,424 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110344.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:02:45,843 INFO [train.py:904] (5/8) Epoch 11, batch 8850, loss[loss=0.1787, simple_loss=0.2654, pruned_loss=0.04597, over 12247.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2781, pruned_loss=0.04644, over 3069969.44 frames. ], batch size: 248, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:02:52,480 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.350e+02 2.510e+02 2.922e+02 3.907e+02 6.547e+02, threshold=5.844e+02, percent-clipped=7.0 2023-04-29 12:03:09,148 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:11,319 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:29,845 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:42,473 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:42,882 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 12:04:23,033 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5676, 4.3519, 4.5914, 4.7396, 4.9116, 4.4055, 4.9180, 4.8880], device='cuda:5'), covar=tensor([0.1478, 0.1019, 0.1452, 0.0654, 0.0448, 0.0826, 0.0383, 0.0535], device='cuda:5'), in_proj_covar=tensor([0.0493, 0.0616, 0.0737, 0.0624, 0.0479, 0.0484, 0.0494, 0.0563], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:04:27,929 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110400.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:04:30,865 INFO [train.py:904] (5/8) Epoch 11, batch 8900, loss[loss=0.1583, simple_loss=0.2501, pruned_loss=0.03325, over 12596.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2783, pruned_loss=0.04588, over 3058755.25 frames. ], batch size: 248, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:04:31,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9752, 3.2066, 3.1584, 2.2480, 2.8295, 3.1287, 3.0984, 1.8167], device='cuda:5'), covar=tensor([0.0411, 0.0033, 0.0041, 0.0278, 0.0077, 0.0078, 0.0053, 0.0411], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0067, 0.0068, 0.0123, 0.0077, 0.0087, 0.0076, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 12:06:36,600 INFO [train.py:904] (5/8) Epoch 11, batch 8950, loss[loss=0.1741, simple_loss=0.2647, pruned_loss=0.04173, over 15357.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2781, pruned_loss=0.04653, over 3047997.66 frames. ], batch size: 191, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:06:45,509 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.155e+02 2.674e+02 3.353e+02 7.252e+02, threshold=5.349e+02, percent-clipped=1.0 2023-04-29 12:07:08,201 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:07:17,878 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:08:26,532 INFO [train.py:904] (5/8) Epoch 11, batch 9000, loss[loss=0.1715, simple_loss=0.2628, pruned_loss=0.0401, over 16841.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.275, pruned_loss=0.04519, over 3061679.55 frames. ], batch size: 124, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:08:26,533 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 12:08:36,935 INFO [train.py:938] (5/8) Epoch 11, validation: loss=0.1545, simple_loss=0.2586, pruned_loss=0.02523, over 944034.00 frames. 2023-04-29 12:08:36,936 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 12:09:18,942 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9413, 5.2562, 5.0317, 5.0491, 4.7353, 4.6765, 4.7388, 5.3071], device='cuda:5'), covar=tensor([0.0964, 0.0823, 0.0872, 0.0592, 0.0787, 0.0820, 0.0891, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0521, 0.0640, 0.0532, 0.0445, 0.0404, 0.0423, 0.0541, 0.0492], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:10:21,916 INFO [train.py:904] (5/8) Epoch 11, batch 9050, loss[loss=0.1789, simple_loss=0.2662, pruned_loss=0.04579, over 16716.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2759, pruned_loss=0.04542, over 3088098.20 frames. ], batch size: 134, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:10:28,905 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.375e+02 3.006e+02 3.857e+02 1.104e+03, threshold=6.012e+02, percent-clipped=5.0 2023-04-29 12:12:06,406 INFO [train.py:904] (5/8) Epoch 11, batch 9100, loss[loss=0.1901, simple_loss=0.2889, pruned_loss=0.04568, over 16152.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2753, pruned_loss=0.0456, over 3089573.06 frames. ], batch size: 165, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:12:25,219 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:13:36,714 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:06,388 INFO [train.py:904] (5/8) Epoch 11, batch 9150, loss[loss=0.1772, simple_loss=0.27, pruned_loss=0.04219, over 15476.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2752, pruned_loss=0.04508, over 3081752.58 frames. ], batch size: 191, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:14:15,981 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.395e+02 2.852e+02 3.560e+02 7.952e+02, threshold=5.704e+02, percent-clipped=1.0 2023-04-29 12:14:23,439 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:52,287 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:54,585 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:56,549 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:15:03,341 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 12:15:46,102 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110700.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:15:48,928 INFO [train.py:904] (5/8) Epoch 11, batch 9200, loss[loss=0.1762, simple_loss=0.2585, pruned_loss=0.04693, over 16609.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2711, pruned_loss=0.04437, over 3077511.05 frames. ], batch size: 57, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:16:24,213 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:16:52,950 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:18,212 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110748.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:24,350 INFO [train.py:904] (5/8) Epoch 11, batch 9250, loss[loss=0.1846, simple_loss=0.2805, pruned_loss=0.04435, over 15269.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2714, pruned_loss=0.04448, over 3091973.95 frames. ], batch size: 191, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:17:32,958 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.583e+02 3.099e+02 3.621e+02 8.759e+02, threshold=6.198e+02, percent-clipped=4.0 2023-04-29 12:17:53,801 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:53,955 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5935, 3.7702, 2.3613, 4.1521, 2.7792, 4.0408, 2.4221, 3.0857], device='cuda:5'), covar=tensor([0.0214, 0.0271, 0.1365, 0.0119, 0.0721, 0.0444, 0.1378, 0.0568], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0156, 0.0183, 0.0120, 0.0163, 0.0193, 0.0192, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 12:18:05,251 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:06,503 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 12:19:16,918 INFO [train.py:904] (5/8) Epoch 11, batch 9300, loss[loss=0.152, simple_loss=0.2405, pruned_loss=0.03177, over 16612.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2695, pruned_loss=0.04369, over 3087407.29 frames. ], batch size: 62, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:19:45,796 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:59,853 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:21:01,162 INFO [train.py:904] (5/8) Epoch 11, batch 9350, loss[loss=0.1905, simple_loss=0.2758, pruned_loss=0.05257, over 16699.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2697, pruned_loss=0.0437, over 3097899.95 frames. ], batch size: 89, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:21:10,067 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.415e+02 2.201e+02 2.572e+02 3.084e+02 6.751e+02, threshold=5.144e+02, percent-clipped=1.0 2023-04-29 12:21:36,178 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:22:41,702 INFO [train.py:904] (5/8) Epoch 11, batch 9400, loss[loss=0.182, simple_loss=0.2611, pruned_loss=0.05146, over 12633.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2698, pruned_loss=0.04373, over 3082793.01 frames. ], batch size: 250, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:23:11,040 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5373, 2.0412, 1.6807, 1.8094, 2.3360, 1.9907, 2.1981, 2.5047], device='cuda:5'), covar=tensor([0.0086, 0.0298, 0.0389, 0.0369, 0.0202, 0.0283, 0.0127, 0.0193], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0199, 0.0192, 0.0191, 0.0196, 0.0195, 0.0193, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:23:36,586 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:23:57,830 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:21,279 INFO [train.py:904] (5/8) Epoch 11, batch 9450, loss[loss=0.1991, simple_loss=0.2876, pruned_loss=0.05534, over 16283.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2712, pruned_loss=0.04389, over 3058662.81 frames. ], batch size: 165, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:24:21,979 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7146, 2.1257, 1.6926, 1.8719, 2.4392, 2.0769, 2.4366, 2.6541], device='cuda:5'), covar=tensor([0.0101, 0.0361, 0.0443, 0.0418, 0.0225, 0.0347, 0.0157, 0.0189], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0199, 0.0193, 0.0192, 0.0196, 0.0196, 0.0193, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:24:27,347 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.422e+02 3.077e+02 3.855e+02 7.743e+02, threshold=6.155e+02, percent-clipped=6.0 2023-04-29 12:24:33,128 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:34,546 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:50,476 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:06,710 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:33,578 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:00,942 INFO [train.py:904] (5/8) Epoch 11, batch 9500, loss[loss=0.1753, simple_loss=0.2806, pruned_loss=0.03496, over 16841.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2707, pruned_loss=0.04335, over 3069881.04 frames. ], batch size: 102, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:26:15,565 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:38,142 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:44,355 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:27:01,731 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2383, 3.7028, 3.7418, 2.5381, 3.3900, 3.6640, 3.5816, 2.5159], device='cuda:5'), covar=tensor([0.0415, 0.0023, 0.0029, 0.0292, 0.0062, 0.0064, 0.0042, 0.0295], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0065, 0.0066, 0.0122, 0.0075, 0.0085, 0.0073, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 12:27:03,468 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-29 12:27:29,868 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2444, 3.3957, 3.6590, 3.6193, 3.6493, 3.4279, 3.4954, 3.5000], device='cuda:5'), covar=tensor([0.0364, 0.0711, 0.0435, 0.0460, 0.0500, 0.0493, 0.0788, 0.0390], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0313, 0.0316, 0.0295, 0.0357, 0.0333, 0.0424, 0.0268], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-29 12:27:46,827 INFO [train.py:904] (5/8) Epoch 11, batch 9550, loss[loss=0.175, simple_loss=0.2636, pruned_loss=0.04326, over 12486.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2708, pruned_loss=0.0437, over 3078299.27 frames. ], batch size: 246, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:27:55,304 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 2.456e+02 2.878e+02 3.283e+02 5.727e+02, threshold=5.755e+02, percent-clipped=0.0 2023-04-29 12:28:52,130 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9682, 4.2244, 4.0676, 4.0759, 3.7557, 3.7811, 3.8812, 4.2216], device='cuda:5'), covar=tensor([0.1053, 0.0922, 0.0866, 0.0625, 0.0702, 0.1522, 0.0834, 0.0935], device='cuda:5'), in_proj_covar=tensor([0.0513, 0.0635, 0.0518, 0.0440, 0.0398, 0.0416, 0.0528, 0.0482], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:28:57,231 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 12:29:10,433 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 12:29:26,775 INFO [train.py:904] (5/8) Epoch 11, batch 9600, loss[loss=0.1839, simple_loss=0.2833, pruned_loss=0.04221, over 16754.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2725, pruned_loss=0.04477, over 3072668.47 frames. ], batch size: 83, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:29:43,641 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3530, 4.3725, 4.7404, 4.6953, 4.7241, 4.4438, 4.4208, 4.2263], device='cuda:5'), covar=tensor([0.0285, 0.0466, 0.0361, 0.0405, 0.0436, 0.0300, 0.0782, 0.0365], device='cuda:5'), in_proj_covar=tensor([0.0301, 0.0312, 0.0315, 0.0295, 0.0356, 0.0333, 0.0423, 0.0268], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-29 12:29:47,747 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 12:30:16,091 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0996, 4.1256, 4.4672, 4.4505, 4.4850, 4.1839, 4.1967, 4.0580], device='cuda:5'), covar=tensor([0.0298, 0.0507, 0.0443, 0.0466, 0.0421, 0.0354, 0.0807, 0.0400], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0313, 0.0317, 0.0297, 0.0357, 0.0334, 0.0425, 0.0269], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:5') 2023-04-29 12:31:14,975 INFO [train.py:904] (5/8) Epoch 11, batch 9650, loss[loss=0.1875, simple_loss=0.279, pruned_loss=0.04805, over 16868.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2737, pruned_loss=0.04484, over 3066287.42 frames. ], batch size: 116, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:31:24,140 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.367e+02 2.756e+02 3.328e+02 5.495e+02, threshold=5.512e+02, percent-clipped=0.0 2023-04-29 12:31:46,877 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4255, 1.9271, 1.5636, 1.6265, 2.2096, 1.8515, 2.0766, 2.3278], device='cuda:5'), covar=tensor([0.0114, 0.0291, 0.0404, 0.0379, 0.0189, 0.0325, 0.0121, 0.0207], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0201, 0.0196, 0.0195, 0.0199, 0.0198, 0.0196, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:32:15,682 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9542, 4.3558, 3.3564, 2.3785, 2.8739, 2.5728, 4.5297, 3.8884], device='cuda:5'), covar=tensor([0.2391, 0.0481, 0.1324, 0.2085, 0.2202, 0.1687, 0.0329, 0.0788], device='cuda:5'), in_proj_covar=tensor([0.0292, 0.0248, 0.0273, 0.0268, 0.0254, 0.0215, 0.0256, 0.0278], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:32:16,127 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-29 12:32:34,072 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8383, 1.8047, 2.1076, 2.8711, 2.5300, 3.0075, 2.0077, 3.0662], device='cuda:5'), covar=tensor([0.0144, 0.0377, 0.0264, 0.0175, 0.0225, 0.0127, 0.0349, 0.0094], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0167, 0.0148, 0.0152, 0.0161, 0.0118, 0.0166, 0.0108], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 12:33:03,909 INFO [train.py:904] (5/8) Epoch 11, batch 9700, loss[loss=0.1796, simple_loss=0.2737, pruned_loss=0.04276, over 16630.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2732, pruned_loss=0.04487, over 3066731.04 frames. ], batch size: 134, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:33:49,086 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:33:52,463 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:33:59,166 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8150, 3.7094, 3.8977, 3.9726, 4.0687, 3.6541, 4.0498, 4.1010], device='cuda:5'), covar=tensor([0.1228, 0.1053, 0.1158, 0.0628, 0.0502, 0.1550, 0.0612, 0.0588], device='cuda:5'), in_proj_covar=tensor([0.0487, 0.0610, 0.0729, 0.0619, 0.0471, 0.0479, 0.0493, 0.0556], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:34:46,302 INFO [train.py:904] (5/8) Epoch 11, batch 9750, loss[loss=0.199, simple_loss=0.2899, pruned_loss=0.054, over 16311.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2725, pruned_loss=0.04515, over 3070883.08 frames. ], batch size: 165, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:34:53,730 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.278e+02 2.798e+02 3.645e+02 8.933e+02, threshold=5.595e+02, percent-clipped=9.0 2023-04-29 12:35:01,296 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4003, 1.9810, 1.7596, 1.7557, 2.2104, 1.8914, 2.0287, 2.3392], device='cuda:5'), covar=tensor([0.0104, 0.0308, 0.0360, 0.0358, 0.0186, 0.0281, 0.0164, 0.0201], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0198, 0.0192, 0.0192, 0.0195, 0.0195, 0.0193, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:35:17,296 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:35:34,590 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2528, 3.2125, 3.3494, 1.6556, 3.5570, 3.5732, 2.8573, 2.7284], device='cuda:5'), covar=tensor([0.0752, 0.0206, 0.0188, 0.1183, 0.0058, 0.0131, 0.0381, 0.0417], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0097, 0.0083, 0.0136, 0.0065, 0.0098, 0.0117, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 12:35:58,520 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:25,545 INFO [train.py:904] (5/8) Epoch 11, batch 9800, loss[loss=0.165, simple_loss=0.2726, pruned_loss=0.02867, over 16813.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2716, pruned_loss=0.04397, over 3069709.90 frames. ], batch size: 83, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:36:49,127 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:51,181 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:37:04,509 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6604, 4.9807, 4.7440, 4.8086, 4.5463, 4.4488, 4.4603, 5.0151], device='cuda:5'), covar=tensor([0.0967, 0.0731, 0.0835, 0.0595, 0.0614, 0.0956, 0.0906, 0.0836], device='cuda:5'), in_proj_covar=tensor([0.0507, 0.0627, 0.0512, 0.0435, 0.0395, 0.0411, 0.0523, 0.0478], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:38:11,446 INFO [train.py:904] (5/8) Epoch 11, batch 9850, loss[loss=0.1784, simple_loss=0.2738, pruned_loss=0.04151, over 12591.00 frames. ], tot_loss[loss=0.18, simple_loss=0.273, pruned_loss=0.04346, over 3077500.03 frames. ], batch size: 248, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:38:20,191 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.271e+02 2.860e+02 3.428e+02 8.615e+02, threshold=5.720e+02, percent-clipped=1.0 2023-04-29 12:39:42,377 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 12:40:02,803 INFO [train.py:904] (5/8) Epoch 11, batch 9900, loss[loss=0.1711, simple_loss=0.2737, pruned_loss=0.03421, over 16776.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2733, pruned_loss=0.04323, over 3084898.42 frames. ], batch size: 76, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:40:35,518 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6254, 3.1167, 3.2146, 1.7973, 2.7718, 2.2058, 3.1662, 3.2898], device='cuda:5'), covar=tensor([0.0316, 0.0734, 0.0508, 0.1874, 0.0758, 0.0923, 0.0666, 0.0818], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0133, 0.0152, 0.0139, 0.0133, 0.0122, 0.0132, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 12:41:34,337 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:42:01,115 INFO [train.py:904] (5/8) Epoch 11, batch 9950, loss[loss=0.1628, simple_loss=0.2591, pruned_loss=0.03326, over 17164.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2755, pruned_loss=0.04341, over 3099300.13 frames. ], batch size: 44, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:42:11,466 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.371e+02 2.780e+02 3.480e+02 6.168e+02, threshold=5.560e+02, percent-clipped=1.0 2023-04-29 12:42:16,970 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 12:42:41,133 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:44:02,034 INFO [train.py:904] (5/8) Epoch 11, batch 10000, loss[loss=0.191, simple_loss=0.2969, pruned_loss=0.04257, over 16372.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2741, pruned_loss=0.04297, over 3101847.70 frames. ], batch size: 146, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:44:45,115 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:44:55,760 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:45:39,804 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 12:45:42,137 INFO [train.py:904] (5/8) Epoch 11, batch 10050, loss[loss=0.1903, simple_loss=0.2839, pruned_loss=0.04837, over 15277.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2746, pruned_loss=0.04341, over 3092270.65 frames. ], batch size: 191, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:45:50,239 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.160e+02 2.693e+02 3.599e+02 6.171e+02, threshold=5.385e+02, percent-clipped=1.0 2023-04-29 12:46:22,432 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:46:38,807 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:47:14,837 INFO [train.py:904] (5/8) Epoch 11, batch 10100, loss[loss=0.1685, simple_loss=0.2657, pruned_loss=0.03563, over 16794.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2753, pruned_loss=0.04395, over 3101234.63 frames. ], batch size: 83, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:47:35,964 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:48:27,778 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 12:48:58,025 INFO [train.py:904] (5/8) Epoch 12, batch 0, loss[loss=0.1909, simple_loss=0.274, pruned_loss=0.05392, over 17227.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.274, pruned_loss=0.05392, over 17227.00 frames. ], batch size: 45, lr: 5.82e-03, grad_scale: 8.0 2023-04-29 12:48:58,025 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 12:49:05,315 INFO [train.py:938] (5/8) Epoch 12, validation: loss=0.1543, simple_loss=0.2577, pruned_loss=0.0254, over 944034.00 frames. 2023-04-29 12:49:05,316 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 12:49:12,535 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.560e+02 3.107e+02 3.982e+02 7.820e+02, threshold=6.214e+02, percent-clipped=3.0 2023-04-29 12:49:21,579 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:49:22,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8876, 5.2142, 5.3626, 5.2169, 5.1215, 5.7493, 5.2758, 5.0056], device='cuda:5'), covar=tensor([0.1108, 0.1830, 0.2044, 0.1833, 0.2396, 0.1004, 0.1648, 0.2405], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0454, 0.0497, 0.0394, 0.0519, 0.0532, 0.0400, 0.0527], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-29 12:50:01,814 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2700, 3.1890, 3.5264, 2.3776, 3.2626, 3.5433, 3.2548, 1.7639], device='cuda:5'), covar=tensor([0.0450, 0.0177, 0.0053, 0.0340, 0.0110, 0.0101, 0.0119, 0.0529], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0066, 0.0067, 0.0124, 0.0076, 0.0085, 0.0075, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 12:50:15,974 INFO [train.py:904] (5/8) Epoch 12, batch 50, loss[loss=0.1966, simple_loss=0.2899, pruned_loss=0.05164, over 17021.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2863, pruned_loss=0.06353, over 751876.61 frames. ], batch size: 55, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:51:25,700 INFO [train.py:904] (5/8) Epoch 12, batch 100, loss[loss=0.1882, simple_loss=0.2707, pruned_loss=0.05285, over 17197.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.28, pruned_loss=0.0587, over 1326821.55 frames. ], batch size: 44, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:51:34,347 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.444e+02 2.855e+02 3.643e+02 7.519e+02, threshold=5.710e+02, percent-clipped=2.0 2023-04-29 12:52:18,945 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111791.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 12:52:31,946 INFO [train.py:904] (5/8) Epoch 12, batch 150, loss[loss=0.1921, simple_loss=0.2919, pruned_loss=0.04618, over 17127.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.278, pruned_loss=0.05714, over 1772558.39 frames. ], batch size: 48, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:52:56,593 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-29 12:53:01,493 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7571, 3.4666, 2.7414, 5.1265, 4.2392, 4.5791, 1.7421, 3.3993], device='cuda:5'), covar=tensor([0.1402, 0.0575, 0.1138, 0.0171, 0.0304, 0.0405, 0.1469, 0.0736], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0158, 0.0179, 0.0142, 0.0186, 0.0207, 0.0181, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 12:53:03,094 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:53:40,724 INFO [train.py:904] (5/8) Epoch 12, batch 200, loss[loss=0.223, simple_loss=0.3019, pruned_loss=0.07205, over 17077.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2771, pruned_loss=0.05615, over 2117111.07 frames. ], batch size: 53, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:41,293 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 12:53:50,236 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.566e+02 3.072e+02 3.744e+02 9.632e+02, threshold=6.144e+02, percent-clipped=5.0 2023-04-29 12:54:16,503 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2390, 3.5098, 3.4558, 2.1853, 2.9077, 2.4409, 3.5978, 3.7232], device='cuda:5'), covar=tensor([0.0265, 0.0654, 0.0622, 0.1595, 0.0767, 0.0894, 0.0543, 0.0781], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0138, 0.0157, 0.0143, 0.0135, 0.0125, 0.0135, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 12:54:21,391 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:54:49,928 INFO [train.py:904] (5/8) Epoch 12, batch 250, loss[loss=0.204, simple_loss=0.2747, pruned_loss=0.06663, over 16874.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2751, pruned_loss=0.0558, over 2382874.20 frames. ], batch size: 96, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:55:27,223 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:55:32,723 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1588, 2.0794, 2.4850, 3.1465, 2.8703, 3.6264, 2.4719, 3.4145], device='cuda:5'), covar=tensor([0.0140, 0.0336, 0.0239, 0.0192, 0.0213, 0.0117, 0.0281, 0.0111], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0171, 0.0153, 0.0159, 0.0168, 0.0123, 0.0170, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 12:55:57,765 INFO [train.py:904] (5/8) Epoch 12, batch 300, loss[loss=0.1763, simple_loss=0.2626, pruned_loss=0.04503, over 17043.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2728, pruned_loss=0.05448, over 2600403.39 frames. ], batch size: 55, lr: 5.82e-03, grad_scale: 1.0 2023-04-29 12:56:09,476 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.376e+02 2.773e+02 3.192e+02 6.381e+02, threshold=5.545e+02, percent-clipped=1.0 2023-04-29 12:56:55,269 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1991, 3.5863, 3.9597, 2.2175, 3.0003, 2.6564, 3.5208, 3.8139], device='cuda:5'), covar=tensor([0.0361, 0.0858, 0.0441, 0.1657, 0.0816, 0.0798, 0.0813, 0.1012], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0138, 0.0156, 0.0142, 0.0135, 0.0124, 0.0134, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 12:57:10,674 INFO [train.py:904] (5/8) Epoch 12, batch 350, loss[loss=0.2004, simple_loss=0.2715, pruned_loss=0.06461, over 12037.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2704, pruned_loss=0.0542, over 2751531.65 frames. ], batch size: 246, lr: 5.81e-03, grad_scale: 1.0 2023-04-29 12:57:17,869 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 12:57:20,129 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4586, 3.7762, 4.2208, 2.2198, 3.2788, 2.6652, 3.9125, 3.9983], device='cuda:5'), covar=tensor([0.0290, 0.0745, 0.0389, 0.1631, 0.0671, 0.0827, 0.0654, 0.0932], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0139, 0.0156, 0.0142, 0.0134, 0.0123, 0.0134, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 12:58:17,745 INFO [train.py:904] (5/8) Epoch 12, batch 400, loss[loss=0.1957, simple_loss=0.2836, pruned_loss=0.05387, over 16671.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2698, pruned_loss=0.05414, over 2866028.07 frames. ], batch size: 57, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:58:27,671 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.508e+02 2.981e+02 3.635e+02 6.653e+02, threshold=5.962e+02, percent-clipped=1.0 2023-04-29 12:58:36,943 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:58:53,616 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-29 12:59:26,055 INFO [train.py:904] (5/8) Epoch 12, batch 450, loss[loss=0.1558, simple_loss=0.2401, pruned_loss=0.03577, over 16999.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2679, pruned_loss=0.05244, over 2970796.12 frames. ], batch size: 41, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:59:47,039 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0279, 5.5805, 5.7175, 5.4443, 5.4079, 6.0903, 5.5720, 5.3431], device='cuda:5'), covar=tensor([0.0794, 0.1750, 0.2080, 0.1963, 0.3249, 0.1069, 0.1492, 0.2242], device='cuda:5'), in_proj_covar=tensor([0.0351, 0.0501, 0.0550, 0.0435, 0.0581, 0.0579, 0.0436, 0.0583], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 12:59:55,938 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:00,717 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:27,080 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:00:33,819 INFO [train.py:904] (5/8) Epoch 12, batch 500, loss[loss=0.1702, simple_loss=0.246, pruned_loss=0.04715, over 16844.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2666, pruned_loss=0.05143, over 3055955.51 frames. ], batch size: 102, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:00:45,215 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.270e+02 2.759e+02 3.532e+02 6.724e+02, threshold=5.519e+02, percent-clipped=2.0 2023-04-29 13:01:02,760 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:01:19,719 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 13:01:30,061 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 13:01:32,814 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8775, 4.9093, 5.4448, 5.3845, 5.4152, 5.0716, 4.9977, 4.8038], device='cuda:5'), covar=tensor([0.0324, 0.0539, 0.0403, 0.0457, 0.0420, 0.0346, 0.0905, 0.0427], device='cuda:5'), in_proj_covar=tensor([0.0333, 0.0346, 0.0348, 0.0328, 0.0389, 0.0367, 0.0468, 0.0298], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 13:01:44,743 INFO [train.py:904] (5/8) Epoch 12, batch 550, loss[loss=0.1901, simple_loss=0.2886, pruned_loss=0.04578, over 17257.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2648, pruned_loss=0.05015, over 3115854.71 frames. ], batch size: 52, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:02:55,025 INFO [train.py:904] (5/8) Epoch 12, batch 600, loss[loss=0.1858, simple_loss=0.2558, pruned_loss=0.05786, over 16769.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2649, pruned_loss=0.05101, over 3171720.15 frames. ], batch size: 102, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:03:06,854 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.365e+02 2.773e+02 3.420e+02 1.272e+03, threshold=5.547e+02, percent-clipped=1.0 2023-04-29 13:04:04,841 INFO [train.py:904] (5/8) Epoch 12, batch 650, loss[loss=0.163, simple_loss=0.2578, pruned_loss=0.0341, over 17067.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2638, pruned_loss=0.05028, over 3210872.03 frames. ], batch size: 50, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:04:22,757 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 13:05:14,237 INFO [train.py:904] (5/8) Epoch 12, batch 700, loss[loss=0.1709, simple_loss=0.2649, pruned_loss=0.03846, over 17197.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2621, pruned_loss=0.04919, over 3241249.23 frames. ], batch size: 45, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:05:18,751 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6932, 3.6890, 4.0349, 2.0473, 4.1855, 4.1795, 3.2330, 2.9738], device='cuda:5'), covar=tensor([0.0750, 0.0195, 0.0152, 0.1091, 0.0060, 0.0140, 0.0337, 0.0445], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0099, 0.0087, 0.0139, 0.0069, 0.0104, 0.0120, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 13:05:26,019 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.509e+02 2.916e+02 3.539e+02 5.225e+02, threshold=5.832e+02, percent-clipped=0.0 2023-04-29 13:06:07,759 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:06:09,248 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 13:06:18,599 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7076, 3.8719, 3.0014, 2.3024, 2.5904, 2.2827, 3.9225, 3.5568], device='cuda:5'), covar=tensor([0.2287, 0.0582, 0.1360, 0.2160, 0.2186, 0.1825, 0.0470, 0.1008], device='cuda:5'), in_proj_covar=tensor([0.0304, 0.0259, 0.0285, 0.0278, 0.0275, 0.0224, 0.0266, 0.0297], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:06:24,428 INFO [train.py:904] (5/8) Epoch 12, batch 750, loss[loss=0.1978, simple_loss=0.2677, pruned_loss=0.06394, over 16806.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2623, pruned_loss=0.04974, over 3259025.25 frames. ], batch size: 102, lr: 5.80e-03, grad_scale: 2.0 2023-04-29 13:06:52,551 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:07:29,061 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:07:32,747 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:07:34,636 INFO [train.py:904] (5/8) Epoch 12, batch 800, loss[loss=0.1845, simple_loss=0.2748, pruned_loss=0.04706, over 17025.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2628, pruned_loss=0.05028, over 3263661.59 frames. ], batch size: 55, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:07:45,050 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.488e+02 3.052e+02 3.720e+02 1.064e+03, threshold=6.105e+02, percent-clipped=2.0 2023-04-29 13:08:33,909 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 13:08:42,878 INFO [train.py:904] (5/8) Epoch 12, batch 850, loss[loss=0.1752, simple_loss=0.2494, pruned_loss=0.05051, over 16271.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2626, pruned_loss=0.05008, over 3273020.57 frames. ], batch size: 165, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:09:51,999 INFO [train.py:904] (5/8) Epoch 12, batch 900, loss[loss=0.167, simple_loss=0.2472, pruned_loss=0.04342, over 16801.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2616, pruned_loss=0.04991, over 3273729.34 frames. ], batch size: 102, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:10:02,362 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.529e+02 2.271e+02 2.835e+02 3.426e+02 5.348e+02, threshold=5.671e+02, percent-clipped=0.0 2023-04-29 13:10:05,757 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:26,107 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:35,777 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7423, 4.1660, 3.0915, 2.2794, 2.7201, 2.5406, 4.3626, 3.6364], device='cuda:5'), covar=tensor([0.2614, 0.0586, 0.1522, 0.2213, 0.2417, 0.1653, 0.0382, 0.1113], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0259, 0.0284, 0.0277, 0.0275, 0.0223, 0.0266, 0.0298], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:10:59,476 INFO [train.py:904] (5/8) Epoch 12, batch 950, loss[loss=0.1579, simple_loss=0.2427, pruned_loss=0.03657, over 16349.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2622, pruned_loss=0.04996, over 3282111.90 frames. ], batch size: 36, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:11:21,373 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:28,373 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:50,174 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 13:12:07,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8829, 4.0870, 2.3328, 4.6185, 2.9115, 4.5342, 2.5310, 3.2817], device='cuda:5'), covar=tensor([0.0255, 0.0329, 0.1614, 0.0179, 0.0815, 0.0474, 0.1512, 0.0675], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0166, 0.0189, 0.0134, 0.0167, 0.0208, 0.0198, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 13:12:07,853 INFO [train.py:904] (5/8) Epoch 12, batch 1000, loss[loss=0.1787, simple_loss=0.2673, pruned_loss=0.04499, over 17111.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2606, pruned_loss=0.04897, over 3293295.05 frames. ], batch size: 47, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:12:18,367 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.238e+02 2.624e+02 3.056e+02 7.407e+02, threshold=5.248e+02, percent-clipped=1.0 2023-04-29 13:12:43,287 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:12:49,792 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2029, 4.6190, 3.3891, 2.5075, 3.0452, 2.6269, 4.8752, 3.9812], device='cuda:5'), covar=tensor([0.2114, 0.0476, 0.1357, 0.2116, 0.2417, 0.1665, 0.0295, 0.0938], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0258, 0.0283, 0.0276, 0.0275, 0.0223, 0.0266, 0.0298], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:13:15,635 INFO [train.py:904] (5/8) Epoch 12, batch 1050, loss[loss=0.1508, simple_loss=0.234, pruned_loss=0.03374, over 16355.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2606, pruned_loss=0.04893, over 3304946.79 frames. ], batch size: 165, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:13:32,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2495, 3.2495, 3.5564, 2.6466, 3.2509, 3.5866, 3.3063, 2.0670], device='cuda:5'), covar=tensor([0.0407, 0.0129, 0.0042, 0.0255, 0.0081, 0.0073, 0.0073, 0.0367], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0070, 0.0069, 0.0125, 0.0079, 0.0089, 0.0078, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 13:13:35,500 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 13:13:36,698 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3213, 4.2595, 4.5839, 2.2058, 4.8263, 4.8449, 3.4621, 3.9460], device='cuda:5'), covar=tensor([0.0560, 0.0166, 0.0196, 0.1056, 0.0039, 0.0073, 0.0331, 0.0244], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0141, 0.0070, 0.0106, 0.0121, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 13:13:42,916 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:14,522 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:22,715 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3592, 3.3522, 3.7227, 2.8553, 3.4037, 3.7238, 3.4197, 2.1126], device='cuda:5'), covar=tensor([0.0391, 0.0131, 0.0037, 0.0234, 0.0076, 0.0079, 0.0070, 0.0376], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0070, 0.0069, 0.0125, 0.0079, 0.0089, 0.0078, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 13:14:23,338 INFO [train.py:904] (5/8) Epoch 12, batch 1100, loss[loss=0.2142, simple_loss=0.2778, pruned_loss=0.07527, over 16831.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2601, pruned_loss=0.04821, over 3308266.03 frames. ], batch size: 116, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:14:34,071 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.374e+02 2.893e+02 3.790e+02 9.509e+02, threshold=5.785e+02, percent-clipped=6.0 2023-04-29 13:14:48,741 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:15:33,623 INFO [train.py:904] (5/8) Epoch 12, batch 1150, loss[loss=0.1729, simple_loss=0.2586, pruned_loss=0.04365, over 17182.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2602, pruned_loss=0.04809, over 3317288.11 frames. ], batch size: 46, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:15:34,084 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6630, 3.7734, 4.1045, 3.0982, 3.6737, 4.1180, 3.8152, 2.4479], device='cuda:5'), covar=tensor([0.0374, 0.0208, 0.0035, 0.0248, 0.0066, 0.0069, 0.0062, 0.0352], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0071, 0.0070, 0.0127, 0.0080, 0.0089, 0.0078, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 13:15:51,126 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 13:15:57,882 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7463, 3.9163, 2.2144, 4.4660, 2.8493, 4.3741, 2.2992, 3.0744], device='cuda:5'), covar=tensor([0.0270, 0.0338, 0.1689, 0.0190, 0.0802, 0.0479, 0.1637, 0.0720], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0168, 0.0191, 0.0136, 0.0168, 0.0210, 0.0200, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 13:16:42,802 INFO [train.py:904] (5/8) Epoch 12, batch 1200, loss[loss=0.1864, simple_loss=0.2675, pruned_loss=0.05263, over 16736.00 frames. ], tot_loss[loss=0.178, simple_loss=0.26, pruned_loss=0.04798, over 3311912.00 frames. ], batch size: 134, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:16:52,678 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.211e+02 2.743e+02 3.250e+02 7.821e+02, threshold=5.486e+02, percent-clipped=2.0 2023-04-29 13:17:49,332 INFO [train.py:904] (5/8) Epoch 12, batch 1250, loss[loss=0.1696, simple_loss=0.2555, pruned_loss=0.04186, over 16659.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2599, pruned_loss=0.0478, over 3317895.31 frames. ], batch size: 57, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:18:12,364 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:16,581 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112921.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:32,355 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:18:41,658 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 13:18:57,920 INFO [train.py:904] (5/8) Epoch 12, batch 1300, loss[loss=0.1736, simple_loss=0.2505, pruned_loss=0.04836, over 12290.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.26, pruned_loss=0.04811, over 3323179.28 frames. ], batch size: 246, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:19:09,591 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.880e+02 2.412e+02 2.831e+02 3.329e+02 6.613e+02, threshold=5.661e+02, percent-clipped=2.0 2023-04-29 13:19:27,115 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:19:41,067 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:20:08,553 INFO [train.py:904] (5/8) Epoch 12, batch 1350, loss[loss=0.1675, simple_loss=0.2461, pruned_loss=0.0445, over 16496.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2596, pruned_loss=0.04809, over 3313627.52 frames. ], batch size: 146, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:21:08,004 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:21:16,370 INFO [train.py:904] (5/8) Epoch 12, batch 1400, loss[loss=0.2047, simple_loss=0.2606, pruned_loss=0.07435, over 16726.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2606, pruned_loss=0.04897, over 3317405.24 frames. ], batch size: 124, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:21:26,437 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.440e+02 3.018e+02 3.818e+02 8.239e+02, threshold=6.035e+02, percent-clipped=8.0 2023-04-29 13:22:04,104 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2783, 5.0542, 5.2342, 5.4825, 5.6532, 4.9560, 5.5922, 5.5930], device='cuda:5'), covar=tensor([0.1314, 0.0965, 0.1450, 0.0573, 0.0429, 0.0625, 0.0452, 0.0448], device='cuda:5'), in_proj_covar=tensor([0.0561, 0.0697, 0.0848, 0.0710, 0.0538, 0.0543, 0.0557, 0.0641], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:22:13,293 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:22:24,192 INFO [train.py:904] (5/8) Epoch 12, batch 1450, loss[loss=0.194, simple_loss=0.2833, pruned_loss=0.05235, over 17066.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2586, pruned_loss=0.04829, over 3318086.46 frames. ], batch size: 53, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:23:35,103 INFO [train.py:904] (5/8) Epoch 12, batch 1500, loss[loss=0.2021, simple_loss=0.2705, pruned_loss=0.06683, over 16913.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2582, pruned_loss=0.04744, over 3326526.97 frames. ], batch size: 90, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:23:45,776 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.499e+02 2.892e+02 3.478e+02 9.992e+02, threshold=5.783e+02, percent-clipped=3.0 2023-04-29 13:23:50,601 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 13:23:51,293 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2972, 5.8163, 5.9648, 5.6333, 5.7200, 6.2849, 5.8981, 5.6238], device='cuda:5'), covar=tensor([0.0798, 0.1684, 0.1681, 0.2165, 0.2778, 0.0907, 0.1233, 0.2244], device='cuda:5'), in_proj_covar=tensor([0.0357, 0.0509, 0.0558, 0.0441, 0.0587, 0.0582, 0.0438, 0.0593], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 13:23:52,635 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7073, 2.5107, 2.2802, 3.4436, 2.7656, 3.6558, 1.5229, 2.7783], device='cuda:5'), covar=tensor([0.1382, 0.0655, 0.1189, 0.0184, 0.0181, 0.0504, 0.1475, 0.0769], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0160, 0.0181, 0.0150, 0.0195, 0.0213, 0.0182, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 13:24:43,446 INFO [train.py:904] (5/8) Epoch 12, batch 1550, loss[loss=0.203, simple_loss=0.2834, pruned_loss=0.06136, over 15461.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2601, pruned_loss=0.04844, over 3324210.34 frames. ], batch size: 190, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:25:06,613 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:27,345 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:29,442 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:36,512 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 13:25:54,134 INFO [train.py:904] (5/8) Epoch 12, batch 1600, loss[loss=0.2036, simple_loss=0.3065, pruned_loss=0.05034, over 16725.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2617, pruned_loss=0.04825, over 3325334.57 frames. ], batch size: 57, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:26:04,710 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.814e+02 2.404e+02 2.815e+02 3.461e+02 5.184e+02, threshold=5.631e+02, percent-clipped=0.0 2023-04-29 13:26:12,638 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:23,439 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:26:28,588 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:33,989 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:53,591 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:27:02,735 INFO [train.py:904] (5/8) Epoch 12, batch 1650, loss[loss=0.1968, simple_loss=0.2685, pruned_loss=0.06256, over 16451.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2621, pruned_loss=0.04874, over 3324338.49 frames. ], batch size: 146, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:27:29,756 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:28:11,048 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0040, 4.5503, 3.4222, 2.3467, 2.9164, 2.6134, 4.8798, 3.9646], device='cuda:5'), covar=tensor([0.2330, 0.0554, 0.1376, 0.2313, 0.2516, 0.1760, 0.0299, 0.0970], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0257, 0.0281, 0.0277, 0.0277, 0.0223, 0.0265, 0.0298], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:28:12,400 INFO [train.py:904] (5/8) Epoch 12, batch 1700, loss[loss=0.1575, simple_loss=0.2465, pruned_loss=0.03424, over 16887.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2641, pruned_loss=0.04965, over 3317432.55 frames. ], batch size: 42, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:28:23,616 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.537e+02 3.070e+02 3.736e+02 6.116e+02, threshold=6.140e+02, percent-clipped=1.0 2023-04-29 13:29:22,278 INFO [train.py:904] (5/8) Epoch 12, batch 1750, loss[loss=0.1912, simple_loss=0.2791, pruned_loss=0.05159, over 16681.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2668, pruned_loss=0.05023, over 3317643.18 frames. ], batch size: 57, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:29:37,757 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-29 13:30:15,880 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7286, 1.7466, 1.5377, 1.5058, 1.8606, 1.5954, 1.7481, 1.9667], device='cuda:5'), covar=tensor([0.0128, 0.0218, 0.0306, 0.0262, 0.0150, 0.0193, 0.0144, 0.0146], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0214, 0.0206, 0.0206, 0.0214, 0.0213, 0.0220, 0.0203], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:30:32,326 INFO [train.py:904] (5/8) Epoch 12, batch 1800, loss[loss=0.1869, simple_loss=0.2686, pruned_loss=0.05262, over 16681.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2682, pruned_loss=0.05051, over 3318714.55 frames. ], batch size: 89, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:30:43,448 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.362e+02 2.904e+02 3.594e+02 5.616e+02, threshold=5.809e+02, percent-clipped=0.0 2023-04-29 13:31:42,923 INFO [train.py:904] (5/8) Epoch 12, batch 1850, loss[loss=0.2106, simple_loss=0.2845, pruned_loss=0.06836, over 16734.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2686, pruned_loss=0.05053, over 3319484.25 frames. ], batch size: 124, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:32:25,486 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7042, 4.0692, 4.2677, 3.0799, 3.6228, 4.2013, 3.8340, 2.6134], device='cuda:5'), covar=tensor([0.0396, 0.0063, 0.0034, 0.0268, 0.0088, 0.0068, 0.0067, 0.0318], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0073, 0.0071, 0.0127, 0.0081, 0.0091, 0.0080, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 13:32:53,593 INFO [train.py:904] (5/8) Epoch 12, batch 1900, loss[loss=0.1595, simple_loss=0.2433, pruned_loss=0.03788, over 16795.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2672, pruned_loss=0.04962, over 3320727.19 frames. ], batch size: 39, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:33:04,795 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 2.197e+02 2.521e+02 3.135e+02 9.394e+02, threshold=5.041e+02, percent-clipped=1.0 2023-04-29 13:33:29,770 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:33:49,004 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:34:05,628 INFO [train.py:904] (5/8) Epoch 12, batch 1950, loss[loss=0.2067, simple_loss=0.2734, pruned_loss=0.06998, over 16746.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2669, pruned_loss=0.04866, over 3313888.91 frames. ], batch size: 124, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:34:27,360 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:34:39,274 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:08,130 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:16,116 INFO [train.py:904] (5/8) Epoch 12, batch 2000, loss[loss=0.1521, simple_loss=0.2434, pruned_loss=0.03039, over 16788.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2653, pruned_loss=0.0483, over 3320168.47 frames. ], batch size: 39, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:35:27,902 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.257e+02 2.755e+02 3.569e+02 6.259e+02, threshold=5.509e+02, percent-clipped=3.0 2023-04-29 13:35:28,307 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5091, 5.9753, 5.6519, 5.7307, 5.3047, 5.2837, 5.3761, 6.0585], device='cuda:5'), covar=tensor([0.1353, 0.0837, 0.1098, 0.0720, 0.0892, 0.0645, 0.1066, 0.0804], device='cuda:5'), in_proj_covar=tensor([0.0577, 0.0716, 0.0584, 0.0503, 0.0449, 0.0462, 0.0598, 0.0548], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:35:52,995 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:36:04,306 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3029, 4.1331, 4.3500, 4.4892, 4.5590, 4.1333, 4.2812, 4.5350], device='cuda:5'), covar=tensor([0.1206, 0.1054, 0.1161, 0.0569, 0.0557, 0.1161, 0.2429, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0571, 0.0714, 0.0858, 0.0727, 0.0547, 0.0560, 0.0566, 0.0654], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:36:25,543 INFO [train.py:904] (5/8) Epoch 12, batch 2050, loss[loss=0.1811, simple_loss=0.2589, pruned_loss=0.05165, over 16562.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2656, pruned_loss=0.04893, over 3312041.60 frames. ], batch size: 75, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:36:31,385 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2163, 4.1823, 4.5345, 2.2652, 4.8052, 4.7401, 3.1845, 3.7445], device='cuda:5'), covar=tensor([0.0592, 0.0186, 0.0181, 0.0994, 0.0043, 0.0122, 0.0368, 0.0317], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0099, 0.0088, 0.0138, 0.0070, 0.0107, 0.0120, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 13:36:32,487 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:37:33,723 INFO [train.py:904] (5/8) Epoch 12, batch 2100, loss[loss=0.2226, simple_loss=0.2952, pruned_loss=0.07503, over 16797.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2672, pruned_loss=0.05002, over 3322228.98 frames. ], batch size: 124, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:37:45,426 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.602e+02 3.022e+02 3.669e+02 5.748e+02, threshold=6.044e+02, percent-clipped=1.0 2023-04-29 13:38:44,794 INFO [train.py:904] (5/8) Epoch 12, batch 2150, loss[loss=0.1918, simple_loss=0.2728, pruned_loss=0.05544, over 16835.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2678, pruned_loss=0.05033, over 3322896.04 frames. ], batch size: 96, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:39:11,692 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5984, 3.1090, 2.7106, 5.0330, 4.0400, 4.4984, 1.6249, 3.1501], device='cuda:5'), covar=tensor([0.1384, 0.0651, 0.1076, 0.0155, 0.0241, 0.0364, 0.1470, 0.0736], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0160, 0.0180, 0.0149, 0.0196, 0.0212, 0.0181, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 13:39:54,122 INFO [train.py:904] (5/8) Epoch 12, batch 2200, loss[loss=0.1802, simple_loss=0.2682, pruned_loss=0.04616, over 17119.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2684, pruned_loss=0.05055, over 3325934.89 frames. ], batch size: 49, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:40:05,126 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.389e+02 2.779e+02 3.462e+02 6.586e+02, threshold=5.558e+02, percent-clipped=1.0 2023-04-29 13:40:48,908 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:41:03,701 INFO [train.py:904] (5/8) Epoch 12, batch 2250, loss[loss=0.1558, simple_loss=0.2491, pruned_loss=0.03125, over 17189.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2688, pruned_loss=0.05102, over 3323485.42 frames. ], batch size: 46, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:41:54,643 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:42:13,511 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3674, 2.1072, 1.6616, 1.9679, 2.4072, 2.2518, 2.4438, 2.6257], device='cuda:5'), covar=tensor([0.0152, 0.0261, 0.0375, 0.0344, 0.0173, 0.0235, 0.0178, 0.0194], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0212, 0.0204, 0.0203, 0.0212, 0.0210, 0.0219, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:42:14,198 INFO [train.py:904] (5/8) Epoch 12, batch 2300, loss[loss=0.1776, simple_loss=0.2536, pruned_loss=0.05082, over 16302.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2685, pruned_loss=0.05015, over 3324549.40 frames. ], batch size: 165, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:42:24,219 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.413e+02 2.819e+02 3.398e+02 7.599e+02, threshold=5.637e+02, percent-clipped=2.0 2023-04-29 13:42:43,133 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:43:25,591 INFO [train.py:904] (5/8) Epoch 12, batch 2350, loss[loss=0.2032, simple_loss=0.2951, pruned_loss=0.05566, over 17179.00 frames. ], tot_loss[loss=0.185, simple_loss=0.269, pruned_loss=0.05045, over 3331169.40 frames. ], batch size: 46, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:43:25,934 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:43:52,529 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3497, 4.1871, 4.3989, 4.5519, 4.6490, 4.2200, 4.4138, 4.6064], device='cuda:5'), covar=tensor([0.1284, 0.0939, 0.1174, 0.0582, 0.0526, 0.1065, 0.1538, 0.0589], device='cuda:5'), in_proj_covar=tensor([0.0563, 0.0707, 0.0848, 0.0722, 0.0540, 0.0554, 0.0562, 0.0644], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:44:35,987 INFO [train.py:904] (5/8) Epoch 12, batch 2400, loss[loss=0.1855, simple_loss=0.265, pruned_loss=0.05301, over 16846.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2704, pruned_loss=0.05148, over 3322711.30 frames. ], batch size: 96, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:44:48,486 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.062e+02 2.572e+02 3.152e+02 4.032e+02 9.919e+02, threshold=6.305e+02, percent-clipped=6.0 2023-04-29 13:45:21,652 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7649, 3.0073, 2.4408, 4.2915, 3.5546, 4.2017, 1.6177, 2.9565], device='cuda:5'), covar=tensor([0.1214, 0.0516, 0.1043, 0.0135, 0.0208, 0.0324, 0.1262, 0.0720], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0160, 0.0181, 0.0151, 0.0197, 0.0212, 0.0181, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 13:45:49,052 INFO [train.py:904] (5/8) Epoch 12, batch 2450, loss[loss=0.1947, simple_loss=0.2903, pruned_loss=0.04954, over 17049.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2718, pruned_loss=0.05144, over 3314241.51 frames. ], batch size: 55, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:46:23,404 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:46:30,518 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8470, 3.9293, 4.3066, 2.0197, 4.5186, 4.4814, 3.1098, 3.4828], device='cuda:5'), covar=tensor([0.0694, 0.0197, 0.0226, 0.1125, 0.0057, 0.0174, 0.0383, 0.0343], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0101, 0.0089, 0.0138, 0.0070, 0.0109, 0.0121, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 13:46:57,384 INFO [train.py:904] (5/8) Epoch 12, batch 2500, loss[loss=0.2201, simple_loss=0.2978, pruned_loss=0.07125, over 16444.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2715, pruned_loss=0.05148, over 3320339.71 frames. ], batch size: 146, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:47:09,684 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.290e+02 2.668e+02 3.158e+02 5.614e+02, threshold=5.335e+02, percent-clipped=0.0 2023-04-29 13:47:48,687 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:47:55,342 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6057, 4.6095, 5.0618, 4.9930, 5.0463, 4.6726, 4.6997, 4.4891], device='cuda:5'), covar=tensor([0.0348, 0.0567, 0.0370, 0.0465, 0.0525, 0.0401, 0.0893, 0.0532], device='cuda:5'), in_proj_covar=tensor([0.0357, 0.0370, 0.0375, 0.0349, 0.0420, 0.0394, 0.0500, 0.0316], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 13:48:06,984 INFO [train.py:904] (5/8) Epoch 12, batch 2550, loss[loss=0.1936, simple_loss=0.2618, pruned_loss=0.06268, over 16672.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2711, pruned_loss=0.05167, over 3328377.47 frames. ], batch size: 134, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:49:15,401 INFO [train.py:904] (5/8) Epoch 12, batch 2600, loss[loss=0.1818, simple_loss=0.2775, pruned_loss=0.043, over 17033.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2702, pruned_loss=0.05103, over 3331164.86 frames. ], batch size: 53, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:49:25,934 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.778e+02 2.446e+02 2.880e+02 3.491e+02 5.288e+02, threshold=5.760e+02, percent-clipped=0.0 2023-04-29 13:49:45,730 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:49:52,262 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4137, 3.5731, 3.7889, 1.8040, 3.9114, 3.8511, 3.0728, 2.8622], device='cuda:5'), covar=tensor([0.0736, 0.0152, 0.0129, 0.1112, 0.0064, 0.0157, 0.0372, 0.0402], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0102, 0.0090, 0.0139, 0.0071, 0.0110, 0.0122, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 13:50:03,858 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8717, 4.1431, 2.4526, 4.7232, 3.1698, 4.6322, 2.6523, 3.2632], device='cuda:5'), covar=tensor([0.0231, 0.0284, 0.1361, 0.0180, 0.0661, 0.0352, 0.1323, 0.0607], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0168, 0.0189, 0.0140, 0.0170, 0.0212, 0.0197, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 13:50:18,095 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3188, 4.2843, 4.4938, 4.3119, 4.2981, 4.9476, 4.5041, 4.1864], device='cuda:5'), covar=tensor([0.1691, 0.2083, 0.2115, 0.2144, 0.2993, 0.1141, 0.1470, 0.2636], device='cuda:5'), in_proj_covar=tensor([0.0363, 0.0519, 0.0563, 0.0441, 0.0597, 0.0591, 0.0446, 0.0597], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 13:50:24,332 INFO [train.py:904] (5/8) Epoch 12, batch 2650, loss[loss=0.189, simple_loss=0.2855, pruned_loss=0.04622, over 17073.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2701, pruned_loss=0.05017, over 3336517.11 frames. ], batch size: 53, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:50:24,607 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:50:51,496 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:51:32,417 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:51:35,147 INFO [train.py:904] (5/8) Epoch 12, batch 2700, loss[loss=0.1699, simple_loss=0.2659, pruned_loss=0.03699, over 16710.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2703, pruned_loss=0.04988, over 3327505.84 frames. ], batch size: 62, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:51:45,487 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.425e+02 2.890e+02 3.523e+02 1.000e+03, threshold=5.781e+02, percent-clipped=5.0 2023-04-29 13:52:44,866 INFO [train.py:904] (5/8) Epoch 12, batch 2750, loss[loss=0.2, simple_loss=0.2904, pruned_loss=0.0548, over 16781.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2703, pruned_loss=0.04926, over 3323693.09 frames. ], batch size: 62, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:53:05,808 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5195, 2.4062, 1.9242, 2.1298, 2.7670, 2.5189, 3.2556, 3.1223], device='cuda:5'), covar=tensor([0.0110, 0.0356, 0.0431, 0.0404, 0.0254, 0.0326, 0.0215, 0.0211], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0214, 0.0205, 0.0206, 0.0214, 0.0212, 0.0223, 0.0205], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:53:07,686 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9158, 4.0769, 2.5284, 4.6917, 3.0604, 4.6301, 2.5943, 3.2117], device='cuda:5'), covar=tensor([0.0245, 0.0313, 0.1367, 0.0194, 0.0782, 0.0400, 0.1434, 0.0690], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0170, 0.0190, 0.0141, 0.0171, 0.0215, 0.0200, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 13:53:21,319 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8490, 4.0072, 2.4636, 4.6369, 2.9348, 4.5568, 2.5293, 3.1529], device='cuda:5'), covar=tensor([0.0226, 0.0322, 0.1349, 0.0204, 0.0824, 0.0421, 0.1405, 0.0675], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0170, 0.0190, 0.0141, 0.0171, 0.0215, 0.0199, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 13:53:54,633 INFO [train.py:904] (5/8) Epoch 12, batch 2800, loss[loss=0.1687, simple_loss=0.2623, pruned_loss=0.03751, over 17123.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2703, pruned_loss=0.04935, over 3332566.40 frames. ], batch size: 48, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:54:06,077 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.386e+02 2.760e+02 3.486e+02 6.287e+02, threshold=5.520e+02, percent-clipped=1.0 2023-04-29 13:54:32,863 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4594, 1.5914, 2.0149, 2.2805, 2.3872, 2.3466, 1.6457, 2.4701], device='cuda:5'), covar=tensor([0.0153, 0.0358, 0.0235, 0.0181, 0.0221, 0.0234, 0.0370, 0.0096], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0177, 0.0159, 0.0165, 0.0175, 0.0131, 0.0175, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 13:54:38,722 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:55:04,071 INFO [train.py:904] (5/8) Epoch 12, batch 2850, loss[loss=0.1818, simple_loss=0.2721, pruned_loss=0.04577, over 16540.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2699, pruned_loss=0.04966, over 3324236.22 frames. ], batch size: 68, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:55:16,370 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 13:56:13,214 INFO [train.py:904] (5/8) Epoch 12, batch 2900, loss[loss=0.1622, simple_loss=0.2351, pruned_loss=0.04467, over 16775.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2677, pruned_loss=0.04915, over 3333332.64 frames. ], batch size: 83, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:56:24,539 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.459e+02 2.961e+02 3.423e+02 5.764e+02, threshold=5.923e+02, percent-clipped=1.0 2023-04-29 13:56:42,479 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 13:57:20,471 INFO [train.py:904] (5/8) Epoch 12, batch 2950, loss[loss=0.1504, simple_loss=0.2407, pruned_loss=0.03006, over 16879.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2681, pruned_loss=0.05018, over 3320750.64 frames. ], batch size: 39, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:58:08,947 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 13:58:28,576 INFO [train.py:904] (5/8) Epoch 12, batch 3000, loss[loss=0.2027, simple_loss=0.2747, pruned_loss=0.06538, over 16726.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2674, pruned_loss=0.0503, over 3324251.63 frames. ], batch size: 134, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:58:28,577 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 13:58:38,475 INFO [train.py:938] (5/8) Epoch 12, validation: loss=0.14, simple_loss=0.2459, pruned_loss=0.01708, over 944034.00 frames. 2023-04-29 13:58:38,476 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 13:58:40,794 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4350, 3.0032, 2.6531, 2.2587, 2.2587, 2.1592, 2.9219, 2.8604], device='cuda:5'), covar=tensor([0.2073, 0.0721, 0.1281, 0.1834, 0.1913, 0.1705, 0.0476, 0.0946], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0256, 0.0281, 0.0277, 0.0279, 0.0221, 0.0268, 0.0300], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:58:50,230 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.323e+02 2.752e+02 3.480e+02 1.028e+03, threshold=5.504e+02, percent-clipped=2.0 2023-04-29 13:58:51,140 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-29 13:59:17,943 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5765, 6.0438, 5.7186, 5.8336, 5.3597, 5.3077, 5.4685, 6.1332], device='cuda:5'), covar=tensor([0.1042, 0.0839, 0.1069, 0.0736, 0.0847, 0.0616, 0.0946, 0.0840], device='cuda:5'), in_proj_covar=tensor([0.0586, 0.0727, 0.0600, 0.0511, 0.0459, 0.0463, 0.0605, 0.0562], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 13:59:45,916 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 13:59:48,680 INFO [train.py:904] (5/8) Epoch 12, batch 3050, loss[loss=0.1976, simple_loss=0.2663, pruned_loss=0.06449, over 16770.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2672, pruned_loss=0.05036, over 3314198.15 frames. ], batch size: 124, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:59:51,586 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-04-29 14:00:09,025 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:00:42,641 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5615, 3.6401, 3.2890, 3.0799, 3.2106, 3.5304, 3.3091, 3.2824], device='cuda:5'), covar=tensor([0.0560, 0.0441, 0.0254, 0.0224, 0.0536, 0.0351, 0.1451, 0.0446], device='cuda:5'), in_proj_covar=tensor([0.0265, 0.0348, 0.0319, 0.0295, 0.0337, 0.0340, 0.0216, 0.0368], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:00:56,637 INFO [train.py:904] (5/8) Epoch 12, batch 3100, loss[loss=0.1893, simple_loss=0.2588, pruned_loss=0.05988, over 16793.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2668, pruned_loss=0.05034, over 3317264.50 frames. ], batch size: 124, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:01:06,013 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2827, 4.3115, 4.6876, 4.6923, 4.7003, 4.4024, 4.4032, 4.2840], device='cuda:5'), covar=tensor([0.0328, 0.0633, 0.0392, 0.0429, 0.0517, 0.0363, 0.0861, 0.0521], device='cuda:5'), in_proj_covar=tensor([0.0357, 0.0373, 0.0375, 0.0349, 0.0418, 0.0393, 0.0506, 0.0317], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 14:01:10,361 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.272e+02 2.804e+02 3.413e+02 6.523e+02, threshold=5.608e+02, percent-clipped=1.0 2023-04-29 14:01:26,093 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-29 14:01:32,639 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:01:40,418 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:02:05,285 INFO [train.py:904] (5/8) Epoch 12, batch 3150, loss[loss=0.1639, simple_loss=0.2552, pruned_loss=0.03625, over 17105.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2668, pruned_loss=0.05069, over 3326214.28 frames. ], batch size: 48, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:02:24,795 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 14:02:45,208 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:02:56,618 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 14:03:14,011 INFO [train.py:904] (5/8) Epoch 12, batch 3200, loss[loss=0.1925, simple_loss=0.277, pruned_loss=0.05395, over 16701.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.266, pruned_loss=0.0504, over 3324693.57 frames. ], batch size: 62, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:03:26,056 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.649e+02 2.483e+02 2.978e+02 3.563e+02 6.234e+02, threshold=5.956e+02, percent-clipped=4.0 2023-04-29 14:03:28,475 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6453, 2.5646, 2.3601, 3.9430, 3.2869, 4.0034, 1.4444, 2.8217], device='cuda:5'), covar=tensor([0.1273, 0.0631, 0.1055, 0.0160, 0.0157, 0.0311, 0.1383, 0.0740], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0160, 0.0181, 0.0153, 0.0199, 0.0213, 0.0181, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 14:03:40,781 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:04:22,409 INFO [train.py:904] (5/8) Epoch 12, batch 3250, loss[loss=0.207, simple_loss=0.2992, pruned_loss=0.05739, over 16808.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2674, pruned_loss=0.05116, over 3327167.18 frames. ], batch size: 57, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:05,239 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114932.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:05:08,085 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3268, 5.2810, 5.0883, 4.4928, 5.1064, 1.8715, 4.8943, 5.1210], device='cuda:5'), covar=tensor([0.0068, 0.0065, 0.0142, 0.0365, 0.0080, 0.2420, 0.0108, 0.0142], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0127, 0.0177, 0.0167, 0.0147, 0.0188, 0.0165, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:05:32,347 INFO [train.py:904] (5/8) Epoch 12, batch 3300, loss[loss=0.162, simple_loss=0.2472, pruned_loss=0.0384, over 17236.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2683, pruned_loss=0.05082, over 3326297.97 frames. ], batch size: 43, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:33,972 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9914, 4.2458, 2.6381, 4.8027, 3.1819, 4.7867, 2.8447, 3.4526], device='cuda:5'), covar=tensor([0.0227, 0.0312, 0.1360, 0.0158, 0.0649, 0.0364, 0.1230, 0.0570], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0168, 0.0189, 0.0140, 0.0169, 0.0213, 0.0198, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 14:05:45,365 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.351e+02 3.077e+02 3.881e+02 7.792e+02, threshold=6.153e+02, percent-clipped=7.0 2023-04-29 14:06:26,186 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0173, 4.0289, 3.8575, 3.6440, 3.6609, 4.0169, 3.6836, 3.7741], device='cuda:5'), covar=tensor([0.0658, 0.0580, 0.0314, 0.0278, 0.0755, 0.0439, 0.0942, 0.0615], device='cuda:5'), in_proj_covar=tensor([0.0270, 0.0354, 0.0324, 0.0298, 0.0342, 0.0345, 0.0218, 0.0376], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:06:42,024 INFO [train.py:904] (5/8) Epoch 12, batch 3350, loss[loss=0.224, simple_loss=0.3106, pruned_loss=0.0687, over 12231.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2686, pruned_loss=0.05083, over 3314549.86 frames. ], batch size: 246, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:07:14,445 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2716, 4.0708, 4.3164, 4.4897, 4.5733, 4.1377, 4.3812, 4.5641], device='cuda:5'), covar=tensor([0.1284, 0.1013, 0.1268, 0.0571, 0.0558, 0.1144, 0.1364, 0.0625], device='cuda:5'), in_proj_covar=tensor([0.0577, 0.0718, 0.0871, 0.0732, 0.0549, 0.0566, 0.0572, 0.0656], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:07:50,790 INFO [train.py:904] (5/8) Epoch 12, batch 3400, loss[loss=0.1608, simple_loss=0.2446, pruned_loss=0.0385, over 16750.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2672, pruned_loss=0.04991, over 3320985.36 frames. ], batch size: 39, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:07:54,534 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7468, 4.4714, 4.7655, 4.9469, 5.1056, 4.4886, 5.1218, 5.0984], device='cuda:5'), covar=tensor([0.1465, 0.1182, 0.1663, 0.0678, 0.0564, 0.1030, 0.0549, 0.0632], device='cuda:5'), in_proj_covar=tensor([0.0575, 0.0715, 0.0869, 0.0729, 0.0547, 0.0565, 0.0569, 0.0653], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:08:04,046 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.344e+02 2.703e+02 3.237e+02 1.034e+03, threshold=5.406e+02, percent-clipped=1.0 2023-04-29 14:08:13,674 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-29 14:08:18,316 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:08:37,800 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:09:00,025 INFO [train.py:904] (5/8) Epoch 12, batch 3450, loss[loss=0.193, simple_loss=0.2607, pruned_loss=0.06268, over 16750.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.267, pruned_loss=0.04988, over 3321086.99 frames. ], batch size: 124, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:09:44,397 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8495, 2.3726, 2.3951, 4.7087, 2.2701, 2.7771, 2.4835, 2.5772], device='cuda:5'), covar=tensor([0.0902, 0.3340, 0.2418, 0.0361, 0.3804, 0.2353, 0.2879, 0.3352], device='cuda:5'), in_proj_covar=tensor([0.0375, 0.0400, 0.0337, 0.0328, 0.0416, 0.0463, 0.0364, 0.0469], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:10:02,314 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:10:08,332 INFO [train.py:904] (5/8) Epoch 12, batch 3500, loss[loss=0.1597, simple_loss=0.2411, pruned_loss=0.03919, over 17214.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2651, pruned_loss=0.04886, over 3330267.58 frames. ], batch size: 44, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:10:23,311 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.367e+02 2.772e+02 3.300e+02 1.193e+03, threshold=5.543e+02, percent-clipped=2.0 2023-04-29 14:11:19,463 INFO [train.py:904] (5/8) Epoch 12, batch 3550, loss[loss=0.2029, simple_loss=0.2808, pruned_loss=0.06252, over 11975.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2648, pruned_loss=0.04892, over 3319629.57 frames. ], batch size: 247, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:11:53,671 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:11:54,845 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9278, 5.0202, 5.4906, 5.4652, 5.3941, 5.0594, 5.0681, 4.8910], device='cuda:5'), covar=tensor([0.0302, 0.0446, 0.0321, 0.0377, 0.0480, 0.0349, 0.0853, 0.0370], device='cuda:5'), in_proj_covar=tensor([0.0362, 0.0379, 0.0377, 0.0356, 0.0426, 0.0396, 0.0509, 0.0320], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 14:12:28,662 INFO [train.py:904] (5/8) Epoch 12, batch 3600, loss[loss=0.1524, simple_loss=0.2329, pruned_loss=0.03593, over 16180.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2641, pruned_loss=0.04829, over 3321752.60 frames. ], batch size: 36, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:12:43,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.756e+02 2.259e+02 2.631e+02 3.384e+02 1.021e+03, threshold=5.262e+02, percent-clipped=2.0 2023-04-29 14:13:40,320 INFO [train.py:904] (5/8) Epoch 12, batch 3650, loss[loss=0.1692, simple_loss=0.2365, pruned_loss=0.05091, over 16463.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2616, pruned_loss=0.04882, over 3314584.98 frames. ], batch size: 75, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:14:14,009 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5727, 3.7060, 3.9731, 2.8396, 3.6822, 4.0949, 3.8574, 2.3532], device='cuda:5'), covar=tensor([0.0395, 0.0116, 0.0046, 0.0279, 0.0065, 0.0065, 0.0055, 0.0366], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0071, 0.0072, 0.0125, 0.0081, 0.0091, 0.0080, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:14:53,093 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4724, 4.4602, 4.6754, 4.4931, 4.5301, 5.1061, 4.6690, 4.3131], device='cuda:5'), covar=tensor([0.1579, 0.1919, 0.1726, 0.2018, 0.2611, 0.1018, 0.1248, 0.2380], device='cuda:5'), in_proj_covar=tensor([0.0360, 0.0513, 0.0557, 0.0436, 0.0591, 0.0578, 0.0440, 0.0592], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:14:55,152 INFO [train.py:904] (5/8) Epoch 12, batch 3700, loss[loss=0.1994, simple_loss=0.2595, pruned_loss=0.06966, over 16701.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2604, pruned_loss=0.0507, over 3299501.68 frames. ], batch size: 83, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:15:02,205 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6626, 5.0540, 4.7878, 4.8325, 4.5700, 4.4667, 4.5722, 5.0922], device='cuda:5'), covar=tensor([0.1077, 0.0777, 0.0985, 0.0602, 0.0745, 0.1114, 0.0879, 0.0879], device='cuda:5'), in_proj_covar=tensor([0.0576, 0.0718, 0.0584, 0.0502, 0.0448, 0.0456, 0.0593, 0.0558], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:15:09,331 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.309e+02 2.720e+02 3.193e+02 6.265e+02, threshold=5.440e+02, percent-clipped=2.0 2023-04-29 14:15:23,054 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:24,707 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:25,906 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:36,232 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0690, 3.2743, 3.3108, 2.1992, 2.7298, 2.3389, 3.5906, 3.4880], device='cuda:5'), covar=tensor([0.0218, 0.0784, 0.0537, 0.1590, 0.0816, 0.0901, 0.0456, 0.0808], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0149, 0.0159, 0.0144, 0.0137, 0.0124, 0.0138, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 14:16:10,087 INFO [train.py:904] (5/8) Epoch 12, batch 3750, loss[loss=0.1891, simple_loss=0.2753, pruned_loss=0.05142, over 16240.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2617, pruned_loss=0.05215, over 3272357.79 frames. ], batch size: 165, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:16:12,079 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 14:16:36,737 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:47,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3463, 3.4373, 3.4986, 2.1385, 2.9780, 2.4648, 3.8612, 3.7110], device='cuda:5'), covar=tensor([0.0187, 0.0766, 0.0558, 0.1694, 0.0777, 0.0882, 0.0371, 0.0709], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0149, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 14:16:54,232 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:55,367 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:17:03,938 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-04-29 14:17:09,505 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:17:15,973 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.87 vs. limit=5.0 2023-04-29 14:17:23,459 INFO [train.py:904] (5/8) Epoch 12, batch 3800, loss[loss=0.208, simple_loss=0.2815, pruned_loss=0.06723, over 16246.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2631, pruned_loss=0.05345, over 3268638.12 frames. ], batch size: 165, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:17:38,965 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.466e+02 2.817e+02 3.424e+02 6.065e+02, threshold=5.633e+02, percent-clipped=2.0 2023-04-29 14:18:16,528 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1812, 4.2482, 4.6090, 4.5930, 4.6155, 4.2773, 4.3364, 4.1633], device='cuda:5'), covar=tensor([0.0332, 0.0549, 0.0329, 0.0373, 0.0417, 0.0354, 0.0738, 0.0529], device='cuda:5'), in_proj_covar=tensor([0.0356, 0.0371, 0.0370, 0.0345, 0.0417, 0.0388, 0.0498, 0.0313], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 14:18:37,590 INFO [train.py:904] (5/8) Epoch 12, batch 3850, loss[loss=0.1803, simple_loss=0.2544, pruned_loss=0.05313, over 16234.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2635, pruned_loss=0.05403, over 3266937.33 frames. ], batch size: 165, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:19:16,874 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:19:52,702 INFO [train.py:904] (5/8) Epoch 12, batch 3900, loss[loss=0.1728, simple_loss=0.2584, pruned_loss=0.04367, over 16646.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2634, pruned_loss=0.05503, over 3272749.38 frames. ], batch size: 57, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:20:07,963 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.153e+02 2.652e+02 3.211e+02 6.333e+02, threshold=5.304e+02, percent-clipped=2.0 2023-04-29 14:20:29,520 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:21:08,908 INFO [train.py:904] (5/8) Epoch 12, batch 3950, loss[loss=0.1841, simple_loss=0.2547, pruned_loss=0.0567, over 16441.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2634, pruned_loss=0.05613, over 3261203.45 frames. ], batch size: 146, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:21:15,336 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2992, 5.2532, 5.0385, 4.3894, 5.2120, 1.8954, 4.9223, 4.8553], device='cuda:5'), covar=tensor([0.0061, 0.0049, 0.0126, 0.0343, 0.0052, 0.2474, 0.0105, 0.0162], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0127, 0.0177, 0.0168, 0.0149, 0.0189, 0.0166, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:22:21,488 INFO [train.py:904] (5/8) Epoch 12, batch 4000, loss[loss=0.1787, simple_loss=0.2557, pruned_loss=0.05087, over 16502.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2629, pruned_loss=0.05594, over 3270549.69 frames. ], batch size: 75, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:22:28,167 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 14:22:34,739 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.376e+02 2.718e+02 3.245e+02 5.190e+02, threshold=5.435e+02, percent-clipped=0.0 2023-04-29 14:23:35,781 INFO [train.py:904] (5/8) Epoch 12, batch 4050, loss[loss=0.1805, simple_loss=0.2654, pruned_loss=0.04776, over 16391.00 frames. ], tot_loss[loss=0.186, simple_loss=0.263, pruned_loss=0.05446, over 3274196.09 frames. ], batch size: 146, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:24:12,128 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:13,920 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:36,676 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115742.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:50,507 INFO [train.py:904] (5/8) Epoch 12, batch 4100, loss[loss=0.212, simple_loss=0.2959, pruned_loss=0.06403, over 16899.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2641, pruned_loss=0.0533, over 3275724.36 frames. ], batch size: 109, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:24:52,878 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 14:25:05,531 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.445e+02 1.888e+02 2.192e+02 2.609e+02 4.553e+02, threshold=4.384e+02, percent-clipped=0.0 2023-04-29 14:25:17,704 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4457, 3.4343, 2.5953, 2.1570, 2.3522, 2.2582, 3.5410, 3.2156], device='cuda:5'), covar=tensor([0.2765, 0.0696, 0.1638, 0.2118, 0.2270, 0.1785, 0.0534, 0.1094], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0261, 0.0288, 0.0284, 0.0289, 0.0226, 0.0271, 0.0305], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:25:48,338 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:26:05,825 INFO [train.py:904] (5/8) Epoch 12, batch 4150, loss[loss=0.2282, simple_loss=0.3099, pruned_loss=0.07323, over 16917.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2723, pruned_loss=0.05651, over 3238449.88 frames. ], batch size: 109, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:23,040 INFO [train.py:904] (5/8) Epoch 12, batch 4200, loss[loss=0.2257, simple_loss=0.3123, pruned_loss=0.06959, over 16244.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2793, pruned_loss=0.05803, over 3220346.80 frames. ], batch size: 35, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:37,144 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.546e+02 2.889e+02 3.538e+02 7.743e+02, threshold=5.778e+02, percent-clipped=11.0 2023-04-29 14:27:47,456 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7680, 1.7897, 2.2430, 2.7594, 2.6600, 2.9846, 2.0625, 2.8799], device='cuda:5'), covar=tensor([0.0154, 0.0366, 0.0241, 0.0183, 0.0198, 0.0130, 0.0311, 0.0123], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0176, 0.0158, 0.0162, 0.0174, 0.0130, 0.0174, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 14:28:11,323 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7097, 3.8723, 2.1485, 4.1216, 2.8165, 4.1552, 2.0929, 2.9124], device='cuda:5'), covar=tensor([0.0190, 0.0219, 0.1454, 0.0153, 0.0648, 0.0441, 0.1508, 0.0637], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0165, 0.0188, 0.0135, 0.0168, 0.0208, 0.0194, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 14:28:22,401 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:28:36,628 INFO [train.py:904] (5/8) Epoch 12, batch 4250, loss[loss=0.193, simple_loss=0.282, pruned_loss=0.05195, over 17124.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2822, pruned_loss=0.05755, over 3216288.18 frames. ], batch size: 48, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:28:57,537 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 14:28:58,644 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9843, 2.3259, 2.3952, 2.7476, 2.0767, 3.2487, 1.7896, 2.8072], device='cuda:5'), covar=tensor([0.0936, 0.0505, 0.0840, 0.0145, 0.0109, 0.0369, 0.1138, 0.0552], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0160, 0.0180, 0.0152, 0.0199, 0.0210, 0.0182, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 14:29:18,459 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4127, 3.4922, 1.9263, 3.8387, 2.4737, 3.8365, 2.0694, 2.6069], device='cuda:5'), covar=tensor([0.0222, 0.0303, 0.1680, 0.0151, 0.0862, 0.0438, 0.1562, 0.0819], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0166, 0.0189, 0.0135, 0.0168, 0.0208, 0.0195, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 14:29:20,898 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9673, 3.4115, 3.5055, 3.4641, 3.4523, 3.3230, 2.9942, 3.3903], device='cuda:5'), covar=tensor([0.0685, 0.0768, 0.0593, 0.0678, 0.0756, 0.0680, 0.1417, 0.0642], device='cuda:5'), in_proj_covar=tensor([0.0341, 0.0355, 0.0356, 0.0335, 0.0401, 0.0375, 0.0478, 0.0302], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 14:29:49,153 INFO [train.py:904] (5/8) Epoch 12, batch 4300, loss[loss=0.186, simple_loss=0.278, pruned_loss=0.04702, over 16534.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.283, pruned_loss=0.05666, over 3199694.24 frames. ], batch size: 68, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:29:50,910 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:29:52,487 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 14:30:04,742 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.520e+02 2.885e+02 3.320e+02 6.529e+02, threshold=5.769e+02, percent-clipped=1.0 2023-04-29 14:30:17,915 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8129, 4.7936, 4.5817, 3.6480, 4.7164, 1.5206, 4.3924, 4.2557], device='cuda:5'), covar=tensor([0.0071, 0.0061, 0.0135, 0.0437, 0.0079, 0.2907, 0.0115, 0.0263], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0122, 0.0169, 0.0160, 0.0142, 0.0182, 0.0159, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:31:07,423 INFO [train.py:904] (5/8) Epoch 12, batch 4350, loss[loss=0.2207, simple_loss=0.3096, pruned_loss=0.0659, over 16741.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2862, pruned_loss=0.05772, over 3179864.59 frames. ], batch size: 83, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:31:27,531 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:46,134 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:47,943 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:06,663 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8289, 2.8201, 2.6540, 4.7262, 3.7043, 4.1461, 1.6533, 3.1046], device='cuda:5'), covar=tensor([0.1260, 0.0719, 0.1136, 0.0099, 0.0351, 0.0390, 0.1461, 0.0790], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0160, 0.0181, 0.0152, 0.0200, 0.0210, 0.0182, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 14:32:22,086 INFO [train.py:904] (5/8) Epoch 12, batch 4400, loss[loss=0.199, simple_loss=0.2865, pruned_loss=0.05576, over 15242.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2883, pruned_loss=0.05896, over 3188522.05 frames. ], batch size: 190, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:32:37,557 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.512e+02 3.045e+02 3.645e+02 6.621e+02, threshold=6.090e+02, percent-clipped=4.0 2023-04-29 14:32:55,647 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:57,361 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:57,574 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:33:35,436 INFO [train.py:904] (5/8) Epoch 12, batch 4450, loss[loss=0.2422, simple_loss=0.3069, pruned_loss=0.08874, over 11546.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.292, pruned_loss=0.06038, over 3199442.95 frames. ], batch size: 246, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:33:40,112 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:33:57,009 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:34:24,823 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0810, 3.3925, 3.5255, 3.4760, 3.4915, 3.3214, 3.1465, 3.3808], device='cuda:5'), covar=tensor([0.0527, 0.0633, 0.0516, 0.0615, 0.0669, 0.0585, 0.1397, 0.0644], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0351, 0.0352, 0.0333, 0.0398, 0.0372, 0.0476, 0.0298], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 14:34:49,339 INFO [train.py:904] (5/8) Epoch 12, batch 4500, loss[loss=0.1873, simple_loss=0.2686, pruned_loss=0.05299, over 16803.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2925, pruned_loss=0.06093, over 3195239.60 frames. ], batch size: 39, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:35:03,474 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 2.115e+02 2.423e+02 2.863e+02 5.089e+02, threshold=4.846e+02, percent-clipped=0.0 2023-04-29 14:35:10,888 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:35:25,718 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116177.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:35:28,429 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9129, 3.3986, 3.3508, 2.1494, 3.0057, 3.3871, 3.1481, 1.8549], device='cuda:5'), covar=tensor([0.0476, 0.0034, 0.0038, 0.0357, 0.0082, 0.0081, 0.0075, 0.0382], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0070, 0.0071, 0.0126, 0.0081, 0.0090, 0.0080, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:36:02,109 INFO [train.py:904] (5/8) Epoch 12, batch 4550, loss[loss=0.2115, simple_loss=0.3042, pruned_loss=0.05936, over 16807.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2931, pruned_loss=0.06216, over 3176898.83 frames. ], batch size: 83, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:37:08,874 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:37:14,080 INFO [train.py:904] (5/8) Epoch 12, batch 4600, loss[loss=0.1902, simple_loss=0.2758, pruned_loss=0.05234, over 16546.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.294, pruned_loss=0.06209, over 3193297.28 frames. ], batch size: 75, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:37:29,435 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 1.985e+02 2.258e+02 2.658e+02 3.690e+02, threshold=4.517e+02, percent-clipped=0.0 2023-04-29 14:37:44,086 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-04-29 14:37:57,197 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0641, 5.0879, 4.8989, 4.6253, 4.5918, 4.9786, 4.7894, 4.6277], device='cuda:5'), covar=tensor([0.0390, 0.0216, 0.0178, 0.0190, 0.0670, 0.0235, 0.0290, 0.0478], device='cuda:5'), in_proj_covar=tensor([0.0241, 0.0316, 0.0289, 0.0269, 0.0309, 0.0308, 0.0197, 0.0336], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:38:26,073 INFO [train.py:904] (5/8) Epoch 12, batch 4650, loss[loss=0.2016, simple_loss=0.2851, pruned_loss=0.05899, over 16751.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2934, pruned_loss=0.06187, over 3204527.28 frames. ], batch size: 124, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:38:28,948 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8154, 4.1479, 3.0692, 2.3552, 2.8307, 2.4316, 4.4283, 3.6096], device='cuda:5'), covar=tensor([0.2539, 0.0574, 0.1522, 0.2016, 0.2310, 0.1786, 0.0388, 0.0929], device='cuda:5'), in_proj_covar=tensor([0.0304, 0.0259, 0.0287, 0.0283, 0.0287, 0.0225, 0.0270, 0.0302], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 14:38:44,477 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-29 14:38:46,594 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6306, 4.4386, 4.3817, 2.9407, 3.8329, 4.3120, 3.9347, 2.5310], device='cuda:5'), covar=tensor([0.0392, 0.0017, 0.0022, 0.0300, 0.0057, 0.0060, 0.0049, 0.0319], device='cuda:5'), in_proj_covar=tensor([0.0126, 0.0069, 0.0070, 0.0125, 0.0080, 0.0089, 0.0079, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:38:55,774 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:39:38,311 INFO [train.py:904] (5/8) Epoch 12, batch 4700, loss[loss=0.1952, simple_loss=0.2815, pruned_loss=0.0545, over 15362.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2904, pruned_loss=0.06042, over 3194562.06 frames. ], batch size: 191, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:39:53,945 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 2.029e+02 2.445e+02 3.086e+02 5.514e+02, threshold=4.890e+02, percent-clipped=4.0 2023-04-29 14:40:05,809 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:40:16,392 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 14:40:27,498 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:40:54,105 INFO [train.py:904] (5/8) Epoch 12, batch 4750, loss[loss=0.1769, simple_loss=0.2695, pruned_loss=0.04211, over 16412.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2867, pruned_loss=0.05882, over 3186692.66 frames. ], batch size: 68, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:41:59,022 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:07,127 INFO [train.py:904] (5/8) Epoch 12, batch 4800, loss[loss=0.197, simple_loss=0.2776, pruned_loss=0.05818, over 16439.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2825, pruned_loss=0.05665, over 3193145.12 frames. ], batch size: 68, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:42:22,201 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116461.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:23,019 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.880e+02 2.223e+02 2.728e+02 4.441e+02, threshold=4.446e+02, percent-clipped=0.0 2023-04-29 14:42:27,878 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5998, 4.3796, 4.3047, 3.0045, 3.6931, 4.2711, 3.8330, 2.3363], device='cuda:5'), covar=tensor([0.0384, 0.0017, 0.0025, 0.0270, 0.0071, 0.0073, 0.0066, 0.0349], device='cuda:5'), in_proj_covar=tensor([0.0127, 0.0069, 0.0070, 0.0125, 0.0081, 0.0090, 0.0080, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:42:39,682 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:43:23,417 INFO [train.py:904] (5/8) Epoch 12, batch 4850, loss[loss=0.1896, simple_loss=0.285, pruned_loss=0.04705, over 15368.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2834, pruned_loss=0.05571, over 3192406.79 frames. ], batch size: 191, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:43:28,304 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:43:30,910 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:43:46,667 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 14:44:05,040 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 14:44:05,367 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 14:44:31,809 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:44:38,143 INFO [train.py:904] (5/8) Epoch 12, batch 4900, loss[loss=0.1888, simple_loss=0.2688, pruned_loss=0.05437, over 16848.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.282, pruned_loss=0.05405, over 3179210.55 frames. ], batch size: 42, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:44:52,630 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.192e+02 2.607e+02 3.082e+02 6.652e+02, threshold=5.215e+02, percent-clipped=3.0 2023-04-29 14:44:59,297 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:45:14,782 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 14:45:42,832 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116596.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:45:52,033 INFO [train.py:904] (5/8) Epoch 12, batch 4950, loss[loss=0.2029, simple_loss=0.2916, pruned_loss=0.05709, over 16886.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2815, pruned_loss=0.05353, over 3182825.33 frames. ], batch size: 116, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:46:16,252 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:46:27,655 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 14:47:04,269 INFO [train.py:904] (5/8) Epoch 12, batch 5000, loss[loss=0.1837, simple_loss=0.2793, pruned_loss=0.04399, over 16871.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2837, pruned_loss=0.05399, over 3180814.60 frames. ], batch size: 96, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:47:17,032 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.254e+02 2.645e+02 3.532e+02 7.072e+02, threshold=5.290e+02, percent-clipped=1.0 2023-04-29 14:47:30,341 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:47:43,245 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:47:43,379 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:15,694 INFO [train.py:904] (5/8) Epoch 12, batch 5050, loss[loss=0.1944, simple_loss=0.2724, pruned_loss=0.05819, over 16514.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.284, pruned_loss=0.05379, over 3204007.64 frames. ], batch size: 62, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:48:18,373 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:38,189 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:54,682 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 14:49:24,609 INFO [train.py:904] (5/8) Epoch 12, batch 5100, loss[loss=0.1929, simple_loss=0.2709, pruned_loss=0.05742, over 16373.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2818, pruned_loss=0.05281, over 3214154.95 frames. ], batch size: 35, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:49:37,517 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116761.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:38,224 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 2.103e+02 2.626e+02 3.146e+02 5.078e+02, threshold=5.251e+02, percent-clipped=1.0 2023-04-29 14:49:43,737 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:52,656 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:50:35,773 INFO [train.py:904] (5/8) Epoch 12, batch 5150, loss[loss=0.2046, simple_loss=0.2941, pruned_loss=0.05756, over 16920.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2822, pruned_loss=0.05249, over 3194774.95 frames. ], batch size: 109, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:50:36,704 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:50:36,876 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:50:47,865 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:03,105 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:47,923 INFO [train.py:904] (5/8) Epoch 12, batch 5200, loss[loss=0.201, simple_loss=0.3003, pruned_loss=0.05085, over 15381.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2807, pruned_loss=0.05179, over 3211769.19 frames. ], batch size: 191, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:52:00,775 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:01,644 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.168e+02 2.438e+02 3.019e+02 6.927e+02, threshold=4.876e+02, percent-clipped=1.0 2023-04-29 14:52:03,243 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6670, 4.7795, 4.9617, 4.7191, 4.9052, 5.3837, 4.8654, 4.5357], device='cuda:5'), covar=tensor([0.1022, 0.1838, 0.1662, 0.2025, 0.2486, 0.0963, 0.1456, 0.2742], device='cuda:5'), in_proj_covar=tensor([0.0352, 0.0487, 0.0525, 0.0422, 0.0568, 0.0558, 0.0422, 0.0575], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 14:52:03,347 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:58,583 INFO [train.py:904] (5/8) Epoch 12, batch 5250, loss[loss=0.1863, simple_loss=0.2791, pruned_loss=0.04678, over 16878.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2777, pruned_loss=0.05122, over 3212502.18 frames. ], batch size: 116, lr: 5.69e-03, grad_scale: 16.0 2023-04-29 14:54:11,420 INFO [train.py:904] (5/8) Epoch 12, batch 5300, loss[loss=0.171, simple_loss=0.2479, pruned_loss=0.047, over 16976.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2746, pruned_loss=0.05027, over 3211397.70 frames. ], batch size: 55, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:54:27,262 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.433e+02 2.208e+02 2.588e+02 3.045e+02 5.450e+02, threshold=5.175e+02, percent-clipped=1.0 2023-04-29 14:54:42,033 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:54:49,731 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:54:49,888 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9875, 1.8567, 2.4812, 2.9830, 2.8207, 3.4064, 2.0808, 3.2333], device='cuda:5'), covar=tensor([0.0134, 0.0362, 0.0226, 0.0174, 0.0192, 0.0097, 0.0358, 0.0083], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0172, 0.0154, 0.0160, 0.0169, 0.0125, 0.0172, 0.0118], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 14:55:21,995 INFO [train.py:904] (5/8) Epoch 12, batch 5350, loss[loss=0.1978, simple_loss=0.2803, pruned_loss=0.05761, over 11941.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2733, pruned_loss=0.04961, over 3208419.71 frames. ], batch size: 247, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:55:58,442 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:56:02,362 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5956, 1.7215, 2.1245, 2.5795, 2.5110, 3.0031, 1.7892, 2.8865], device='cuda:5'), covar=tensor([0.0185, 0.0386, 0.0282, 0.0240, 0.0238, 0.0114, 0.0379, 0.0096], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0172, 0.0154, 0.0160, 0.0170, 0.0125, 0.0172, 0.0118], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 14:56:22,638 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-29 14:56:31,817 INFO [train.py:904] (5/8) Epoch 12, batch 5400, loss[loss=0.1756, simple_loss=0.2624, pruned_loss=0.0444, over 16613.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2757, pruned_loss=0.05016, over 3203290.01 frames. ], batch size: 62, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:56:43,952 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:56:49,066 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 2.253e+02 2.451e+02 2.837e+02 5.241e+02, threshold=4.902e+02, percent-clipped=1.0 2023-04-29 14:57:11,289 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 14:57:43,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3040, 3.4045, 1.8982, 3.7513, 2.4519, 3.6779, 2.1386, 2.6644], device='cuda:5'), covar=tensor([0.0235, 0.0348, 0.1618, 0.0122, 0.0810, 0.0491, 0.1351, 0.0730], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0164, 0.0188, 0.0130, 0.0167, 0.0205, 0.0194, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 14:57:46,202 INFO [train.py:904] (5/8) Epoch 12, batch 5450, loss[loss=0.191, simple_loss=0.2858, pruned_loss=0.04815, over 17198.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.279, pruned_loss=0.05176, over 3203014.45 frames. ], batch size: 44, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:57:46,721 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:58:57,077 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:58:59,928 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 14:59:00,359 INFO [train.py:904] (5/8) Epoch 12, batch 5500, loss[loss=0.2425, simple_loss=0.3265, pruned_loss=0.07923, over 16463.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2864, pruned_loss=0.05669, over 3177390.80 frames. ], batch size: 75, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:59:09,314 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:13,859 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:18,158 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.773e+02 3.440e+02 4.409e+02 8.971e+02, threshold=6.880e+02, percent-clipped=17.0 2023-04-29 15:00:18,010 INFO [train.py:904] (5/8) Epoch 12, batch 5550, loss[loss=0.2375, simple_loss=0.3199, pruned_loss=0.0775, over 16690.00 frames. ], tot_loss[loss=0.211, simple_loss=0.295, pruned_loss=0.06345, over 3123690.23 frames. ], batch size: 124, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:00:30,367 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:00:49,816 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:00:51,453 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-29 15:01:39,155 INFO [train.py:904] (5/8) Epoch 12, batch 5600, loss[loss=0.2624, simple_loss=0.335, pruned_loss=0.09489, over 15194.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3006, pruned_loss=0.06882, over 3073028.79 frames. ], batch size: 190, lr: 5.68e-03, grad_scale: 8.0 2023-04-29 15:01:58,904 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.404e+02 3.697e+02 4.299e+02 5.006e+02 8.998e+02, threshold=8.599e+02, percent-clipped=4.0 2023-04-29 15:02:16,783 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:02:30,456 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117282.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:02:37,058 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 15:02:48,736 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4818, 3.4862, 2.6105, 2.0754, 2.4185, 2.1195, 3.6944, 3.3208], device='cuda:5'), covar=tensor([0.2758, 0.0717, 0.1712, 0.2336, 0.2263, 0.1956, 0.0499, 0.1069], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0257, 0.0284, 0.0281, 0.0281, 0.0221, 0.0268, 0.0299], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:03:02,074 INFO [train.py:904] (5/8) Epoch 12, batch 5650, loss[loss=0.2631, simple_loss=0.334, pruned_loss=0.09609, over 11403.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.305, pruned_loss=0.07188, over 3071003.77 frames. ], batch size: 248, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:03:33,952 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:45,054 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:04:18,934 INFO [train.py:904] (5/8) Epoch 12, batch 5700, loss[loss=0.2576, simple_loss=0.3432, pruned_loss=0.08606, over 16871.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3063, pruned_loss=0.07334, over 3069428.08 frames. ], batch size: 116, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:04:32,812 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:04:41,590 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.520e+02 3.684e+02 4.316e+02 5.371e+02 1.144e+03, threshold=8.631e+02, percent-clipped=1.0 2023-04-29 15:05:08,756 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0454, 1.9614, 2.0827, 3.6521, 1.8973, 2.3894, 2.1092, 2.1260], device='cuda:5'), covar=tensor([0.1105, 0.3320, 0.2394, 0.0442, 0.3869, 0.2072, 0.2893, 0.3271], device='cuda:5'), in_proj_covar=tensor([0.0366, 0.0395, 0.0331, 0.0316, 0.0412, 0.0454, 0.0359, 0.0460], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:05:21,418 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:05:39,189 INFO [train.py:904] (5/8) Epoch 12, batch 5750, loss[loss=0.2685, simple_loss=0.3302, pruned_loss=0.1034, over 11249.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.309, pruned_loss=0.07545, over 3028347.40 frames. ], batch size: 248, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:05:48,655 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:07:00,143 INFO [train.py:904] (5/8) Epoch 12, batch 5800, loss[loss=0.202, simple_loss=0.2886, pruned_loss=0.0577, over 16849.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3098, pruned_loss=0.07496, over 3022005.67 frames. ], batch size: 116, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:07:09,861 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:07:21,329 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.077e+02 3.389e+02 3.943e+02 4.772e+02 8.236e+02, threshold=7.885e+02, percent-clipped=0.0 2023-04-29 15:08:16,568 INFO [train.py:904] (5/8) Epoch 12, batch 5850, loss[loss=0.2068, simple_loss=0.2979, pruned_loss=0.0579, over 16689.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3068, pruned_loss=0.07223, over 3049591.32 frames. ], batch size: 76, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:08:24,236 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:08:32,731 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4683, 2.3250, 2.8074, 3.1922, 3.0730, 3.9258, 2.2763, 3.6363], device='cuda:5'), covar=tensor([0.0119, 0.0322, 0.0195, 0.0184, 0.0192, 0.0067, 0.0356, 0.0078], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0169, 0.0153, 0.0157, 0.0167, 0.0124, 0.0170, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 15:09:37,163 INFO [train.py:904] (5/8) Epoch 12, batch 5900, loss[loss=0.2892, simple_loss=0.3421, pruned_loss=0.1182, over 11384.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3064, pruned_loss=0.0719, over 3059134.98 frames. ], batch size: 249, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:10:01,563 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.797e+02 2.838e+02 3.606e+02 4.204e+02 7.799e+02, threshold=7.213e+02, percent-clipped=0.0 2023-04-29 15:10:18,957 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:10:55,846 INFO [train.py:904] (5/8) Epoch 12, batch 5950, loss[loss=0.2158, simple_loss=0.3064, pruned_loss=0.06257, over 16312.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3075, pruned_loss=0.07064, over 3065953.84 frames. ], batch size: 35, lr: 5.67e-03, grad_scale: 2.0 2023-04-29 15:11:07,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2197, 2.4182, 2.0547, 2.2212, 2.8082, 2.4395, 2.9794, 3.0190], device='cuda:5'), covar=tensor([0.0080, 0.0306, 0.0407, 0.0341, 0.0192, 0.0300, 0.0175, 0.0177], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0202, 0.0198, 0.0196, 0.0201, 0.0200, 0.0207, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:12:14,118 INFO [train.py:904] (5/8) Epoch 12, batch 6000, loss[loss=0.2187, simple_loss=0.3002, pruned_loss=0.06859, over 15295.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.306, pruned_loss=0.06943, over 3086063.95 frames. ], batch size: 190, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:12:14,118 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 15:12:25,322 INFO [train.py:938] (5/8) Epoch 12, validation: loss=0.161, simple_loss=0.2739, pruned_loss=0.02405, over 944034.00 frames. 2023-04-29 15:12:25,323 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 15:12:46,505 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.036e+02 2.944e+02 3.518e+02 4.227e+02 7.444e+02, threshold=7.035e+02, percent-clipped=1.0 2023-04-29 15:13:18,119 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:13:46,111 INFO [train.py:904] (5/8) Epoch 12, batch 6050, loss[loss=0.2329, simple_loss=0.3252, pruned_loss=0.07032, over 16576.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3046, pruned_loss=0.06851, over 3105043.20 frames. ], batch size: 68, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:13:46,968 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7764, 4.8055, 4.6558, 3.9146, 4.6821, 1.5913, 4.4195, 4.4319], device='cuda:5'), covar=tensor([0.0103, 0.0095, 0.0156, 0.0380, 0.0095, 0.2529, 0.0152, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0119, 0.0163, 0.0157, 0.0136, 0.0179, 0.0152, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:14:23,516 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5310, 3.1237, 3.0608, 1.8551, 2.6685, 2.1306, 3.1789, 3.2677], device='cuda:5'), covar=tensor([0.0281, 0.0604, 0.0605, 0.1867, 0.0827, 0.0940, 0.0593, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0147, 0.0161, 0.0145, 0.0138, 0.0125, 0.0140, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 15:15:02,184 INFO [train.py:904] (5/8) Epoch 12, batch 6100, loss[loss=0.2267, simple_loss=0.2951, pruned_loss=0.07914, over 11730.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3038, pruned_loss=0.0672, over 3117138.14 frames. ], batch size: 248, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:15:24,824 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.974e+02 3.801e+02 4.781e+02 1.516e+03, threshold=7.602e+02, percent-clipped=11.0 2023-04-29 15:16:19,696 INFO [train.py:904] (5/8) Epoch 12, batch 6150, loss[loss=0.2047, simple_loss=0.2892, pruned_loss=0.06012, over 16552.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.3015, pruned_loss=0.06656, over 3120425.02 frames. ], batch size: 62, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:17:38,950 INFO [train.py:904] (5/8) Epoch 12, batch 6200, loss[loss=0.2244, simple_loss=0.3108, pruned_loss=0.06903, over 16321.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.3, pruned_loss=0.06629, over 3120228.07 frames. ], batch size: 146, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:18:00,668 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 3.055e+02 3.626e+02 4.280e+02 7.203e+02, threshold=7.253e+02, percent-clipped=0.0 2023-04-29 15:18:18,047 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:18:52,885 INFO [train.py:904] (5/8) Epoch 12, batch 6250, loss[loss=0.2084, simple_loss=0.296, pruned_loss=0.06043, over 16421.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2991, pruned_loss=0.06588, over 3117815.71 frames. ], batch size: 146, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:18:56,944 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-04-29 15:19:28,072 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:19:34,637 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1415, 4.2691, 4.5926, 4.5531, 4.5471, 4.3438, 4.3035, 4.1640], device='cuda:5'), covar=tensor([0.0347, 0.0566, 0.0380, 0.0421, 0.0472, 0.0396, 0.0871, 0.0494], device='cuda:5'), in_proj_covar=tensor([0.0347, 0.0359, 0.0361, 0.0343, 0.0412, 0.0383, 0.0488, 0.0309], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 15:20:06,479 INFO [train.py:904] (5/8) Epoch 12, batch 6300, loss[loss=0.2241, simple_loss=0.304, pruned_loss=0.07207, over 16882.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2992, pruned_loss=0.06557, over 3110684.98 frames. ], batch size: 109, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:20:28,835 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 3.047e+02 3.596e+02 4.345e+02 1.152e+03, threshold=7.193e+02, percent-clipped=4.0 2023-04-29 15:20:57,493 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:21:24,717 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-29 15:21:25,211 INFO [train.py:904] (5/8) Epoch 12, batch 6350, loss[loss=0.211, simple_loss=0.2927, pruned_loss=0.0647, over 16217.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2998, pruned_loss=0.06683, over 3120596.63 frames. ], batch size: 165, lr: 5.66e-03, grad_scale: 4.0 2023-04-29 15:22:11,816 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:21,819 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:36,897 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3494, 1.6339, 1.9605, 2.1933, 2.3176, 2.4587, 1.6909, 2.3600], device='cuda:5'), covar=tensor([0.0151, 0.0351, 0.0194, 0.0257, 0.0216, 0.0141, 0.0359, 0.0097], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0173, 0.0155, 0.0160, 0.0170, 0.0126, 0.0173, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 15:22:37,552 INFO [train.py:904] (5/8) Epoch 12, batch 6400, loss[loss=0.1907, simple_loss=0.2792, pruned_loss=0.05115, over 16291.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.301, pruned_loss=0.06855, over 3098215.74 frames. ], batch size: 165, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:22:57,665 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.235e+02 3.125e+02 4.036e+02 4.702e+02 8.359e+02, threshold=8.072e+02, percent-clipped=7.0 2023-04-29 15:23:50,436 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118101.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:23:50,966 INFO [train.py:904] (5/8) Epoch 12, batch 6450, loss[loss=0.2052, simple_loss=0.293, pruned_loss=0.05872, over 16917.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3012, pruned_loss=0.0682, over 3091141.93 frames. ], batch size: 116, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:23:53,386 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1702, 1.9675, 2.0717, 3.7833, 1.9259, 2.3852, 2.0676, 2.1538], device='cuda:5'), covar=tensor([0.1017, 0.3445, 0.2460, 0.0447, 0.3826, 0.2267, 0.3156, 0.3223], device='cuda:5'), in_proj_covar=tensor([0.0365, 0.0394, 0.0331, 0.0317, 0.0412, 0.0455, 0.0360, 0.0461], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:24:29,120 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4438, 4.5581, 4.7333, 4.5543, 4.6085, 5.1314, 4.6943, 4.4720], device='cuda:5'), covar=tensor([0.1312, 0.1716, 0.1811, 0.1966, 0.2259, 0.0990, 0.1462, 0.2283], device='cuda:5'), in_proj_covar=tensor([0.0358, 0.0497, 0.0546, 0.0429, 0.0582, 0.0569, 0.0432, 0.0586], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 15:25:08,004 INFO [train.py:904] (5/8) Epoch 12, batch 6500, loss[loss=0.2105, simple_loss=0.2954, pruned_loss=0.06283, over 16728.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2989, pruned_loss=0.06728, over 3105007.78 frames. ], batch size: 134, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:25:29,349 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 2.874e+02 3.539e+02 4.309e+02 7.479e+02, threshold=7.078e+02, percent-clipped=0.0 2023-04-29 15:25:32,621 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:26:28,551 INFO [train.py:904] (5/8) Epoch 12, batch 6550, loss[loss=0.2866, simple_loss=0.345, pruned_loss=0.1141, over 11733.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3018, pruned_loss=0.06898, over 3082367.93 frames. ], batch size: 247, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:27:10,771 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 15:27:44,400 INFO [train.py:904] (5/8) Epoch 12, batch 6600, loss[loss=0.2164, simple_loss=0.3026, pruned_loss=0.06513, over 15466.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3049, pruned_loss=0.0698, over 3097440.32 frames. ], batch size: 191, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:28:05,473 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.686e+02 3.186e+02 3.899e+02 4.864e+02 8.341e+02, threshold=7.798e+02, percent-clipped=4.0 2023-04-29 15:28:10,208 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:28:37,391 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4375, 3.0224, 2.8658, 1.7712, 2.6130, 2.0951, 2.9354, 3.2263], device='cuda:5'), covar=tensor([0.0306, 0.0607, 0.0680, 0.1881, 0.0833, 0.0966, 0.0729, 0.0737], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0146, 0.0160, 0.0144, 0.0138, 0.0125, 0.0138, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 15:28:42,122 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:01,042 INFO [train.py:904] (5/8) Epoch 12, batch 6650, loss[loss=0.2602, simple_loss=0.3134, pruned_loss=0.1035, over 11140.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.305, pruned_loss=0.07055, over 3085424.22 frames. ], batch size: 246, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:29:44,402 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:45,598 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:30:16,331 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:30:18,844 INFO [train.py:904] (5/8) Epoch 12, batch 6700, loss[loss=0.2189, simple_loss=0.3006, pruned_loss=0.06858, over 16890.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3045, pruned_loss=0.07113, over 3072210.58 frames. ], batch size: 109, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:30:33,871 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6164, 2.4356, 2.2605, 3.1831, 2.3446, 3.5384, 1.3855, 2.7280], device='cuda:5'), covar=tensor([0.1398, 0.0717, 0.1255, 0.0159, 0.0186, 0.0403, 0.1691, 0.0787], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0162, 0.0183, 0.0151, 0.0198, 0.0208, 0.0183, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 15:30:39,921 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 3.120e+02 3.560e+02 4.149e+02 6.786e+02, threshold=7.119e+02, percent-clipped=0.0 2023-04-29 15:31:17,636 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:31:27,359 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:31:35,216 INFO [train.py:904] (5/8) Epoch 12, batch 6750, loss[loss=0.2616, simple_loss=0.3345, pruned_loss=0.09439, over 12047.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3036, pruned_loss=0.0713, over 3059054.33 frames. ], batch size: 248, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:31:58,010 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4007, 2.9594, 2.9180, 1.9457, 2.5782, 2.0926, 3.0401, 3.1241], device='cuda:5'), covar=tensor([0.0298, 0.0638, 0.0656, 0.1848, 0.0840, 0.0993, 0.0676, 0.0845], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0147, 0.0161, 0.0145, 0.0139, 0.0126, 0.0139, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 15:32:49,915 INFO [train.py:904] (5/8) Epoch 12, batch 6800, loss[loss=0.2541, simple_loss=0.3357, pruned_loss=0.08624, over 16238.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3026, pruned_loss=0.07005, over 3076380.83 frames. ], batch size: 165, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:33:11,649 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 3.071e+02 3.777e+02 4.759e+02 7.416e+02, threshold=7.554e+02, percent-clipped=1.0 2023-04-29 15:33:15,257 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6788, 2.0584, 1.5444, 1.8483, 2.4600, 2.0782, 2.5296, 2.7124], device='cuda:5'), covar=tensor([0.0149, 0.0375, 0.0545, 0.0446, 0.0218, 0.0351, 0.0198, 0.0213], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0204, 0.0201, 0.0200, 0.0205, 0.0203, 0.0210, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:33:27,425 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9023, 5.1845, 4.9107, 4.9000, 4.6770, 4.6130, 4.6162, 5.2315], device='cuda:5'), covar=tensor([0.0971, 0.0742, 0.0908, 0.0693, 0.0710, 0.0877, 0.1032, 0.0856], device='cuda:5'), in_proj_covar=tensor([0.0557, 0.0691, 0.0570, 0.0489, 0.0439, 0.0454, 0.0573, 0.0539], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:34:04,799 INFO [train.py:904] (5/8) Epoch 12, batch 6850, loss[loss=0.2294, simple_loss=0.3231, pruned_loss=0.06785, over 16872.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3035, pruned_loss=0.07081, over 3055732.28 frames. ], batch size: 109, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:34:05,726 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 15:34:16,578 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0071, 1.9958, 2.0921, 3.5884, 2.0003, 2.3615, 2.1149, 2.1363], device='cuda:5'), covar=tensor([0.1200, 0.3356, 0.2430, 0.0498, 0.3873, 0.2248, 0.3156, 0.3123], device='cuda:5'), in_proj_covar=tensor([0.0364, 0.0393, 0.0331, 0.0316, 0.0412, 0.0454, 0.0360, 0.0461], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:34:36,249 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 15:35:15,881 INFO [train.py:904] (5/8) Epoch 12, batch 6900, loss[loss=0.2168, simple_loss=0.3025, pruned_loss=0.06552, over 16678.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3057, pruned_loss=0.07032, over 3057871.19 frames. ], batch size: 62, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:35:36,847 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 3.007e+02 3.634e+02 4.757e+02 1.105e+03, threshold=7.268e+02, percent-clipped=1.0 2023-04-29 15:36:30,544 INFO [train.py:904] (5/8) Epoch 12, batch 6950, loss[loss=0.2339, simple_loss=0.3145, pruned_loss=0.0767, over 15404.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3083, pruned_loss=0.07228, over 3057109.12 frames. ], batch size: 191, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:37:04,643 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:14,746 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:35,449 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:44,792 INFO [train.py:904] (5/8) Epoch 12, batch 7000, loss[loss=0.2267, simple_loss=0.3153, pruned_loss=0.06904, over 15334.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3077, pruned_loss=0.0709, over 3070876.98 frames. ], batch size: 190, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:38:05,445 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 2.982e+02 3.795e+02 4.664e+02 8.323e+02, threshold=7.591e+02, percent-clipped=2.0 2023-04-29 15:38:33,857 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:44,685 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:50,456 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:50,617 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5531, 2.3436, 2.4586, 4.2673, 2.1804, 2.7209, 2.3738, 2.5360], device='cuda:5'), covar=tensor([0.0914, 0.3084, 0.2202, 0.0379, 0.3633, 0.2157, 0.2981, 0.2873], device='cuda:5'), in_proj_covar=tensor([0.0362, 0.0391, 0.0328, 0.0313, 0.0409, 0.0450, 0.0359, 0.0457], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:38:59,534 INFO [train.py:904] (5/8) Epoch 12, batch 7050, loss[loss=0.2051, simple_loss=0.2942, pruned_loss=0.05799, over 16441.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3084, pruned_loss=0.07053, over 3074309.05 frames. ], batch size: 68, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:40:01,792 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:40:14,611 INFO [train.py:904] (5/8) Epoch 12, batch 7100, loss[loss=0.1994, simple_loss=0.2866, pruned_loss=0.05608, over 17249.00 frames. ], tot_loss[loss=0.224, simple_loss=0.307, pruned_loss=0.07048, over 3071826.18 frames. ], batch size: 52, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:40:36,862 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 3.120e+02 3.747e+02 4.649e+02 1.761e+03, threshold=7.495e+02, percent-clipped=5.0 2023-04-29 15:40:42,824 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8407, 3.3998, 3.3046, 2.0901, 3.1431, 3.3549, 3.1531, 1.7685], device='cuda:5'), covar=tensor([0.0499, 0.0034, 0.0038, 0.0365, 0.0070, 0.0095, 0.0068, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0068, 0.0071, 0.0125, 0.0080, 0.0092, 0.0080, 0.0118], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 15:41:29,296 INFO [train.py:904] (5/8) Epoch 12, batch 7150, loss[loss=0.2257, simple_loss=0.3276, pruned_loss=0.06191, over 16952.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3052, pruned_loss=0.06979, over 3082530.76 frames. ], batch size: 109, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:41:32,182 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3603, 3.3497, 3.3884, 3.4963, 3.5168, 3.2538, 3.5017, 3.5571], device='cuda:5'), covar=tensor([0.1018, 0.0804, 0.0982, 0.0538, 0.0562, 0.2053, 0.0865, 0.0675], device='cuda:5'), in_proj_covar=tensor([0.0518, 0.0648, 0.0778, 0.0660, 0.0496, 0.0509, 0.0524, 0.0596], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:42:01,983 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:42:41,521 INFO [train.py:904] (5/8) Epoch 12, batch 7200, loss[loss=0.175, simple_loss=0.2664, pruned_loss=0.04183, over 16539.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3024, pruned_loss=0.06785, over 3075992.83 frames. ], batch size: 68, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:43:03,915 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.566e+02 3.185e+02 3.996e+02 6.189e+02, threshold=6.369e+02, percent-clipped=0.0 2023-04-29 15:43:12,503 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:43:22,005 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 15:44:00,070 INFO [train.py:904] (5/8) Epoch 12, batch 7250, loss[loss=0.199, simple_loss=0.278, pruned_loss=0.05997, over 16728.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2996, pruned_loss=0.0664, over 3071304.59 frames. ], batch size: 124, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:44:30,184 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0653, 4.2698, 4.7181, 2.4325, 4.9995, 4.9847, 3.4463, 3.6059], device='cuda:5'), covar=tensor([0.0711, 0.0182, 0.0134, 0.1060, 0.0033, 0.0078, 0.0334, 0.0420], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0100, 0.0087, 0.0137, 0.0068, 0.0104, 0.0121, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 15:44:35,493 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:04,527 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:15,120 INFO [train.py:904] (5/8) Epoch 12, batch 7300, loss[loss=0.2165, simple_loss=0.3098, pruned_loss=0.06159, over 16609.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2983, pruned_loss=0.06571, over 3085988.65 frames. ], batch size: 62, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:45:26,339 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1687, 1.3977, 1.8346, 1.9628, 2.1101, 2.3707, 1.6664, 2.2274], device='cuda:5'), covar=tensor([0.0159, 0.0372, 0.0203, 0.0249, 0.0239, 0.0134, 0.0331, 0.0091], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0171, 0.0154, 0.0157, 0.0170, 0.0124, 0.0170, 0.0117], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 15:45:36,404 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 3.034e+02 3.599e+02 4.380e+02 7.583e+02, threshold=7.199e+02, percent-clipped=5.0 2023-04-29 15:45:45,782 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:47,881 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7735, 2.6501, 2.6289, 4.6615, 3.7222, 4.1366, 1.5565, 3.0502], device='cuda:5'), covar=tensor([0.1275, 0.0747, 0.1117, 0.0109, 0.0326, 0.0323, 0.1542, 0.0763], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0162, 0.0183, 0.0151, 0.0200, 0.0209, 0.0184, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 15:45:51,953 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 15:46:03,491 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:04,904 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7245, 2.3986, 2.2777, 3.2626, 2.4546, 3.5531, 1.4636, 2.6698], device='cuda:5'), covar=tensor([0.1286, 0.0702, 0.1198, 0.0157, 0.0267, 0.0395, 0.1593, 0.0855], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0183, 0.0151, 0.0200, 0.0209, 0.0184, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 15:46:06,457 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:15,623 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:28,524 INFO [train.py:904] (5/8) Epoch 12, batch 7350, loss[loss=0.2017, simple_loss=0.2902, pruned_loss=0.05657, over 16866.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.299, pruned_loss=0.06634, over 3087203.05 frames. ], batch size: 116, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:47:14,447 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:47:42,949 INFO [train.py:904] (5/8) Epoch 12, batch 7400, loss[loss=0.2105, simple_loss=0.2986, pruned_loss=0.06125, over 17228.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3007, pruned_loss=0.06748, over 3083929.87 frames. ], batch size: 45, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:48:06,309 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 3.304e+02 4.015e+02 4.796e+02 1.422e+03, threshold=8.030e+02, percent-clipped=7.0 2023-04-29 15:48:17,981 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3512, 3.3749, 1.9229, 3.7717, 2.4546, 3.7247, 2.0077, 2.5784], device='cuda:5'), covar=tensor([0.0255, 0.0404, 0.1660, 0.0154, 0.0897, 0.0567, 0.1691, 0.0868], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0165, 0.0190, 0.0130, 0.0168, 0.0205, 0.0197, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 15:48:19,167 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:48:47,080 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:49:01,221 INFO [train.py:904] (5/8) Epoch 12, batch 7450, loss[loss=0.2325, simple_loss=0.3153, pruned_loss=0.07485, over 16929.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3021, pruned_loss=0.06847, over 3090927.87 frames. ], batch size: 109, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:49:55,919 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:50:06,329 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 15:50:20,478 INFO [train.py:904] (5/8) Epoch 12, batch 7500, loss[loss=0.2645, simple_loss=0.33, pruned_loss=0.09948, over 11168.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3028, pruned_loss=0.06862, over 3058117.11 frames. ], batch size: 247, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:50:24,061 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 15:50:42,267 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.952e+02 3.435e+02 4.438e+02 7.679e+02, threshold=6.870e+02, percent-clipped=0.0 2023-04-29 15:51:35,631 INFO [train.py:904] (5/8) Epoch 12, batch 7550, loss[loss=0.2118, simple_loss=0.2961, pruned_loss=0.06371, over 16291.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3023, pruned_loss=0.06934, over 3048192.97 frames. ], batch size: 165, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:52:50,120 INFO [train.py:904] (5/8) Epoch 12, batch 7600, loss[loss=0.2482, simple_loss=0.3115, pruned_loss=0.09244, over 11143.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3012, pruned_loss=0.0691, over 3058271.07 frames. ], batch size: 247, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:53:12,405 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.101e+02 2.984e+02 3.563e+02 4.291e+02 1.017e+03, threshold=7.126e+02, percent-clipped=2.0 2023-04-29 15:53:28,022 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5241, 4.6272, 4.7930, 4.6584, 4.6260, 5.1516, 4.7203, 4.4831], device='cuda:5'), covar=tensor([0.1211, 0.1769, 0.2091, 0.1834, 0.2412, 0.1002, 0.1499, 0.2430], device='cuda:5'), in_proj_covar=tensor([0.0359, 0.0499, 0.0546, 0.0434, 0.0582, 0.0569, 0.0431, 0.0586], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 15:53:30,689 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9665, 2.2940, 1.6256, 1.9585, 2.6308, 2.3368, 2.8835, 2.8732], device='cuda:5'), covar=tensor([0.0123, 0.0363, 0.0596, 0.0461, 0.0229, 0.0346, 0.0248, 0.0219], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0205, 0.0201, 0.0200, 0.0206, 0.0202, 0.0208, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 15:53:43,933 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:54:05,873 INFO [train.py:904] (5/8) Epoch 12, batch 7650, loss[loss=0.2484, simple_loss=0.3143, pruned_loss=0.09125, over 11008.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3019, pruned_loss=0.07024, over 3043186.87 frames. ], batch size: 246, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:54:55,657 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:55:20,166 INFO [train.py:904] (5/8) Epoch 12, batch 7700, loss[loss=0.2247, simple_loss=0.3226, pruned_loss=0.0634, over 16798.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3014, pruned_loss=0.06955, over 3079088.72 frames. ], batch size: 102, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:55:42,614 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.234e+02 3.396e+02 4.418e+02 5.495e+02 1.012e+03, threshold=8.835e+02, percent-clipped=5.0 2023-04-29 15:56:36,114 INFO [train.py:904] (5/8) Epoch 12, batch 7750, loss[loss=0.2539, simple_loss=0.3165, pruned_loss=0.09562, over 11722.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3012, pruned_loss=0.06882, over 3106893.23 frames. ], batch size: 248, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:56:37,491 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-04-29 15:57:09,858 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0136, 3.0727, 1.8135, 3.2560, 2.3616, 3.2985, 2.0611, 2.5452], device='cuda:5'), covar=tensor([0.0257, 0.0350, 0.1577, 0.0171, 0.0737, 0.0525, 0.1367, 0.0672], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0164, 0.0190, 0.0130, 0.0168, 0.0205, 0.0197, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 15:57:18,801 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:57:45,531 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 15:57:48,712 INFO [train.py:904] (5/8) Epoch 12, batch 7800, loss[loss=0.2098, simple_loss=0.2924, pruned_loss=0.06353, over 15341.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3025, pruned_loss=0.0696, over 3111881.68 frames. ], batch size: 190, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:58:11,185 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 3.141e+02 3.879e+02 4.510e+02 7.221e+02, threshold=7.757e+02, percent-clipped=0.0 2023-04-29 15:59:04,890 INFO [train.py:904] (5/8) Epoch 12, batch 7850, loss[loss=0.2229, simple_loss=0.3009, pruned_loss=0.07242, over 15307.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3032, pruned_loss=0.06921, over 3108197.50 frames. ], batch size: 191, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 15:59:24,417 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:59:46,618 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1018, 5.4451, 5.1460, 5.1782, 4.8903, 4.8176, 4.8231, 5.5475], device='cuda:5'), covar=tensor([0.1006, 0.0818, 0.0995, 0.0782, 0.0728, 0.0827, 0.1029, 0.0770], device='cuda:5'), in_proj_covar=tensor([0.0550, 0.0680, 0.0567, 0.0483, 0.0435, 0.0450, 0.0571, 0.0532], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:00:21,535 INFO [train.py:904] (5/8) Epoch 12, batch 7900, loss[loss=0.2521, simple_loss=0.3072, pruned_loss=0.09848, over 11663.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3016, pruned_loss=0.06847, over 3121654.37 frames. ], batch size: 247, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:00:45,732 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 2.838e+02 3.537e+02 4.302e+02 9.445e+02, threshold=7.074e+02, percent-clipped=1.0 2023-04-29 16:00:56,382 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:01:38,586 INFO [train.py:904] (5/8) Epoch 12, batch 7950, loss[loss=0.2745, simple_loss=0.3368, pruned_loss=0.1061, over 11362.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3023, pruned_loss=0.0687, over 3127990.16 frames. ], batch size: 246, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:02:39,408 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1254, 3.2762, 3.3410, 2.1481, 3.1269, 3.3383, 3.2765, 1.9615], device='cuda:5'), covar=tensor([0.0440, 0.0045, 0.0042, 0.0356, 0.0073, 0.0091, 0.0060, 0.0371], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0069, 0.0071, 0.0127, 0.0080, 0.0092, 0.0081, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 16:02:53,353 INFO [train.py:904] (5/8) Epoch 12, batch 8000, loss[loss=0.2107, simple_loss=0.2952, pruned_loss=0.06311, over 16652.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3033, pruned_loss=0.0704, over 3093496.61 frames. ], batch size: 57, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 16:02:58,634 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6137, 4.7002, 4.4755, 4.2133, 4.1073, 4.5611, 4.4846, 4.2158], device='cuda:5'), covar=tensor([0.0629, 0.0479, 0.0281, 0.0290, 0.0995, 0.0432, 0.0347, 0.0660], device='cuda:5'), in_proj_covar=tensor([0.0240, 0.0324, 0.0285, 0.0264, 0.0303, 0.0307, 0.0197, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:03:12,796 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:03:17,121 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 2.898e+02 3.526e+02 5.291e+02 1.062e+03, threshold=7.052e+02, percent-clipped=9.0 2023-04-29 16:04:07,964 INFO [train.py:904] (5/8) Epoch 12, batch 8050, loss[loss=0.2651, simple_loss=0.3241, pruned_loss=0.103, over 11817.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.303, pruned_loss=0.07, over 3078718.25 frames. ], batch size: 247, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:04:26,763 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4580, 4.3381, 4.5297, 4.7077, 4.8122, 4.3916, 4.8369, 4.7957], device='cuda:5'), covar=tensor([0.1700, 0.1052, 0.1431, 0.0571, 0.0526, 0.0867, 0.0487, 0.0570], device='cuda:5'), in_proj_covar=tensor([0.0523, 0.0655, 0.0785, 0.0665, 0.0505, 0.0513, 0.0535, 0.0607], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:04:42,938 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119726.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:04:49,757 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:05:17,377 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 16:05:21,196 INFO [train.py:904] (5/8) Epoch 12, batch 8100, loss[loss=0.2482, simple_loss=0.3245, pruned_loss=0.08598, over 15305.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3026, pruned_loss=0.06909, over 3088755.04 frames. ], batch size: 190, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:05:47,745 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.274e+02 3.055e+02 3.528e+02 4.290e+02 7.598e+02, threshold=7.056e+02, percent-clipped=1.0 2023-04-29 16:05:48,809 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 16:05:52,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5025, 3.5026, 3.4554, 2.8207, 3.3252, 2.1121, 3.2039, 2.8102], device='cuda:5'), covar=tensor([0.0126, 0.0098, 0.0144, 0.0195, 0.0084, 0.1956, 0.0110, 0.0191], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0119, 0.0165, 0.0156, 0.0137, 0.0181, 0.0152, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:06:01,242 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:28,074 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:34,776 INFO [train.py:904] (5/8) Epoch 12, batch 8150, loss[loss=0.1913, simple_loss=0.2752, pruned_loss=0.05373, over 16340.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2995, pruned_loss=0.06757, over 3093405.73 frames. ], batch size: 165, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:06:57,741 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6443, 5.9571, 5.6105, 5.6861, 5.2852, 5.0788, 5.3672, 6.0496], device='cuda:5'), covar=tensor([0.0972, 0.0699, 0.0891, 0.0629, 0.0798, 0.0680, 0.1029, 0.0763], device='cuda:5'), in_proj_covar=tensor([0.0544, 0.0672, 0.0559, 0.0478, 0.0429, 0.0447, 0.0566, 0.0523], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:07:50,609 INFO [train.py:904] (5/8) Epoch 12, batch 8200, loss[loss=0.2207, simple_loss=0.3098, pruned_loss=0.06584, over 16212.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2968, pruned_loss=0.06672, over 3090367.43 frames. ], batch size: 165, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:08:18,227 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 3.028e+02 3.797e+02 4.737e+02 9.915e+02, threshold=7.593e+02, percent-clipped=4.0 2023-04-29 16:08:19,319 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:08:36,980 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 16:08:38,900 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 16:09:09,087 INFO [train.py:904] (5/8) Epoch 12, batch 8250, loss[loss=0.1858, simple_loss=0.2749, pruned_loss=0.04834, over 16543.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2954, pruned_loss=0.06455, over 3069000.84 frames. ], batch size: 57, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:09:20,730 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0062, 1.8033, 1.6310, 1.5627, 1.9280, 1.6461, 1.7176, 1.9993], device='cuda:5'), covar=tensor([0.0133, 0.0220, 0.0308, 0.0279, 0.0176, 0.0226, 0.0143, 0.0171], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0201, 0.0197, 0.0196, 0.0202, 0.0199, 0.0204, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:10:28,032 INFO [train.py:904] (5/8) Epoch 12, batch 8300, loss[loss=0.1842, simple_loss=0.2718, pruned_loss=0.04831, over 16293.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2921, pruned_loss=0.06078, over 3080431.58 frames. ], batch size: 35, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:10:47,018 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9306, 4.2370, 4.0083, 4.0682, 3.6586, 3.8098, 3.8438, 4.2065], device='cuda:5'), covar=tensor([0.1059, 0.0968, 0.0977, 0.0742, 0.0976, 0.1652, 0.0963, 0.1089], device='cuda:5'), in_proj_covar=tensor([0.0539, 0.0666, 0.0554, 0.0474, 0.0424, 0.0443, 0.0559, 0.0518], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:10:57,564 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.352e+02 2.901e+02 3.420e+02 5.863e+02, threshold=5.801e+02, percent-clipped=0.0 2023-04-29 16:11:17,378 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5603, 3.2790, 2.7023, 2.0143, 2.1390, 2.1588, 3.4148, 3.1117], device='cuda:5'), covar=tensor([0.2555, 0.0709, 0.1689, 0.2855, 0.2574, 0.2058, 0.0488, 0.1162], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0253, 0.0283, 0.0278, 0.0277, 0.0220, 0.0265, 0.0293], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:11:52,967 INFO [train.py:904] (5/8) Epoch 12, batch 8350, loss[loss=0.1913, simple_loss=0.2886, pruned_loss=0.04698, over 16277.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2909, pruned_loss=0.0588, over 3064100.91 frames. ], batch size: 146, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:12:24,539 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120021.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:13:14,452 INFO [train.py:904] (5/8) Epoch 12, batch 8400, loss[loss=0.1823, simple_loss=0.2754, pruned_loss=0.04454, over 16740.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2883, pruned_loss=0.05627, over 3081528.55 frames. ], batch size: 124, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:13:25,003 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 16:13:42,958 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.479e+02 2.965e+02 3.428e+02 6.852e+02, threshold=5.930e+02, percent-clipped=2.0 2023-04-29 16:14:31,509 INFO [train.py:904] (5/8) Epoch 12, batch 8450, loss[loss=0.1838, simple_loss=0.2807, pruned_loss=0.04351, over 16288.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2865, pruned_loss=0.05451, over 3084036.88 frames. ], batch size: 165, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:14:32,487 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 16:14:52,201 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5194, 3.2906, 2.6837, 1.9987, 2.1515, 2.1810, 3.4310, 2.9686], device='cuda:5'), covar=tensor([0.2499, 0.0703, 0.1502, 0.2691, 0.2714, 0.2029, 0.0412, 0.1240], device='cuda:5'), in_proj_covar=tensor([0.0296, 0.0248, 0.0276, 0.0272, 0.0270, 0.0216, 0.0259, 0.0287], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:15:47,458 INFO [train.py:904] (5/8) Epoch 12, batch 8500, loss[loss=0.1674, simple_loss=0.254, pruned_loss=0.04043, over 15262.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2826, pruned_loss=0.05218, over 3078467.23 frames. ], batch size: 190, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:16:15,430 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.373e+02 2.767e+02 3.491e+02 7.213e+02, threshold=5.534e+02, percent-clipped=3.0 2023-04-29 16:16:15,840 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:16:34,816 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0981, 1.3923, 1.8204, 2.0775, 2.1553, 2.2337, 1.6580, 2.3003], device='cuda:5'), covar=tensor([0.0201, 0.0364, 0.0223, 0.0226, 0.0223, 0.0173, 0.0347, 0.0106], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0171, 0.0155, 0.0156, 0.0169, 0.0123, 0.0169, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 16:17:07,733 INFO [train.py:904] (5/8) Epoch 12, batch 8550, loss[loss=0.1995, simple_loss=0.2885, pruned_loss=0.05525, over 15347.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2796, pruned_loss=0.05077, over 3048831.25 frames. ], batch size: 191, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:17:37,327 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:18:37,663 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-04-29 16:18:41,895 INFO [train.py:904] (5/8) Epoch 12, batch 8600, loss[loss=0.2048, simple_loss=0.2931, pruned_loss=0.05825, over 16948.00 frames. ], tot_loss[loss=0.19, simple_loss=0.28, pruned_loss=0.05002, over 3041522.49 frames. ], batch size: 109, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:18:45,455 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 16:19:19,268 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 16:19:19,446 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.485e+02 2.557e+02 3.073e+02 4.030e+02 6.956e+02, threshold=6.147e+02, percent-clipped=5.0 2023-04-29 16:20:19,210 INFO [train.py:904] (5/8) Epoch 12, batch 8650, loss[loss=0.1737, simple_loss=0.2621, pruned_loss=0.04267, over 12231.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2781, pruned_loss=0.04853, over 3049433.95 frames. ], batch size: 249, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:21:01,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6436, 4.9768, 4.6860, 4.7285, 4.4598, 4.4095, 4.4013, 4.9957], device='cuda:5'), covar=tensor([0.1042, 0.0808, 0.1046, 0.0660, 0.0682, 0.1126, 0.1038, 0.0908], device='cuda:5'), in_proj_covar=tensor([0.0533, 0.0655, 0.0543, 0.0466, 0.0419, 0.0437, 0.0548, 0.0511], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:21:01,243 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:21:34,705 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8343, 4.6040, 4.8150, 5.0120, 5.1850, 4.6081, 5.1704, 5.1473], device='cuda:5'), covar=tensor([0.1483, 0.1105, 0.1543, 0.0662, 0.0484, 0.0897, 0.0511, 0.0629], device='cuda:5'), in_proj_covar=tensor([0.0510, 0.0641, 0.0770, 0.0653, 0.0494, 0.0506, 0.0524, 0.0595], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:22:02,145 INFO [train.py:904] (5/8) Epoch 12, batch 8700, loss[loss=0.1887, simple_loss=0.2769, pruned_loss=0.05031, over 16898.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2756, pruned_loss=0.04727, over 3066681.63 frames. ], batch size: 116, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:22:28,876 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6093, 3.4474, 3.6390, 1.7474, 3.8278, 3.9376, 3.0393, 3.0401], device='cuda:5'), covar=tensor([0.0614, 0.0207, 0.0166, 0.1147, 0.0053, 0.0095, 0.0340, 0.0363], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0095, 0.0082, 0.0130, 0.0065, 0.0099, 0.0115, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 16:22:33,633 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=120369.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:22:34,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 2.303e+02 2.810e+02 3.260e+02 5.191e+02, threshold=5.620e+02, percent-clipped=0.0 2023-04-29 16:23:31,703 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4884, 1.9637, 2.0926, 4.0036, 1.9166, 2.2510, 2.0983, 2.1021], device='cuda:5'), covar=tensor([0.1012, 0.4239, 0.2657, 0.0509, 0.5114, 0.2904, 0.3590, 0.4160], device='cuda:5'), in_proj_covar=tensor([0.0355, 0.0387, 0.0325, 0.0309, 0.0407, 0.0440, 0.0351, 0.0447], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:23:36,539 INFO [train.py:904] (5/8) Epoch 12, batch 8750, loss[loss=0.1934, simple_loss=0.2929, pruned_loss=0.04699, over 16430.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2755, pruned_loss=0.0469, over 3067574.34 frames. ], batch size: 146, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:23:40,835 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7097, 3.2735, 3.2736, 1.7699, 2.7896, 2.2239, 3.2234, 3.4004], device='cuda:5'), covar=tensor([0.0302, 0.0732, 0.0524, 0.2025, 0.0820, 0.0971, 0.0699, 0.0879], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0141, 0.0155, 0.0141, 0.0133, 0.0123, 0.0134, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 16:24:49,470 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0138, 2.6882, 2.8386, 1.9844, 2.5969, 2.1253, 2.7234, 2.8434], device='cuda:5'), covar=tensor([0.0284, 0.0838, 0.0519, 0.1767, 0.0800, 0.0970, 0.0595, 0.0712], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0140, 0.0154, 0.0141, 0.0133, 0.0123, 0.0133, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 16:25:30,729 INFO [train.py:904] (5/8) Epoch 12, batch 8800, loss[loss=0.1787, simple_loss=0.275, pruned_loss=0.04125, over 16837.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2733, pruned_loss=0.04541, over 3073510.55 frames. ], batch size: 90, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:26:08,429 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.538e+02 3.135e+02 3.629e+02 6.620e+02, threshold=6.271e+02, percent-clipped=2.0 2023-04-29 16:27:17,120 INFO [train.py:904] (5/8) Epoch 12, batch 8850, loss[loss=0.1744, simple_loss=0.2636, pruned_loss=0.04267, over 12136.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2755, pruned_loss=0.0452, over 3040678.65 frames. ], batch size: 250, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:27:34,409 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2997, 3.3395, 1.7833, 3.7585, 2.3197, 3.6644, 1.9961, 2.6955], device='cuda:5'), covar=tensor([0.0285, 0.0375, 0.1804, 0.0148, 0.1059, 0.0452, 0.1787, 0.0776], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0157, 0.0183, 0.0125, 0.0164, 0.0197, 0.0193, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 16:29:04,479 INFO [train.py:904] (5/8) Epoch 12, batch 8900, loss[loss=0.1723, simple_loss=0.2694, pruned_loss=0.03757, over 16764.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2764, pruned_loss=0.04447, over 3068803.95 frames. ], batch size: 76, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:29:39,306 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.788e+02 2.488e+02 2.995e+02 3.537e+02 1.098e+03, threshold=5.991e+02, percent-clipped=1.0 2023-04-29 16:29:46,628 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0309, 1.7876, 1.6537, 1.4687, 1.9432, 1.6094, 1.7668, 1.9883], device='cuda:5'), covar=tensor([0.0109, 0.0208, 0.0284, 0.0286, 0.0140, 0.0209, 0.0127, 0.0150], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0204, 0.0199, 0.0198, 0.0203, 0.0201, 0.0202, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:31:10,998 INFO [train.py:904] (5/8) Epoch 12, batch 8950, loss[loss=0.1946, simple_loss=0.2723, pruned_loss=0.05841, over 12253.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2767, pruned_loss=0.04537, over 3048895.89 frames. ], batch size: 247, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:31:17,112 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 16:31:30,379 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5153, 1.6360, 2.0029, 2.5355, 2.4801, 2.8373, 1.7849, 2.6936], device='cuda:5'), covar=tensor([0.0151, 0.0401, 0.0287, 0.0199, 0.0226, 0.0121, 0.0365, 0.0107], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0170, 0.0155, 0.0155, 0.0167, 0.0121, 0.0168, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 16:33:00,375 INFO [train.py:904] (5/8) Epoch 12, batch 9000, loss[loss=0.169, simple_loss=0.2608, pruned_loss=0.03862, over 16218.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2732, pruned_loss=0.04371, over 3066499.36 frames. ], batch size: 165, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:33:00,375 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 16:33:10,347 INFO [train.py:938] (5/8) Epoch 12, validation: loss=0.1532, simple_loss=0.2571, pruned_loss=0.02465, over 944034.00 frames. 2023-04-29 16:33:10,347 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 16:33:49,114 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.337e+02 2.622e+02 3.241e+02 7.245e+02, threshold=5.244e+02, percent-clipped=4.0 2023-04-29 16:33:56,473 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-04-29 16:33:58,147 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7647, 3.8041, 4.0964, 4.1109, 4.1092, 3.8945, 3.8994, 3.8926], device='cuda:5'), covar=tensor([0.0326, 0.0752, 0.0475, 0.0425, 0.0420, 0.0402, 0.0691, 0.0355], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0340, 0.0336, 0.0320, 0.0385, 0.0362, 0.0448, 0.0291], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 16:34:01,236 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:34:54,277 INFO [train.py:904] (5/8) Epoch 12, batch 9050, loss[loss=0.1752, simple_loss=0.2641, pruned_loss=0.04317, over 16351.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2748, pruned_loss=0.04437, over 3066001.98 frames. ], batch size: 146, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:35:22,050 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9188, 2.3692, 2.3059, 3.0871, 2.0202, 3.2960, 1.6475, 2.7664], device='cuda:5'), covar=tensor([0.1394, 0.0628, 0.1036, 0.0147, 0.0080, 0.0363, 0.1644, 0.0693], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0160, 0.0180, 0.0145, 0.0190, 0.0205, 0.0182, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 16:36:04,478 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:36:39,417 INFO [train.py:904] (5/8) Epoch 12, batch 9100, loss[loss=0.173, simple_loss=0.2596, pruned_loss=0.0432, over 12432.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2739, pruned_loss=0.04488, over 3047273.68 frames. ], batch size: 248, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:36:51,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7547, 3.9473, 2.1319, 4.3918, 2.8582, 4.3080, 2.3825, 3.1913], device='cuda:5'), covar=tensor([0.0212, 0.0230, 0.1482, 0.0112, 0.0708, 0.0349, 0.1440, 0.0527], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0156, 0.0182, 0.0125, 0.0162, 0.0195, 0.0192, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 16:37:15,456 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.513e+02 2.910e+02 3.503e+02 5.095e+02, threshold=5.819e+02, percent-clipped=0.0 2023-04-29 16:37:39,837 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 16:38:36,950 INFO [train.py:904] (5/8) Epoch 12, batch 9150, loss[loss=0.1759, simple_loss=0.2653, pruned_loss=0.04331, over 12239.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2742, pruned_loss=0.04435, over 3047976.20 frames. ], batch size: 250, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:40:21,521 INFO [train.py:904] (5/8) Epoch 12, batch 9200, loss[loss=0.1662, simple_loss=0.2581, pruned_loss=0.03712, over 16793.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.27, pruned_loss=0.04334, over 3055848.73 frames. ], batch size: 124, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:40:28,414 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3092, 4.3509, 4.1667, 3.8749, 3.8985, 4.2789, 4.0235, 3.9378], device='cuda:5'), covar=tensor([0.0539, 0.0509, 0.0271, 0.0263, 0.0749, 0.0417, 0.0582, 0.0613], device='cuda:5'), in_proj_covar=tensor([0.0232, 0.0310, 0.0274, 0.0254, 0.0289, 0.0293, 0.0189, 0.0320], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-29 16:40:55,546 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.400e+02 2.881e+02 3.588e+02 6.775e+02, threshold=5.761e+02, percent-clipped=4.0 2023-04-29 16:42:00,513 INFO [train.py:904] (5/8) Epoch 12, batch 9250, loss[loss=0.1762, simple_loss=0.2695, pruned_loss=0.04148, over 16364.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2704, pruned_loss=0.04387, over 3049615.99 frames. ], batch size: 166, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:42:10,066 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9318, 2.7547, 2.5402, 1.9548, 2.4721, 2.7588, 2.6560, 1.9275], device='cuda:5'), covar=tensor([0.0336, 0.0047, 0.0042, 0.0291, 0.0115, 0.0076, 0.0071, 0.0343], device='cuda:5'), in_proj_covar=tensor([0.0124, 0.0065, 0.0067, 0.0122, 0.0078, 0.0086, 0.0077, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 16:42:51,387 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3107, 4.6671, 4.4419, 4.4334, 4.0519, 4.1303, 4.1314, 4.6809], device='cuda:5'), covar=tensor([0.1032, 0.0906, 0.0981, 0.0747, 0.0850, 0.1331, 0.1018, 0.0841], device='cuda:5'), in_proj_covar=tensor([0.0525, 0.0654, 0.0537, 0.0461, 0.0417, 0.0434, 0.0543, 0.0509], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:43:49,111 INFO [train.py:904] (5/8) Epoch 12, batch 9300, loss[loss=0.1698, simple_loss=0.2583, pruned_loss=0.04065, over 16713.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2688, pruned_loss=0.04334, over 3048136.54 frames. ], batch size: 134, lr: 5.60e-03, grad_scale: 4.0 2023-04-29 16:44:31,803 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.549e+02 3.003e+02 3.626e+02 5.994e+02, threshold=6.005e+02, percent-clipped=1.0 2023-04-29 16:45:32,164 INFO [train.py:904] (5/8) Epoch 12, batch 9350, loss[loss=0.1807, simple_loss=0.2755, pruned_loss=0.04299, over 16869.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2694, pruned_loss=0.04365, over 3067642.64 frames. ], batch size: 116, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:45:35,662 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:46:10,456 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-29 16:46:33,086 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:47:13,260 INFO [train.py:904] (5/8) Epoch 12, batch 9400, loss[loss=0.1705, simple_loss=0.2527, pruned_loss=0.04413, over 12305.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2694, pruned_loss=0.04323, over 3070954.19 frames. ], batch size: 246, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:47:39,240 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121064.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:47:50,522 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.435e+02 2.897e+02 3.662e+02 7.614e+02, threshold=5.793e+02, percent-clipped=2.0 2023-04-29 16:48:55,047 INFO [train.py:904] (5/8) Epoch 12, batch 9450, loss[loss=0.1634, simple_loss=0.2626, pruned_loss=0.03213, over 16872.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2708, pruned_loss=0.04316, over 3086895.86 frames. ], batch size: 102, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:50:28,992 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 16:50:34,763 INFO [train.py:904] (5/8) Epoch 12, batch 9500, loss[loss=0.1928, simple_loss=0.2867, pruned_loss=0.04945, over 15380.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2698, pruned_loss=0.04266, over 3091310.47 frames. ], batch size: 191, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:51:01,960 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:51:13,620 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.296e+02 2.757e+02 3.587e+02 6.751e+02, threshold=5.515e+02, percent-clipped=2.0 2023-04-29 16:51:45,989 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9349, 5.3525, 5.4766, 5.2454, 5.2724, 5.8671, 5.3448, 5.0613], device='cuda:5'), covar=tensor([0.0694, 0.1757, 0.2023, 0.1649, 0.2363, 0.0846, 0.1392, 0.2118], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0465, 0.0516, 0.0400, 0.0538, 0.0538, 0.0407, 0.0538], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-29 16:52:20,002 INFO [train.py:904] (5/8) Epoch 12, batch 9550, loss[loss=0.1981, simple_loss=0.294, pruned_loss=0.05108, over 15152.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2693, pruned_loss=0.0425, over 3106176.66 frames. ], batch size: 191, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:52:34,911 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 16:53:08,734 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:53:59,933 INFO [train.py:904] (5/8) Epoch 12, batch 9600, loss[loss=0.2095, simple_loss=0.3045, pruned_loss=0.05725, over 15617.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2714, pruned_loss=0.04365, over 3103006.22 frames. ], batch size: 191, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:54:09,959 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0330, 4.0217, 4.4012, 4.3766, 4.3878, 4.1316, 4.1245, 4.0497], device='cuda:5'), covar=tensor([0.0310, 0.0576, 0.0409, 0.0439, 0.0511, 0.0351, 0.0850, 0.0421], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0331, 0.0331, 0.0317, 0.0379, 0.0356, 0.0440, 0.0285], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-29 16:54:35,189 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.441e+02 3.037e+02 3.482e+02 7.490e+02, threshold=6.075e+02, percent-clipped=2.0 2023-04-29 16:55:45,664 INFO [train.py:904] (5/8) Epoch 12, batch 9650, loss[loss=0.176, simple_loss=0.2707, pruned_loss=0.04065, over 16241.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2735, pruned_loss=0.04443, over 3104072.76 frames. ], batch size: 165, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:56:10,122 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 16:56:52,042 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121332.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:57:07,129 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0122, 1.9448, 2.3858, 3.0411, 2.7987, 3.1887, 2.1603, 3.2700], device='cuda:5'), covar=tensor([0.0155, 0.0378, 0.0268, 0.0192, 0.0217, 0.0130, 0.0368, 0.0100], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0169, 0.0155, 0.0157, 0.0166, 0.0121, 0.0168, 0.0114], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 16:57:25,244 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 16:57:30,750 INFO [train.py:904] (5/8) Epoch 12, batch 9700, loss[loss=0.1873, simple_loss=0.2745, pruned_loss=0.05009, over 16686.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2731, pruned_loss=0.04464, over 3110116.62 frames. ], batch size: 134, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:57:44,336 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:58:07,291 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.472e+02 3.175e+02 3.928e+02 9.920e+02, threshold=6.351e+02, percent-clipped=4.0 2023-04-29 16:58:11,349 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 16:58:31,090 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:59:14,335 INFO [train.py:904] (5/8) Epoch 12, batch 9750, loss[loss=0.1804, simple_loss=0.2767, pruned_loss=0.04204, over 16273.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2712, pruned_loss=0.04407, over 3104525.70 frames. ], batch size: 165, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:59:23,924 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:00:53,749 INFO [train.py:904] (5/8) Epoch 12, batch 9800, loss[loss=0.1631, simple_loss=0.2639, pruned_loss=0.03115, over 16448.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2708, pruned_loss=0.04292, over 3096824.83 frames. ], batch size: 68, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:01:24,639 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:01:29,895 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.372e+02 2.708e+02 3.035e+02 5.819e+02, threshold=5.415e+02, percent-clipped=0.0 2023-04-29 17:01:34,289 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7944, 1.2539, 1.5882, 1.6464, 1.8040, 1.8232, 1.5657, 1.7972], device='cuda:5'), covar=tensor([0.0231, 0.0309, 0.0168, 0.0213, 0.0216, 0.0171, 0.0319, 0.0100], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0169, 0.0155, 0.0157, 0.0167, 0.0121, 0.0169, 0.0113], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:5') 2023-04-29 17:02:38,500 INFO [train.py:904] (5/8) Epoch 12, batch 9850, loss[loss=0.1837, simple_loss=0.2691, pruned_loss=0.04913, over 12635.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2716, pruned_loss=0.04255, over 3074607.41 frames. ], batch size: 247, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:03:17,656 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:04:15,377 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.48 vs. limit=5.0 2023-04-29 17:04:19,972 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 17:04:30,202 INFO [train.py:904] (5/8) Epoch 12, batch 9900, loss[loss=0.1815, simple_loss=0.2845, pruned_loss=0.03926, over 16988.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2717, pruned_loss=0.04211, over 3058769.68 frames. ], batch size: 125, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:05:12,986 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.385e+02 2.886e+02 3.421e+02 6.283e+02, threshold=5.771e+02, percent-clipped=3.0 2023-04-29 17:06:26,653 INFO [train.py:904] (5/8) Epoch 12, batch 9950, loss[loss=0.1956, simple_loss=0.2907, pruned_loss=0.05028, over 15492.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2743, pruned_loss=0.04281, over 3059205.45 frames. ], batch size: 191, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:07:47,209 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.80 vs. limit=5.0 2023-04-29 17:08:27,939 INFO [train.py:904] (5/8) Epoch 12, batch 10000, loss[loss=0.2068, simple_loss=0.2926, pruned_loss=0.06051, over 12731.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.273, pruned_loss=0.04258, over 3056945.81 frames. ], batch size: 250, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:08:44,214 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:08:59,942 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:09:06,355 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 2.322e+02 2.719e+02 3.378e+02 7.215e+02, threshold=5.438e+02, percent-clipped=2.0 2023-04-29 17:10:10,842 INFO [train.py:904] (5/8) Epoch 12, batch 10050, loss[loss=0.212, simple_loss=0.3038, pruned_loss=0.06009, over 16225.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2726, pruned_loss=0.04232, over 3062821.73 frames. ], batch size: 165, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:10:21,423 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:11:46,096 INFO [train.py:904] (5/8) Epoch 12, batch 10100, loss[loss=0.1556, simple_loss=0.2503, pruned_loss=0.03044, over 15503.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2722, pruned_loss=0.0422, over 3052517.57 frames. ], batch size: 191, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:12:06,124 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:12:23,726 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.328e+02 2.737e+02 3.349e+02 5.090e+02, threshold=5.474e+02, percent-clipped=0.0 2023-04-29 17:12:43,408 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8982, 2.2982, 2.2384, 3.0515, 1.8963, 3.2660, 1.7212, 2.6531], device='cuda:5'), covar=tensor([0.1266, 0.0667, 0.1098, 0.0149, 0.0121, 0.0457, 0.1472, 0.0731], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0155, 0.0177, 0.0141, 0.0182, 0.0200, 0.0179, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:5') 2023-04-29 17:12:46,278 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7933, 3.0956, 2.7940, 4.8203, 3.9181, 4.3300, 1.6603, 3.3163], device='cuda:5'), covar=tensor([0.1283, 0.0618, 0.1016, 0.0120, 0.0186, 0.0355, 0.1470, 0.0606], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0155, 0.0177, 0.0141, 0.0182, 0.0201, 0.0179, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:5') 2023-04-29 17:13:30,562 INFO [train.py:904] (5/8) Epoch 13, batch 0, loss[loss=0.272, simple_loss=0.3154, pruned_loss=0.1143, over 16748.00 frames. ], tot_loss[loss=0.272, simple_loss=0.3154, pruned_loss=0.1143, over 16748.00 frames. ], batch size: 83, lr: 5.36e-03, grad_scale: 8.0 2023-04-29 17:13:30,562 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 17:13:38,111 INFO [train.py:938] (5/8) Epoch 13, validation: loss=0.1523, simple_loss=0.2559, pruned_loss=0.02431, over 944034.00 frames. 2023-04-29 17:13:38,112 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 17:14:04,031 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:14:49,499 INFO [train.py:904] (5/8) Epoch 13, batch 50, loss[loss=0.1747, simple_loss=0.2679, pruned_loss=0.04075, over 17239.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2852, pruned_loss=0.06293, over 751929.79 frames. ], batch size: 45, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:15:11,442 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:15:18,822 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.864e+02 2.677e+02 3.187e+02 4.069e+02 9.250e+02, threshold=6.374e+02, percent-clipped=6.0 2023-04-29 17:15:58,276 INFO [train.py:904] (5/8) Epoch 13, batch 100, loss[loss=0.2165, simple_loss=0.2913, pruned_loss=0.0708, over 16706.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2783, pruned_loss=0.05736, over 1325343.09 frames. ], batch size: 134, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:17:04,489 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:17:07,086 INFO [train.py:904] (5/8) Epoch 13, batch 150, loss[loss=0.2254, simple_loss=0.293, pruned_loss=0.07892, over 16706.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2764, pruned_loss=0.05572, over 1765389.24 frames. ], batch size: 134, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:17:27,866 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:17:35,634 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.742e+02 2.538e+02 3.283e+02 4.116e+02 1.264e+03, threshold=6.566e+02, percent-clipped=3.0 2023-04-29 17:18:18,203 INFO [train.py:904] (5/8) Epoch 13, batch 200, loss[loss=0.2056, simple_loss=0.2741, pruned_loss=0.06857, over 16533.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2761, pruned_loss=0.05522, over 2109695.25 frames. ], batch size: 68, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:18:19,832 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3811, 3.4626, 3.4127, 2.8985, 3.2952, 2.0634, 3.0660, 2.7553], device='cuda:5'), covar=tensor([0.0117, 0.0092, 0.0139, 0.0171, 0.0073, 0.2001, 0.0106, 0.0185], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0118, 0.0162, 0.0148, 0.0135, 0.0182, 0.0150, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:18:30,495 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:18:35,254 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 17:19:01,168 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8812, 4.2409, 3.1110, 2.2767, 2.7873, 2.4643, 4.5926, 3.6707], device='cuda:5'), covar=tensor([0.2576, 0.0589, 0.1669, 0.2594, 0.2503, 0.1921, 0.0406, 0.1184], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0251, 0.0282, 0.0276, 0.0265, 0.0219, 0.0262, 0.0290], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:19:26,247 INFO [train.py:904] (5/8) Epoch 13, batch 250, loss[loss=0.1695, simple_loss=0.2647, pruned_loss=0.0372, over 17274.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2733, pruned_loss=0.05446, over 2380803.03 frames. ], batch size: 52, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:19:41,455 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:19:53,888 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 17:19:54,203 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.354e+02 2.915e+02 3.517e+02 1.210e+03, threshold=5.831e+02, percent-clipped=2.0 2023-04-29 17:20:34,753 INFO [train.py:904] (5/8) Epoch 13, batch 300, loss[loss=0.2047, simple_loss=0.266, pruned_loss=0.07172, over 16687.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2702, pruned_loss=0.05288, over 2599239.26 frames. ], batch size: 134, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:20:48,062 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122111.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:20:55,777 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:21:02,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0536, 3.8608, 3.9791, 4.2088, 4.3048, 3.9290, 4.1304, 4.2804], device='cuda:5'), covar=tensor([0.1288, 0.1194, 0.1574, 0.0760, 0.0628, 0.1411, 0.1606, 0.0797], device='cuda:5'), in_proj_covar=tensor([0.0527, 0.0660, 0.0795, 0.0670, 0.0505, 0.0519, 0.0533, 0.0609], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:21:42,725 INFO [train.py:904] (5/8) Epoch 13, batch 350, loss[loss=0.1851, simple_loss=0.2643, pruned_loss=0.05295, over 16486.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2675, pruned_loss=0.05144, over 2766258.94 frames. ], batch size: 146, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:22:13,947 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.554e+02 2.369e+02 2.759e+02 3.305e+02 7.145e+02, threshold=5.518e+02, percent-clipped=2.0 2023-04-29 17:22:20,568 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 17:22:44,452 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122195.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:22:53,149 INFO [train.py:904] (5/8) Epoch 13, batch 400, loss[loss=0.1919, simple_loss=0.2587, pruned_loss=0.06256, over 16818.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2664, pruned_loss=0.05132, over 2887920.76 frames. ], batch size: 90, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:24:03,761 INFO [train.py:904] (5/8) Epoch 13, batch 450, loss[loss=0.1743, simple_loss=0.2498, pruned_loss=0.04939, over 16732.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2639, pruned_loss=0.05046, over 2987892.65 frames. ], batch size: 124, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:24:10,412 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:24:29,466 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9459, 1.8953, 2.4574, 2.9525, 2.6316, 3.4083, 2.1727, 3.2789], device='cuda:5'), covar=tensor([0.0195, 0.0373, 0.0251, 0.0249, 0.0268, 0.0119, 0.0343, 0.0115], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0174, 0.0161, 0.0164, 0.0174, 0.0127, 0.0175, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 17:24:32,947 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.250e+02 2.134e+02 2.674e+02 3.325e+02 9.122e+02, threshold=5.349e+02, percent-clipped=2.0 2023-04-29 17:25:13,890 INFO [train.py:904] (5/8) Epoch 13, batch 500, loss[loss=0.1874, simple_loss=0.2798, pruned_loss=0.04748, over 17262.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2621, pruned_loss=0.049, over 3054126.55 frames. ], batch size: 52, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:25:19,458 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:25:28,578 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 17:25:36,837 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:26:21,452 INFO [train.py:904] (5/8) Epoch 13, batch 550, loss[loss=0.2117, simple_loss=0.2792, pruned_loss=0.07212, over 15576.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2624, pruned_loss=0.04899, over 3117448.65 frames. ], batch size: 191, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:26:50,252 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.266e+02 2.671e+02 3.065e+02 5.763e+02, threshold=5.342e+02, percent-clipped=1.0 2023-04-29 17:27:00,780 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:27:22,538 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9820, 1.9265, 2.4763, 2.9511, 2.6854, 3.3450, 2.2472, 3.3258], device='cuda:5'), covar=tensor([0.0184, 0.0373, 0.0229, 0.0215, 0.0249, 0.0134, 0.0322, 0.0108], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0175, 0.0160, 0.0163, 0.0174, 0.0128, 0.0174, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 17:27:30,037 INFO [train.py:904] (5/8) Epoch 13, batch 600, loss[loss=0.1504, simple_loss=0.2365, pruned_loss=0.03219, over 16805.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2615, pruned_loss=0.04894, over 3163639.49 frames. ], batch size: 42, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:27:39,209 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9513, 4.0712, 2.0803, 4.6206, 2.9088, 4.5796, 2.2117, 3.2433], device='cuda:5'), covar=tensor([0.0229, 0.0295, 0.1512, 0.0182, 0.0740, 0.0348, 0.1474, 0.0549], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0164, 0.0188, 0.0136, 0.0170, 0.0205, 0.0197, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 17:27:49,677 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8501, 3.1879, 3.0613, 5.1596, 4.4142, 4.6967, 1.7087, 3.3455], device='cuda:5'), covar=tensor([0.1268, 0.0612, 0.0971, 0.0168, 0.0191, 0.0324, 0.1446, 0.0644], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0157, 0.0179, 0.0147, 0.0189, 0.0204, 0.0180, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 17:28:17,241 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-04-29 17:28:37,795 INFO [train.py:904] (5/8) Epoch 13, batch 650, loss[loss=0.1635, simple_loss=0.2523, pruned_loss=0.03737, over 17213.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2602, pruned_loss=0.04794, over 3206200.99 frames. ], batch size: 44, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:29:01,211 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:29:07,723 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.494e+02 2.897e+02 3.411e+02 6.772e+02, threshold=5.794e+02, percent-clipped=6.0 2023-04-29 17:29:08,039 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:29:47,561 INFO [train.py:904] (5/8) Epoch 13, batch 700, loss[loss=0.1652, simple_loss=0.2414, pruned_loss=0.04447, over 16362.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.26, pruned_loss=0.0472, over 3229066.68 frames. ], batch size: 36, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:30:26,059 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:56,886 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:57,591 INFO [train.py:904] (5/8) Epoch 13, batch 750, loss[loss=0.1681, simple_loss=0.26, pruned_loss=0.03813, over 17085.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2603, pruned_loss=0.04735, over 3247461.19 frames. ], batch size: 53, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:31:00,516 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2796, 4.3141, 4.7084, 4.7019, 4.7352, 4.3829, 4.4045, 4.2605], device='cuda:5'), covar=tensor([0.0413, 0.0662, 0.0460, 0.0454, 0.0438, 0.0407, 0.0909, 0.0614], device='cuda:5'), in_proj_covar=tensor([0.0354, 0.0374, 0.0371, 0.0353, 0.0420, 0.0396, 0.0493, 0.0320], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 17:31:27,842 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.324e+02 2.801e+02 3.357e+02 5.795e+02, threshold=5.603e+02, percent-clipped=1.0 2023-04-29 17:32:03,002 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0914, 5.0460, 4.9149, 4.5077, 4.5464, 4.9703, 4.9648, 4.6096], device='cuda:5'), covar=tensor([0.0553, 0.0509, 0.0265, 0.0287, 0.0946, 0.0393, 0.0307, 0.0661], device='cuda:5'), in_proj_covar=tensor([0.0257, 0.0345, 0.0307, 0.0287, 0.0327, 0.0328, 0.0208, 0.0358], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:32:09,562 INFO [train.py:904] (5/8) Epoch 13, batch 800, loss[loss=0.1607, simple_loss=0.2582, pruned_loss=0.03159, over 17138.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2599, pruned_loss=0.04705, over 3265902.94 frames. ], batch size: 48, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:32:15,182 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122606.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:15,732 INFO [train.py:904] (5/8) Epoch 13, batch 850, loss[loss=0.1672, simple_loss=0.2618, pruned_loss=0.0363, over 17163.00 frames. ], tot_loss[loss=0.176, simple_loss=0.259, pruned_loss=0.04649, over 3281786.62 frames. ], batch size: 46, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:33:18,971 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:23,692 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7156, 3.9068, 2.4938, 4.3139, 3.0306, 4.3104, 2.4275, 3.0765], device='cuda:5'), covar=tensor([0.0261, 0.0323, 0.1317, 0.0261, 0.0685, 0.0461, 0.1358, 0.0662], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0164, 0.0188, 0.0136, 0.0168, 0.0206, 0.0196, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 17:33:44,228 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.256e+02 2.755e+02 3.448e+02 5.006e+02, threshold=5.510e+02, percent-clipped=0.0 2023-04-29 17:33:46,992 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:34:23,509 INFO [train.py:904] (5/8) Epoch 13, batch 900, loss[loss=0.166, simple_loss=0.2586, pruned_loss=0.03664, over 17224.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2586, pruned_loss=0.0463, over 3297368.47 frames. ], batch size: 52, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:34:35,575 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2156, 5.2254, 4.9972, 4.4766, 5.0219, 1.9793, 4.8056, 4.9825], device='cuda:5'), covar=tensor([0.0074, 0.0066, 0.0156, 0.0352, 0.0080, 0.2286, 0.0113, 0.0144], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0125, 0.0170, 0.0158, 0.0143, 0.0187, 0.0158, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:35:33,185 INFO [train.py:904] (5/8) Epoch 13, batch 950, loss[loss=0.1845, simple_loss=0.2607, pruned_loss=0.05411, over 16791.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2588, pruned_loss=0.04667, over 3304772.05 frames. ], batch size: 124, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:35:55,874 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5086, 2.1570, 2.4099, 4.1890, 2.2591, 2.5703, 2.2659, 2.3599], device='cuda:5'), covar=tensor([0.1034, 0.3512, 0.2304, 0.0464, 0.3444, 0.2324, 0.3282, 0.2761], device='cuda:5'), in_proj_covar=tensor([0.0371, 0.0403, 0.0339, 0.0326, 0.0420, 0.0462, 0.0369, 0.0471], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:36:02,670 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.301e+02 2.627e+02 3.023e+02 6.278e+02, threshold=5.253e+02, percent-clipped=3.0 2023-04-29 17:36:03,710 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:36:40,505 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0108, 4.1005, 4.4095, 4.4109, 4.4102, 4.1226, 4.1750, 4.1083], device='cuda:5'), covar=tensor([0.0377, 0.0659, 0.0366, 0.0388, 0.0454, 0.0432, 0.0750, 0.0517], device='cuda:5'), in_proj_covar=tensor([0.0355, 0.0374, 0.0369, 0.0354, 0.0421, 0.0398, 0.0494, 0.0321], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 17:36:43,330 INFO [train.py:904] (5/8) Epoch 13, batch 1000, loss[loss=0.1832, simple_loss=0.2577, pruned_loss=0.0543, over 15491.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.258, pruned_loss=0.04642, over 3312942.63 frames. ], batch size: 190, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:37:10,560 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:14,623 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:46,885 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 17:37:51,967 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:52,803 INFO [train.py:904] (5/8) Epoch 13, batch 1050, loss[loss=0.1623, simple_loss=0.2553, pruned_loss=0.0347, over 17203.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2579, pruned_loss=0.04613, over 3305214.80 frames. ], batch size: 45, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:38:20,485 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:38:21,176 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.183e+02 2.662e+02 3.059e+02 5.432e+02, threshold=5.323e+02, percent-clipped=2.0 2023-04-29 17:38:32,639 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9258, 2.6341, 2.5402, 2.0376, 2.5804, 2.6771, 2.6280, 1.8685], device='cuda:5'), covar=tensor([0.0364, 0.0079, 0.0056, 0.0296, 0.0096, 0.0099, 0.0084, 0.0359], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0073, 0.0072, 0.0129, 0.0084, 0.0092, 0.0082, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 17:38:58,266 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:39:01,465 INFO [train.py:904] (5/8) Epoch 13, batch 1100, loss[loss=0.1717, simple_loss=0.2451, pruned_loss=0.04911, over 16212.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2573, pruned_loss=0.04577, over 3299503.46 frames. ], batch size: 165, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:39:45,802 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:40:11,772 INFO [train.py:904] (5/8) Epoch 13, batch 1150, loss[loss=0.1465, simple_loss=0.2325, pruned_loss=0.03023, over 17185.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2571, pruned_loss=0.04558, over 3305862.16 frames. ], batch size: 44, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:40:39,513 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.197e+02 2.560e+02 3.227e+02 9.975e+02, threshold=5.120e+02, percent-clipped=3.0 2023-04-29 17:40:43,434 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:41:20,388 INFO [train.py:904] (5/8) Epoch 13, batch 1200, loss[loss=0.1744, simple_loss=0.2672, pruned_loss=0.04083, over 17038.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2565, pruned_loss=0.04529, over 3309935.84 frames. ], batch size: 53, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:41:50,215 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123023.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:42:30,140 INFO [train.py:904] (5/8) Epoch 13, batch 1250, loss[loss=0.1878, simple_loss=0.2582, pruned_loss=0.05874, over 16494.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2568, pruned_loss=0.0462, over 3314206.94 frames. ], batch size: 75, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:42:38,229 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9286, 4.2036, 4.0168, 4.1138, 3.7575, 3.8132, 3.8499, 4.2237], device='cuda:5'), covar=tensor([0.1079, 0.0934, 0.0946, 0.0701, 0.0749, 0.1631, 0.0886, 0.0984], device='cuda:5'), in_proj_covar=tensor([0.0579, 0.0733, 0.0592, 0.0517, 0.0462, 0.0471, 0.0612, 0.0558], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:42:59,818 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.819e+02 2.369e+02 2.845e+02 3.353e+02 6.021e+02, threshold=5.690e+02, percent-clipped=1.0 2023-04-29 17:43:40,476 INFO [train.py:904] (5/8) Epoch 13, batch 1300, loss[loss=0.1782, simple_loss=0.2562, pruned_loss=0.05012, over 16696.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2577, pruned_loss=0.04673, over 3305322.52 frames. ], batch size: 134, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:44:12,464 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:44:49,682 INFO [train.py:904] (5/8) Epoch 13, batch 1350, loss[loss=0.1719, simple_loss=0.2714, pruned_loss=0.03617, over 17243.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2577, pruned_loss=0.04684, over 3307040.38 frames. ], batch size: 52, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:44:53,233 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5060, 5.9502, 5.6813, 5.7097, 5.3049, 5.2168, 5.3626, 6.0816], device='cuda:5'), covar=tensor([0.1191, 0.0862, 0.0901, 0.0751, 0.0749, 0.0707, 0.1059, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0574, 0.0727, 0.0585, 0.0515, 0.0458, 0.0468, 0.0607, 0.0554], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:44:57,609 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7770, 4.5513, 4.8139, 4.9997, 5.2075, 4.4980, 5.1820, 5.1818], device='cuda:5'), covar=tensor([0.1652, 0.1275, 0.1694, 0.0757, 0.0590, 0.0938, 0.0600, 0.0595], device='cuda:5'), in_proj_covar=tensor([0.0562, 0.0707, 0.0854, 0.0714, 0.0539, 0.0559, 0.0568, 0.0655], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:45:17,071 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:45:17,979 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.686e+02 2.284e+02 2.773e+02 3.220e+02 7.351e+02, threshold=5.547e+02, percent-clipped=2.0 2023-04-29 17:45:41,378 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4867, 4.6729, 4.7201, 3.5693, 3.9892, 4.7013, 4.1339, 2.7939], device='cuda:5'), covar=tensor([0.0309, 0.0044, 0.0033, 0.0245, 0.0089, 0.0070, 0.0065, 0.0383], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0072, 0.0072, 0.0129, 0.0084, 0.0093, 0.0083, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 17:45:58,578 INFO [train.py:904] (5/8) Epoch 13, batch 1400, loss[loss=0.1957, simple_loss=0.2639, pruned_loss=0.0638, over 16702.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2581, pruned_loss=0.04739, over 3318219.66 frames. ], batch size: 134, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:46:02,053 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8927, 1.9408, 2.4994, 2.9208, 2.6828, 3.3901, 2.2296, 3.3723], device='cuda:5'), covar=tensor([0.0199, 0.0374, 0.0232, 0.0224, 0.0259, 0.0137, 0.0327, 0.0102], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0176, 0.0161, 0.0165, 0.0176, 0.0129, 0.0176, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 17:46:35,873 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:47:09,291 INFO [train.py:904] (5/8) Epoch 13, batch 1450, loss[loss=0.1727, simple_loss=0.2642, pruned_loss=0.04058, over 17047.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2565, pruned_loss=0.04707, over 3326035.25 frames. ], batch size: 50, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:47:38,911 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.301e+02 2.596e+02 3.249e+02 6.793e+02, threshold=5.192e+02, percent-clipped=2.0 2023-04-29 17:47:46,712 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 17:48:19,760 INFO [train.py:904] (5/8) Epoch 13, batch 1500, loss[loss=0.1781, simple_loss=0.2483, pruned_loss=0.05401, over 16809.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2562, pruned_loss=0.04626, over 3327574.23 frames. ], batch size: 102, lr: 5.33e-03, grad_scale: 4.0 2023-04-29 17:48:20,869 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9611, 2.7340, 2.8552, 2.1857, 2.7537, 2.1553, 2.7936, 2.8780], device='cuda:5'), covar=tensor([0.0298, 0.0793, 0.0454, 0.1570, 0.0708, 0.0864, 0.0580, 0.0759], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0149, 0.0160, 0.0145, 0.0138, 0.0125, 0.0137, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 17:48:27,522 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4854, 5.8490, 5.5851, 5.6741, 5.2302, 5.2402, 5.3039, 5.9542], device='cuda:5'), covar=tensor([0.1258, 0.0886, 0.1040, 0.0695, 0.0778, 0.0609, 0.1095, 0.0936], device='cuda:5'), in_proj_covar=tensor([0.0575, 0.0730, 0.0588, 0.0517, 0.0461, 0.0471, 0.0608, 0.0559], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:48:37,128 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:49:20,403 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9213, 4.6315, 4.9414, 5.1361, 5.3481, 4.6386, 5.3062, 5.2977], device='cuda:5'), covar=tensor([0.1648, 0.1378, 0.1825, 0.0764, 0.0506, 0.0810, 0.0506, 0.0595], device='cuda:5'), in_proj_covar=tensor([0.0574, 0.0720, 0.0869, 0.0727, 0.0549, 0.0569, 0.0580, 0.0669], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:49:30,722 INFO [train.py:904] (5/8) Epoch 13, batch 1550, loss[loss=0.1992, simple_loss=0.267, pruned_loss=0.06564, over 16884.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2577, pruned_loss=0.04711, over 3316353.86 frames. ], batch size: 116, lr: 5.32e-03, grad_scale: 4.0 2023-04-29 17:49:35,980 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2789, 4.4455, 4.1939, 3.9934, 3.4785, 4.4409, 4.2724, 4.0593], device='cuda:5'), covar=tensor([0.0968, 0.0665, 0.0595, 0.0482, 0.2006, 0.0570, 0.0695, 0.0796], device='cuda:5'), in_proj_covar=tensor([0.0265, 0.0358, 0.0318, 0.0297, 0.0340, 0.0341, 0.0214, 0.0371], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 17:50:00,255 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.426e+02 2.890e+02 3.629e+02 7.930e+02, threshold=5.780e+02, percent-clipped=5.0 2023-04-29 17:50:01,832 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:50:39,399 INFO [train.py:904] (5/8) Epoch 13, batch 1600, loss[loss=0.1816, simple_loss=0.2689, pruned_loss=0.04713, over 16650.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2605, pruned_loss=0.04844, over 3300076.56 frames. ], batch size: 57, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:51:15,073 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:51:43,826 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 17:51:47,311 INFO [train.py:904] (5/8) Epoch 13, batch 1650, loss[loss=0.1717, simple_loss=0.2519, pruned_loss=0.04572, over 16804.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2626, pruned_loss=0.04876, over 3295458.65 frames. ], batch size: 102, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:52:18,026 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.454e+02 2.938e+02 3.518e+02 6.758e+02, threshold=5.875e+02, percent-clipped=2.0 2023-04-29 17:52:19,654 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3551, 3.8525, 4.1094, 2.2737, 3.2901, 2.5109, 3.9635, 3.9064], device='cuda:5'), covar=tensor([0.0262, 0.0740, 0.0399, 0.1656, 0.0687, 0.0947, 0.0517, 0.0919], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0148, 0.0158, 0.0144, 0.0137, 0.0125, 0.0136, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 17:52:38,858 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123489.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:52:55,818 INFO [train.py:904] (5/8) Epoch 13, batch 1700, loss[loss=0.1966, simple_loss=0.2759, pruned_loss=0.05863, over 16745.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2647, pruned_loss=0.04971, over 3298784.90 frames. ], batch size: 124, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:53:31,952 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:54:04,802 INFO [train.py:904] (5/8) Epoch 13, batch 1750, loss[loss=0.1667, simple_loss=0.2582, pruned_loss=0.03759, over 17217.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2648, pruned_loss=0.04927, over 3303039.12 frames. ], batch size: 44, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:54:17,157 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8328, 4.8340, 5.2545, 5.2190, 5.2753, 4.9202, 4.8924, 4.6297], device='cuda:5'), covar=tensor([0.0274, 0.0487, 0.0395, 0.0434, 0.0391, 0.0351, 0.0855, 0.0437], device='cuda:5'), in_proj_covar=tensor([0.0360, 0.0380, 0.0379, 0.0358, 0.0426, 0.0403, 0.0500, 0.0324], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 17:54:34,133 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.358e+02 2.760e+02 3.262e+02 5.842e+02, threshold=5.520e+02, percent-clipped=0.0 2023-04-29 17:54:37,989 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:55:14,404 INFO [train.py:904] (5/8) Epoch 13, batch 1800, loss[loss=0.2084, simple_loss=0.2912, pruned_loss=0.06279, over 15463.00 frames. ], tot_loss[loss=0.182, simple_loss=0.266, pruned_loss=0.04899, over 3313233.40 frames. ], batch size: 190, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:56:23,366 INFO [train.py:904] (5/8) Epoch 13, batch 1850, loss[loss=0.1666, simple_loss=0.2603, pruned_loss=0.03644, over 17172.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2659, pruned_loss=0.04878, over 3320882.00 frames. ], batch size: 46, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:56:47,460 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:56:52,513 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.459e+02 2.923e+02 3.563e+02 7.345e+02, threshold=5.846e+02, percent-clipped=2.0 2023-04-29 17:57:33,312 INFO [train.py:904] (5/8) Epoch 13, batch 1900, loss[loss=0.195, simple_loss=0.2713, pruned_loss=0.05938, over 16887.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.265, pruned_loss=0.04798, over 3325362.21 frames. ], batch size: 116, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:58:17,797 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1151, 5.5001, 5.2621, 5.2864, 4.9393, 4.8046, 4.9210, 5.6117], device='cuda:5'), covar=tensor([0.1129, 0.0825, 0.0943, 0.0716, 0.0738, 0.0819, 0.1032, 0.0793], device='cuda:5'), in_proj_covar=tensor([0.0581, 0.0731, 0.0592, 0.0520, 0.0462, 0.0471, 0.0609, 0.0560], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:58:31,006 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 17:58:39,340 INFO [train.py:904] (5/8) Epoch 13, batch 1950, loss[loss=0.186, simple_loss=0.2631, pruned_loss=0.05443, over 16816.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2657, pruned_loss=0.0478, over 3326258.14 frames. ], batch size: 124, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:58:56,955 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-29 17:59:09,940 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.247e+02 2.704e+02 3.281e+02 7.395e+02, threshold=5.408e+02, percent-clipped=2.0 2023-04-29 17:59:24,277 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0243, 4.6251, 3.3502, 2.3868, 2.8898, 2.6296, 4.8104, 3.8802], device='cuda:5'), covar=tensor([0.2396, 0.0440, 0.1483, 0.2427, 0.2690, 0.1798, 0.0309, 0.1066], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0262, 0.0289, 0.0284, 0.0283, 0.0228, 0.0272, 0.0306], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 17:59:25,267 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:59:48,920 INFO [train.py:904] (5/8) Epoch 13, batch 2000, loss[loss=0.1805, simple_loss=0.2703, pruned_loss=0.04535, over 17214.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2647, pruned_loss=0.04762, over 3320739.03 frames. ], batch size: 45, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:00:54,439 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5404, 3.4692, 3.9530, 2.7386, 3.5231, 3.9182, 3.6659, 2.2716], device='cuda:5'), covar=tensor([0.0385, 0.0216, 0.0032, 0.0288, 0.0086, 0.0079, 0.0073, 0.0372], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0072, 0.0071, 0.0127, 0.0083, 0.0092, 0.0082, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 18:00:59,631 INFO [train.py:904] (5/8) Epoch 13, batch 2050, loss[loss=0.2126, simple_loss=0.2732, pruned_loss=0.07599, over 16772.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2644, pruned_loss=0.04779, over 3328254.76 frames. ], batch size: 124, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:01:17,986 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9483, 2.5352, 2.6736, 1.8742, 2.7885, 2.8341, 2.3961, 2.3793], device='cuda:5'), covar=tensor([0.0713, 0.0202, 0.0216, 0.0912, 0.0100, 0.0197, 0.0438, 0.0397], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0101, 0.0090, 0.0138, 0.0070, 0.0111, 0.0121, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 18:01:28,865 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:01:30,784 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.367e+02 2.767e+02 3.495e+02 6.657e+02, threshold=5.534e+02, percent-clipped=3.0 2023-04-29 18:02:09,985 INFO [train.py:904] (5/8) Epoch 13, batch 2100, loss[loss=0.2127, simple_loss=0.2812, pruned_loss=0.07212, over 16846.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2651, pruned_loss=0.04836, over 3318160.21 frames. ], batch size: 116, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:02:45,123 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0302, 3.3741, 3.3912, 2.1752, 2.9031, 2.5022, 3.4467, 3.5942], device='cuda:5'), covar=tensor([0.0236, 0.0742, 0.0494, 0.1548, 0.0715, 0.0849, 0.0500, 0.0745], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0150, 0.0160, 0.0146, 0.0138, 0.0126, 0.0138, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 18:02:54,430 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:20,334 INFO [train.py:904] (5/8) Epoch 13, batch 2150, loss[loss=0.1912, simple_loss=0.2743, pruned_loss=0.054, over 15604.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2668, pruned_loss=0.04946, over 3312719.38 frames. ], batch size: 190, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:03:45,051 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123970.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:50,025 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.337e+02 2.985e+02 3.427e+02 6.976e+02, threshold=5.971e+02, percent-clipped=3.0 2023-04-29 18:04:22,816 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7044, 3.0654, 3.1206, 2.0034, 2.8156, 2.2007, 3.1881, 3.3041], device='cuda:5'), covar=tensor([0.0262, 0.0791, 0.0519, 0.1777, 0.0769, 0.0943, 0.0678, 0.0832], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0151, 0.0161, 0.0147, 0.0138, 0.0126, 0.0139, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 18:04:30,748 INFO [train.py:904] (5/8) Epoch 13, batch 2200, loss[loss=0.1952, simple_loss=0.2674, pruned_loss=0.0615, over 16834.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2674, pruned_loss=0.05017, over 3306876.95 frames. ], batch size: 96, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:04:45,316 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 18:04:53,401 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:19,594 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:40,222 INFO [train.py:904] (5/8) Epoch 13, batch 2250, loss[loss=0.1672, simple_loss=0.2511, pruned_loss=0.04167, over 17231.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2686, pruned_loss=0.05112, over 3305733.94 frames. ], batch size: 44, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:05:53,694 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124062.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:09,556 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.445e+02 2.929e+02 3.478e+02 7.502e+02, threshold=5.857e+02, percent-clipped=2.0 2023-04-29 18:06:22,766 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:31,255 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0438, 5.0396, 4.8617, 4.0174, 4.8986, 1.7773, 4.6431, 4.5865], device='cuda:5'), covar=tensor([0.0113, 0.0091, 0.0167, 0.0468, 0.0103, 0.2790, 0.0145, 0.0258], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0132, 0.0177, 0.0166, 0.0149, 0.0191, 0.0167, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:06:42,881 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:48,023 INFO [train.py:904] (5/8) Epoch 13, batch 2300, loss[loss=0.1823, simple_loss=0.2531, pruned_loss=0.05576, over 16724.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2668, pruned_loss=0.0497, over 3317526.06 frames. ], batch size: 89, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:07:17,896 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124123.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:07:29,437 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:07:38,389 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2743, 5.8337, 6.0535, 5.7033, 5.8482, 6.3270, 5.9351, 5.6585], device='cuda:5'), covar=tensor([0.0898, 0.1792, 0.1702, 0.2120, 0.2429, 0.0911, 0.1220, 0.2091], device='cuda:5'), in_proj_covar=tensor([0.0375, 0.0535, 0.0581, 0.0463, 0.0622, 0.0607, 0.0462, 0.0610], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 18:07:56,751 INFO [train.py:904] (5/8) Epoch 13, batch 2350, loss[loss=0.2024, simple_loss=0.2817, pruned_loss=0.06158, over 16245.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2662, pruned_loss=0.0492, over 3324186.37 frames. ], batch size: 165, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:08:26,418 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.342e+02 2.726e+02 3.219e+02 8.801e+02, threshold=5.452e+02, percent-clipped=1.0 2023-04-29 18:08:48,310 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1170, 4.2007, 4.6769, 2.2007, 4.8097, 4.8529, 3.4534, 3.9194], device='cuda:5'), covar=tensor([0.0671, 0.0194, 0.0191, 0.1092, 0.0071, 0.0135, 0.0339, 0.0304], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0102, 0.0091, 0.0139, 0.0071, 0.0112, 0.0123, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 18:09:06,176 INFO [train.py:904] (5/8) Epoch 13, batch 2400, loss[loss=0.2128, simple_loss=0.2845, pruned_loss=0.07056, over 16863.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2678, pruned_loss=0.04987, over 3312964.67 frames. ], batch size: 116, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:09:42,820 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:10:15,693 INFO [train.py:904] (5/8) Epoch 13, batch 2450, loss[loss=0.1725, simple_loss=0.2677, pruned_loss=0.0386, over 17121.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2692, pruned_loss=0.05016, over 3309502.12 frames. ], batch size: 49, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:10:43,376 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124272.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:10:45,687 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 18:10:46,041 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.400e+02 2.819e+02 3.354e+02 7.582e+02, threshold=5.638e+02, percent-clipped=2.0 2023-04-29 18:11:20,660 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 18:11:24,242 INFO [train.py:904] (5/8) Epoch 13, batch 2500, loss[loss=0.1742, simple_loss=0.2693, pruned_loss=0.03955, over 17070.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2689, pruned_loss=0.04979, over 3304799.04 frames. ], batch size: 50, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:12:03,416 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0416, 4.1609, 2.7874, 4.8415, 3.2804, 4.7943, 2.7427, 3.4902], device='cuda:5'), covar=tensor([0.0241, 0.0330, 0.1243, 0.0212, 0.0732, 0.0340, 0.1284, 0.0615], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0170, 0.0191, 0.0146, 0.0172, 0.0215, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 18:12:09,464 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:12:18,756 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4379, 5.3487, 5.3214, 4.9536, 4.9172, 5.3421, 5.2980, 4.9679], device='cuda:5'), covar=tensor([0.0659, 0.0454, 0.0236, 0.0256, 0.1024, 0.0405, 0.0275, 0.0662], device='cuda:5'), in_proj_covar=tensor([0.0271, 0.0367, 0.0324, 0.0303, 0.0346, 0.0346, 0.0219, 0.0379], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 18:12:31,513 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:12:37,104 INFO [train.py:904] (5/8) Epoch 13, batch 2550, loss[loss=0.1825, simple_loss=0.2596, pruned_loss=0.05275, over 16696.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2682, pruned_loss=0.04934, over 3309949.39 frames. ], batch size: 89, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:06,420 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.275e+02 2.738e+02 3.268e+02 6.282e+02, threshold=5.476e+02, percent-clipped=1.0 2023-04-29 18:13:24,557 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3968, 3.7744, 4.0766, 2.3361, 3.2202, 2.5671, 3.9036, 3.9183], device='cuda:5'), covar=tensor([0.0283, 0.0797, 0.0450, 0.1654, 0.0721, 0.0918, 0.0603, 0.0936], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0152, 0.0162, 0.0147, 0.0138, 0.0127, 0.0139, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 18:13:31,390 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124392.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:13:44,310 INFO [train.py:904] (5/8) Epoch 13, batch 2600, loss[loss=0.1705, simple_loss=0.2598, pruned_loss=0.04057, over 17218.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2676, pruned_loss=0.04887, over 3317548.71 frames. ], batch size: 45, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:54,861 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124409.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:14:00,219 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2920, 5.2864, 5.1119, 4.6101, 5.1434, 2.0882, 4.9102, 5.0858], device='cuda:5'), covar=tensor([0.0088, 0.0070, 0.0143, 0.0338, 0.0089, 0.2205, 0.0123, 0.0149], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0133, 0.0180, 0.0169, 0.0152, 0.0193, 0.0169, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:14:06,947 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124418.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:26,675 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:54,857 INFO [train.py:904] (5/8) Epoch 13, batch 2650, loss[loss=0.1741, simple_loss=0.2567, pruned_loss=0.04572, over 16130.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2682, pruned_loss=0.04878, over 3316114.32 frames. ], batch size: 164, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:15:06,772 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0311, 5.0581, 5.5842, 5.5663, 5.5869, 5.1664, 5.1355, 4.8505], device='cuda:5'), covar=tensor([0.0275, 0.0445, 0.0321, 0.0385, 0.0423, 0.0332, 0.0866, 0.0396], device='cuda:5'), in_proj_covar=tensor([0.0363, 0.0382, 0.0381, 0.0361, 0.0431, 0.0406, 0.0504, 0.0324], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 18:15:25,988 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.199e+02 2.542e+02 3.057e+02 5.216e+02, threshold=5.084e+02, percent-clipped=0.0 2023-04-29 18:15:49,369 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-04-29 18:15:52,793 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:16:04,565 INFO [train.py:904] (5/8) Epoch 13, batch 2700, loss[loss=0.2004, simple_loss=0.275, pruned_loss=0.06293, over 16467.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2693, pruned_loss=0.0487, over 3324385.15 frames. ], batch size: 75, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:16:40,434 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:17:13,331 INFO [train.py:904] (5/8) Epoch 13, batch 2750, loss[loss=0.1711, simple_loss=0.2725, pruned_loss=0.03488, over 17111.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2694, pruned_loss=0.04816, over 3326083.18 frames. ], batch size: 47, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:17:20,340 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4576, 2.3134, 1.8232, 1.9875, 2.6348, 2.4435, 2.7782, 2.8101], device='cuda:5'), covar=tensor([0.0158, 0.0311, 0.0404, 0.0403, 0.0179, 0.0262, 0.0170, 0.0199], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0218, 0.0209, 0.0210, 0.0220, 0.0216, 0.0227, 0.0207], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:17:23,425 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 18:17:32,881 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9882, 4.9650, 5.4576, 5.4583, 5.4775, 5.1041, 5.0883, 4.9084], device='cuda:5'), covar=tensor([0.0298, 0.0514, 0.0359, 0.0399, 0.0378, 0.0319, 0.0852, 0.0360], device='cuda:5'), in_proj_covar=tensor([0.0365, 0.0385, 0.0383, 0.0362, 0.0434, 0.0408, 0.0507, 0.0325], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 18:17:44,014 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.177e+02 2.567e+02 3.156e+02 4.821e+02, threshold=5.134e+02, percent-clipped=0.0 2023-04-29 18:17:46,646 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:18:10,151 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6592, 4.5890, 4.6103, 4.0768, 4.5798, 1.8937, 4.3853, 4.3609], device='cuda:5'), covar=tensor([0.0096, 0.0081, 0.0131, 0.0290, 0.0090, 0.2292, 0.0132, 0.0165], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0132, 0.0178, 0.0168, 0.0150, 0.0190, 0.0167, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:18:22,960 INFO [train.py:904] (5/8) Epoch 13, batch 2800, loss[loss=0.1732, simple_loss=0.2672, pruned_loss=0.03966, over 16973.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2691, pruned_loss=0.04806, over 3333135.26 frames. ], batch size: 53, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:18:24,451 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5433, 4.5833, 4.7474, 4.6356, 4.5428, 5.2055, 4.7321, 4.3604], device='cuda:5'), covar=tensor([0.1324, 0.1956, 0.2014, 0.2084, 0.2885, 0.1031, 0.1483, 0.2514], device='cuda:5'), in_proj_covar=tensor([0.0375, 0.0534, 0.0580, 0.0462, 0.0625, 0.0607, 0.0460, 0.0612], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 18:18:43,560 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9856, 4.8934, 4.8037, 4.2807, 4.8480, 1.9409, 4.6058, 4.6705], device='cuda:5'), covar=tensor([0.0080, 0.0077, 0.0151, 0.0309, 0.0079, 0.2336, 0.0124, 0.0176], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0132, 0.0179, 0.0168, 0.0151, 0.0191, 0.0168, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:18:59,124 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:19:12,864 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:19:31,054 INFO [train.py:904] (5/8) Epoch 13, batch 2850, loss[loss=0.3053, simple_loss=0.3592, pruned_loss=0.1257, over 11710.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2697, pruned_loss=0.04925, over 3314845.14 frames. ], batch size: 247, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:19:46,013 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124663.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:00,129 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.347e+02 2.816e+02 3.672e+02 8.891e+02, threshold=5.632e+02, percent-clipped=4.0 2023-04-29 18:20:24,532 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:34,557 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124699.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:37,526 INFO [train.py:904] (5/8) Epoch 13, batch 2900, loss[loss=0.2126, simple_loss=0.2897, pruned_loss=0.06779, over 16393.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2689, pruned_loss=0.04995, over 3306132.95 frames. ], batch size: 68, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:20:40,129 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:20:43,568 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8859, 4.1119, 2.3859, 4.7482, 3.0317, 4.6547, 2.4088, 3.2877], device='cuda:5'), covar=tensor([0.0252, 0.0333, 0.1449, 0.0153, 0.0769, 0.0387, 0.1439, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0170, 0.0191, 0.0145, 0.0172, 0.0216, 0.0201, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 18:20:59,033 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:07,377 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:29,890 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:46,237 INFO [train.py:904] (5/8) Epoch 13, batch 2950, loss[loss=0.1732, simple_loss=0.2557, pruned_loss=0.04529, over 16136.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2677, pruned_loss=0.0498, over 3314412.47 frames. ], batch size: 35, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:05,859 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:16,914 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.794e+02 2.554e+02 2.996e+02 3.501e+02 7.058e+02, threshold=5.993e+02, percent-clipped=3.0 2023-04-29 18:22:35,664 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:55,211 INFO [train.py:904] (5/8) Epoch 13, batch 3000, loss[loss=0.1566, simple_loss=0.2375, pruned_loss=0.03784, over 17002.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2679, pruned_loss=0.05021, over 3318658.36 frames. ], batch size: 41, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:55,212 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 18:23:04,003 INFO [train.py:938] (5/8) Epoch 13, validation: loss=0.1391, simple_loss=0.2452, pruned_loss=0.01648, over 944034.00 frames. 2023-04-29 18:23:04,004 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 18:23:29,526 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:24:14,169 INFO [train.py:904] (5/8) Epoch 13, batch 3050, loss[loss=0.1802, simple_loss=0.259, pruned_loss=0.05069, over 16788.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2678, pruned_loss=0.05078, over 3312556.93 frames. ], batch size: 102, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:24:44,159 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9287, 4.3768, 3.2385, 2.2779, 2.8966, 2.5085, 4.6869, 3.8603], device='cuda:5'), covar=tensor([0.2493, 0.0541, 0.1500, 0.2481, 0.2525, 0.1795, 0.0311, 0.1028], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0258, 0.0285, 0.0282, 0.0281, 0.0227, 0.0270, 0.0306], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:24:44,776 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.296e+02 3.000e+02 3.565e+02 8.414e+02, threshold=5.999e+02, percent-clipped=2.0 2023-04-29 18:24:55,438 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:25:16,814 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7561, 2.7744, 2.4816, 2.5472, 3.0632, 2.8194, 3.5454, 3.2791], device='cuda:5'), covar=tensor([0.0136, 0.0282, 0.0325, 0.0354, 0.0214, 0.0291, 0.0204, 0.0203], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0218, 0.0209, 0.0210, 0.0219, 0.0216, 0.0228, 0.0208], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:25:25,290 INFO [train.py:904] (5/8) Epoch 13, batch 3100, loss[loss=0.1929, simple_loss=0.2636, pruned_loss=0.06108, over 16435.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2665, pruned_loss=0.04992, over 3318693.75 frames. ], batch size: 146, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:26:01,307 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:26:01,709 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 18:26:28,133 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2730, 5.6670, 5.4457, 5.5003, 5.1279, 5.0792, 5.1434, 5.7918], device='cuda:5'), covar=tensor([0.1108, 0.0868, 0.1000, 0.0683, 0.0824, 0.0642, 0.1040, 0.0797], device='cuda:5'), in_proj_covar=tensor([0.0590, 0.0745, 0.0600, 0.0525, 0.0469, 0.0475, 0.0616, 0.0566], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:26:34,832 INFO [train.py:904] (5/8) Epoch 13, batch 3150, loss[loss=0.1918, simple_loss=0.2669, pruned_loss=0.05837, over 15494.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2659, pruned_loss=0.04952, over 3316232.53 frames. ], batch size: 191, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:26:35,263 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4664, 3.3652, 3.6723, 1.8261, 3.7472, 3.7494, 3.0316, 2.8162], device='cuda:5'), covar=tensor([0.0714, 0.0214, 0.0150, 0.1130, 0.0076, 0.0153, 0.0373, 0.0428], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0101, 0.0089, 0.0138, 0.0070, 0.0111, 0.0121, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 18:27:05,839 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.205e+02 2.626e+02 3.111e+02 5.476e+02, threshold=5.252e+02, percent-clipped=0.0 2023-04-29 18:27:09,150 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124976.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:10,509 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9513, 1.9495, 2.4581, 2.8642, 2.8313, 2.8648, 2.0759, 3.0848], device='cuda:5'), covar=tensor([0.0148, 0.0370, 0.0238, 0.0205, 0.0200, 0.0192, 0.0352, 0.0120], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0181, 0.0165, 0.0171, 0.0179, 0.0133, 0.0179, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:27:23,546 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0466, 5.4642, 5.6775, 5.4077, 5.4433, 6.0820, 5.5279, 5.2598], device='cuda:5'), covar=tensor([0.0875, 0.2062, 0.1904, 0.2160, 0.2717, 0.0907, 0.1631, 0.2436], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0533, 0.0581, 0.0463, 0.0625, 0.0607, 0.0462, 0.0616], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 18:27:27,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5682, 2.2367, 2.3353, 4.3824, 2.2531, 2.6743, 2.3409, 2.4566], device='cuda:5'), covar=tensor([0.1010, 0.3388, 0.2460, 0.0404, 0.3651, 0.2397, 0.3149, 0.3387], device='cuda:5'), in_proj_covar=tensor([0.0375, 0.0407, 0.0339, 0.0325, 0.0418, 0.0470, 0.0371, 0.0475], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:27:34,189 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:43,168 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0529, 5.4400, 5.6336, 5.3966, 5.4372, 6.0814, 5.5547, 5.2740], device='cuda:5'), covar=tensor([0.0905, 0.1951, 0.1802, 0.1891, 0.2487, 0.0805, 0.1462, 0.2248], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0532, 0.0579, 0.0462, 0.0624, 0.0606, 0.0461, 0.0614], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 18:27:45,743 INFO [train.py:904] (5/8) Epoch 13, batch 3200, loss[loss=0.179, simple_loss=0.2546, pruned_loss=0.05172, over 16708.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2643, pruned_loss=0.04872, over 3317607.45 frames. ], batch size: 124, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:49,053 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:28:10,527 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:28:18,740 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-29 18:28:19,934 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 18:28:56,257 INFO [train.py:904] (5/8) Epoch 13, batch 3250, loss[loss=0.156, simple_loss=0.244, pruned_loss=0.03399, over 17132.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2651, pruned_loss=0.04946, over 3312602.63 frames. ], batch size: 46, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:28:56,586 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125052.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:29:21,776 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 18:29:27,132 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.287e+02 2.914e+02 3.406e+02 7.385e+02, threshold=5.827e+02, percent-clipped=6.0 2023-04-29 18:29:46,798 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:29:46,999 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7646, 3.1176, 2.8602, 4.9118, 4.0536, 4.3703, 1.4974, 3.1826], device='cuda:5'), covar=tensor([0.1289, 0.0621, 0.0997, 0.0148, 0.0242, 0.0372, 0.1563, 0.0697], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0162, 0.0183, 0.0159, 0.0199, 0.0211, 0.0184, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 18:30:05,574 INFO [train.py:904] (5/8) Epoch 13, batch 3300, loss[loss=0.186, simple_loss=0.2696, pruned_loss=0.05117, over 17207.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.265, pruned_loss=0.04934, over 3313045.16 frames. ], batch size: 44, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:30:42,367 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125128.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:53,994 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:58,833 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0575, 4.7439, 5.0967, 5.3201, 5.4776, 4.7667, 5.4463, 5.4576], device='cuda:5'), covar=tensor([0.1715, 0.1336, 0.1626, 0.0590, 0.0495, 0.0820, 0.0440, 0.0562], device='cuda:5'), in_proj_covar=tensor([0.0585, 0.0736, 0.0887, 0.0741, 0.0559, 0.0583, 0.0590, 0.0680], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:31:15,066 INFO [train.py:904] (5/8) Epoch 13, batch 3350, loss[loss=0.1794, simple_loss=0.2666, pruned_loss=0.04611, over 16421.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2651, pruned_loss=0.04911, over 3313491.93 frames. ], batch size: 68, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:31:45,941 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.375e+02 2.737e+02 3.421e+02 8.798e+02, threshold=5.473e+02, percent-clipped=2.0 2023-04-29 18:31:49,063 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:32:05,817 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7729, 3.2267, 2.9520, 4.9783, 4.2471, 4.5365, 1.7215, 3.4327], device='cuda:5'), covar=tensor([0.1364, 0.0631, 0.1012, 0.0238, 0.0247, 0.0383, 0.1516, 0.0633], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0183, 0.0159, 0.0199, 0.0212, 0.0185, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 18:32:06,938 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:32:24,432 INFO [train.py:904] (5/8) Epoch 13, batch 3400, loss[loss=0.1755, simple_loss=0.2654, pruned_loss=0.04278, over 17043.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2646, pruned_loss=0.0487, over 3324683.90 frames. ], batch size: 53, lr: 5.29e-03, grad_scale: 4.0 2023-04-29 18:32:49,350 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-04-29 18:32:58,979 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 18:33:29,076 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-04-29 18:33:35,403 INFO [train.py:904] (5/8) Epoch 13, batch 3450, loss[loss=0.1748, simple_loss=0.2579, pruned_loss=0.04584, over 16489.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2635, pruned_loss=0.04809, over 3319480.53 frames. ], batch size: 68, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:33:41,289 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9506, 2.0126, 2.5050, 2.7991, 2.7282, 3.0524, 2.0573, 3.1367], device='cuda:5'), covar=tensor([0.0166, 0.0380, 0.0238, 0.0237, 0.0228, 0.0165, 0.0375, 0.0132], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0179, 0.0163, 0.0170, 0.0178, 0.0133, 0.0178, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:34:07,281 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.393e+02 2.670e+02 3.290e+02 7.203e+02, threshold=5.341e+02, percent-clipped=1.0 2023-04-29 18:34:35,550 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:34:46,248 INFO [train.py:904] (5/8) Epoch 13, batch 3500, loss[loss=0.1748, simple_loss=0.2537, pruned_loss=0.04792, over 16882.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2617, pruned_loss=0.04719, over 3326363.35 frames. ], batch size: 96, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:35:09,886 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:35:43,534 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125342.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:35:57,849 INFO [train.py:904] (5/8) Epoch 13, batch 3550, loss[loss=0.1759, simple_loss=0.2471, pruned_loss=0.05233, over 16746.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.261, pruned_loss=0.04677, over 3321860.99 frames. ], batch size: 124, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:36:03,878 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:13,677 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 18:36:19,302 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:25,601 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5002, 3.2901, 3.6267, 1.9556, 3.6959, 3.7021, 3.0414, 2.9393], device='cuda:5'), covar=tensor([0.0701, 0.0220, 0.0159, 0.1071, 0.0079, 0.0174, 0.0330, 0.0402], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0136, 0.0070, 0.0111, 0.0121, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 18:36:30,597 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 2.251e+02 2.708e+02 3.279e+02 8.250e+02, threshold=5.415e+02, percent-clipped=4.0 2023-04-29 18:36:32,834 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:37:08,980 INFO [train.py:904] (5/8) Epoch 13, batch 3600, loss[loss=0.1543, simple_loss=0.2287, pruned_loss=0.03995, over 16981.00 frames. ], tot_loss[loss=0.176, simple_loss=0.26, pruned_loss=0.04599, over 3311555.71 frames. ], batch size: 41, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:37:31,308 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 18:37:35,909 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:37:55,605 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9433, 4.7406, 4.9675, 5.1830, 5.3302, 4.7203, 5.3555, 5.3360], device='cuda:5'), covar=tensor([0.1648, 0.1222, 0.1657, 0.0682, 0.0569, 0.0789, 0.0486, 0.0536], device='cuda:5'), in_proj_covar=tensor([0.0579, 0.0728, 0.0882, 0.0739, 0.0551, 0.0580, 0.0584, 0.0675], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:37:59,828 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:38:14,930 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 18:38:21,938 INFO [train.py:904] (5/8) Epoch 13, batch 3650, loss[loss=0.1833, simple_loss=0.2489, pruned_loss=0.05882, over 16883.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2596, pruned_loss=0.04636, over 3321940.25 frames. ], batch size: 109, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:38:37,734 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 18:38:57,400 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.272e+02 2.828e+02 3.424e+02 6.489e+02, threshold=5.656e+02, percent-clipped=1.0 2023-04-29 18:38:59,666 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:07,792 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:11,363 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125484.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:36,237 INFO [train.py:904] (5/8) Epoch 13, batch 3700, loss[loss=0.1863, simple_loss=0.2545, pruned_loss=0.05906, over 16723.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2588, pruned_loss=0.04837, over 3310641.28 frames. ], batch size: 134, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:40:10,903 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:40:51,177 INFO [train.py:904] (5/8) Epoch 13, batch 3750, loss[loss=0.183, simple_loss=0.2559, pruned_loss=0.05505, over 16797.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2603, pruned_loss=0.05005, over 3290063.08 frames. ], batch size: 83, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:41:24,166 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.222e+02 2.735e+02 3.211e+02 5.083e+02, threshold=5.471e+02, percent-clipped=0.0 2023-04-29 18:41:55,620 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 18:42:05,179 INFO [train.py:904] (5/8) Epoch 13, batch 3800, loss[loss=0.1674, simple_loss=0.2441, pruned_loss=0.04535, over 16659.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.261, pruned_loss=0.05101, over 3299594.82 frames. ], batch size: 89, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:42:07,577 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:42:44,967 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7991, 1.8937, 2.4094, 2.7471, 2.7794, 2.6601, 1.9207, 2.9783], device='cuda:5'), covar=tensor([0.0143, 0.0329, 0.0220, 0.0198, 0.0195, 0.0184, 0.0346, 0.0079], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0178, 0.0164, 0.0170, 0.0178, 0.0133, 0.0178, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:43:18,388 INFO [train.py:904] (5/8) Epoch 13, batch 3850, loss[loss=0.1773, simple_loss=0.2587, pruned_loss=0.04797, over 16703.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2615, pruned_loss=0.05183, over 3286012.04 frames. ], batch size: 62, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:43:26,603 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6963, 4.7236, 5.0682, 5.0585, 5.0972, 4.7518, 4.7516, 4.4632], device='cuda:5'), covar=tensor([0.0296, 0.0571, 0.0357, 0.0431, 0.0454, 0.0370, 0.0900, 0.0513], device='cuda:5'), in_proj_covar=tensor([0.0362, 0.0382, 0.0378, 0.0362, 0.0428, 0.0403, 0.0503, 0.0323], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 18:43:32,082 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 2023-04-29 18:43:35,168 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:43:50,564 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.444e+02 2.805e+02 3.444e+02 9.574e+02, threshold=5.610e+02, percent-clipped=2.0 2023-04-29 18:43:51,694 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2413, 2.6611, 2.0588, 2.4523, 3.0543, 2.7196, 3.1675, 3.1582], device='cuda:5'), covar=tensor([0.0117, 0.0275, 0.0432, 0.0347, 0.0168, 0.0286, 0.0169, 0.0190], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0215, 0.0207, 0.0208, 0.0215, 0.0213, 0.0225, 0.0207], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:44:29,777 INFO [train.py:904] (5/8) Epoch 13, batch 3900, loss[loss=0.1758, simple_loss=0.2522, pruned_loss=0.0497, over 16224.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2613, pruned_loss=0.05271, over 3280615.56 frames. ], batch size: 165, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:44:45,088 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:44:47,525 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4127, 2.1358, 2.2297, 4.2534, 2.1552, 2.6183, 2.2193, 2.4087], device='cuda:5'), covar=tensor([0.1083, 0.3564, 0.2397, 0.0405, 0.3592, 0.2274, 0.3443, 0.2663], device='cuda:5'), in_proj_covar=tensor([0.0373, 0.0407, 0.0340, 0.0325, 0.0419, 0.0471, 0.0371, 0.0478], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:45:03,342 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:45:14,797 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125732.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:45:42,351 INFO [train.py:904] (5/8) Epoch 13, batch 3950, loss[loss=0.1766, simple_loss=0.2597, pruned_loss=0.04681, over 17051.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2611, pruned_loss=0.05356, over 3278670.15 frames. ], batch size: 50, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:46:16,760 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.289e+02 2.629e+02 3.210e+02 7.394e+02, threshold=5.259e+02, percent-clipped=3.0 2023-04-29 18:46:20,060 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:29,435 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:32,737 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:44,380 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 18:46:55,508 INFO [train.py:904] (5/8) Epoch 13, batch 4000, loss[loss=0.1791, simple_loss=0.2579, pruned_loss=0.05009, over 16746.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2608, pruned_loss=0.05352, over 3273427.68 frames. ], batch size: 124, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:47:38,004 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125832.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:48:05,370 INFO [train.py:904] (5/8) Epoch 13, batch 4050, loss[loss=0.1802, simple_loss=0.2619, pruned_loss=0.04929, over 17171.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2609, pruned_loss=0.05272, over 3279433.53 frames. ], batch size: 46, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:48:15,078 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9116, 2.5438, 2.7271, 1.8155, 2.7888, 2.8554, 2.4027, 2.3383], device='cuda:5'), covar=tensor([0.0785, 0.0232, 0.0173, 0.0917, 0.0081, 0.0166, 0.0443, 0.0452], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0137, 0.0070, 0.0110, 0.0121, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 18:48:23,591 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-04-29 18:48:36,499 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 1.993e+02 2.274e+02 2.780e+02 5.444e+02, threshold=4.549e+02, percent-clipped=3.0 2023-04-29 18:48:45,155 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1372, 3.4624, 3.4938, 1.9374, 2.9058, 2.3059, 3.6353, 3.6211], device='cuda:5'), covar=tensor([0.0233, 0.0685, 0.0589, 0.1978, 0.0836, 0.0945, 0.0538, 0.0767], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0152, 0.0160, 0.0146, 0.0138, 0.0126, 0.0137, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 18:49:15,124 INFO [train.py:904] (5/8) Epoch 13, batch 4100, loss[loss=0.1825, simple_loss=0.2635, pruned_loss=0.05074, over 12292.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2622, pruned_loss=0.05209, over 3266678.56 frames. ], batch size: 246, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:49:16,118 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-29 18:50:13,804 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6639, 3.8914, 2.7386, 2.1958, 2.7876, 2.3321, 4.0338, 3.6079], device='cuda:5'), covar=tensor([0.2680, 0.0653, 0.1778, 0.2250, 0.2401, 0.1897, 0.0484, 0.0887], device='cuda:5'), in_proj_covar=tensor([0.0304, 0.0260, 0.0290, 0.0287, 0.0288, 0.0229, 0.0275, 0.0307], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:50:30,156 INFO [train.py:904] (5/8) Epoch 13, batch 4150, loss[loss=0.2608, simple_loss=0.3223, pruned_loss=0.09965, over 11309.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2694, pruned_loss=0.05508, over 3219727.13 frames. ], batch size: 248, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:50:42,589 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:05,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.529e+02 2.394e+02 2.865e+02 3.650e+02 6.775e+02, threshold=5.730e+02, percent-clipped=10.0 2023-04-29 18:51:40,568 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:48,710 INFO [train.py:904] (5/8) Epoch 13, batch 4200, loss[loss=0.2091, simple_loss=0.2978, pruned_loss=0.06023, over 16362.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2764, pruned_loss=0.05658, over 3200952.28 frames. ], batch size: 146, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:52:04,531 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 18:52:13,333 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 18:52:24,614 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 18:52:34,312 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:02,530 INFO [train.py:904] (5/8) Epoch 13, batch 4250, loss[loss=0.1779, simple_loss=0.2692, pruned_loss=0.0433, over 16850.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2802, pruned_loss=0.05669, over 3194198.65 frames. ], batch size: 42, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:53:13,125 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126059.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:14,167 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:36,050 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.365e+02 2.942e+02 3.644e+02 5.722e+02, threshold=5.884e+02, percent-clipped=0.0 2023-04-29 18:53:39,730 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:43,877 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:45,163 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:54:16,805 INFO [train.py:904] (5/8) Epoch 13, batch 4300, loss[loss=0.2112, simple_loss=0.3002, pruned_loss=0.06106, over 16288.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2811, pruned_loss=0.05579, over 3190107.45 frames. ], batch size: 165, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:54:43,121 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7323, 3.6991, 3.8447, 3.6292, 3.8552, 4.2008, 3.8962, 3.5449], device='cuda:5'), covar=tensor([0.2102, 0.2094, 0.2059, 0.2452, 0.2422, 0.1771, 0.1300, 0.2547], device='cuda:5'), in_proj_covar=tensor([0.0364, 0.0511, 0.0554, 0.0438, 0.0592, 0.0580, 0.0441, 0.0593], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 18:54:48,574 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3947, 5.3702, 5.2211, 4.8585, 4.8832, 5.2260, 5.1505, 4.8928], device='cuda:5'), covar=tensor([0.0458, 0.0229, 0.0203, 0.0225, 0.0808, 0.0273, 0.0234, 0.0607], device='cuda:5'), in_proj_covar=tensor([0.0258, 0.0348, 0.0306, 0.0287, 0.0329, 0.0331, 0.0208, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:54:51,666 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:55:30,685 INFO [train.py:904] (5/8) Epoch 13, batch 4350, loss[loss=0.2049, simple_loss=0.2816, pruned_loss=0.06411, over 11529.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2845, pruned_loss=0.05656, over 3185455.69 frames. ], batch size: 247, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:55:48,440 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2926, 5.5344, 5.2505, 5.3214, 5.0321, 4.8358, 4.9826, 5.6314], device='cuda:5'), covar=tensor([0.0738, 0.0667, 0.0913, 0.0678, 0.0682, 0.0736, 0.0897, 0.0783], device='cuda:5'), in_proj_covar=tensor([0.0565, 0.0713, 0.0578, 0.0507, 0.0454, 0.0461, 0.0594, 0.0547], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:56:06,024 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.441e+02 2.808e+02 3.338e+02 6.392e+02, threshold=5.616e+02, percent-clipped=2.0 2023-04-29 18:56:46,671 INFO [train.py:904] (5/8) Epoch 13, batch 4400, loss[loss=0.2367, simple_loss=0.3102, pruned_loss=0.08161, over 11707.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.286, pruned_loss=0.05746, over 3196904.43 frames. ], batch size: 247, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:57:59,273 INFO [train.py:904] (5/8) Epoch 13, batch 4450, loss[loss=0.2194, simple_loss=0.3096, pruned_loss=0.06458, over 16959.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2896, pruned_loss=0.05869, over 3208863.17 frames. ], batch size: 109, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:58:10,193 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:58:33,039 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.129e+02 2.519e+02 3.084e+02 5.887e+02, threshold=5.038e+02, percent-clipped=2.0 2023-04-29 18:59:14,461 INFO [train.py:904] (5/8) Epoch 13, batch 4500, loss[loss=0.1856, simple_loss=0.2733, pruned_loss=0.04895, over 16438.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.289, pruned_loss=0.0589, over 3208753.08 frames. ], batch size: 75, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:59:20,342 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9041, 1.9603, 2.1411, 3.4575, 1.9537, 2.2766, 2.1272, 2.0824], device='cuda:5'), covar=tensor([0.1163, 0.3191, 0.2349, 0.0518, 0.3905, 0.2358, 0.2877, 0.3306], device='cuda:5'), in_proj_covar=tensor([0.0372, 0.0406, 0.0337, 0.0320, 0.0418, 0.0471, 0.0371, 0.0475], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 18:59:21,428 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:00:25,581 INFO [train.py:904] (5/8) Epoch 13, batch 4550, loss[loss=0.2018, simple_loss=0.2882, pruned_loss=0.05773, over 16277.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2901, pruned_loss=0.0596, over 3219894.79 frames. ], batch size: 165, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:00:28,867 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:00:37,218 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6287, 3.7572, 4.4453, 2.0107, 4.6591, 4.7111, 2.9821, 3.5363], device='cuda:5'), covar=tensor([0.0861, 0.0239, 0.0157, 0.1160, 0.0050, 0.0079, 0.0446, 0.0388], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0103, 0.0089, 0.0138, 0.0070, 0.0110, 0.0123, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 19:00:59,198 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.007e+02 2.258e+02 2.674e+02 4.772e+02, threshold=4.515e+02, percent-clipped=0.0 2023-04-29 19:01:06,178 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:07,312 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:18,954 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9161, 2.7928, 2.0732, 2.7857, 3.3100, 2.9080, 3.4894, 3.4020], device='cuda:5'), covar=tensor([0.0051, 0.0291, 0.0432, 0.0294, 0.0151, 0.0266, 0.0166, 0.0175], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0213, 0.0208, 0.0208, 0.0213, 0.0213, 0.0221, 0.0206], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:01:23,800 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-04-29 19:01:26,591 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 19:01:36,161 INFO [train.py:904] (5/8) Epoch 13, batch 4600, loss[loss=0.1993, simple_loss=0.2833, pruned_loss=0.05763, over 17096.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2908, pruned_loss=0.0597, over 3212400.79 frames. ], batch size: 47, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:02:22,722 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:39,264 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:53,246 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:55,164 INFO [train.py:904] (5/8) Epoch 13, batch 4650, loss[loss=0.2048, simple_loss=0.2856, pruned_loss=0.06206, over 16874.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2902, pruned_loss=0.05995, over 3209194.71 frames. ], batch size: 42, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:03:00,885 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 19:03:27,526 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 1.885e+02 2.230e+02 2.796e+02 5.506e+02, threshold=4.460e+02, percent-clipped=0.0 2023-04-29 19:03:29,714 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 19:04:07,666 INFO [train.py:904] (5/8) Epoch 13, batch 4700, loss[loss=0.1889, simple_loss=0.2776, pruned_loss=0.05009, over 16492.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2873, pruned_loss=0.05883, over 3209999.67 frames. ], batch size: 146, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:04:21,606 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:05:21,246 INFO [train.py:904] (5/8) Epoch 13, batch 4750, loss[loss=0.1771, simple_loss=0.265, pruned_loss=0.04459, over 16698.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2835, pruned_loss=0.0568, over 3205642.51 frames. ], batch size: 62, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:05:53,858 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.457e+02 1.907e+02 2.465e+02 2.943e+02 4.218e+02, threshold=4.930e+02, percent-clipped=1.0 2023-04-29 19:06:30,733 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 19:06:32,283 INFO [train.py:904] (5/8) Epoch 13, batch 4800, loss[loss=0.1972, simple_loss=0.279, pruned_loss=0.0577, over 11956.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2809, pruned_loss=0.05514, over 3189505.61 frames. ], batch size: 246, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:07:26,368 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 19:07:43,569 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7210, 3.0147, 2.6381, 4.9587, 3.8379, 4.3498, 1.6669, 3.0842], device='cuda:5'), covar=tensor([0.1351, 0.0662, 0.1149, 0.0104, 0.0317, 0.0362, 0.1514, 0.0775], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0185, 0.0159, 0.0201, 0.0208, 0.0184, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 19:07:46,147 INFO [train.py:904] (5/8) Epoch 13, batch 4850, loss[loss=0.1822, simple_loss=0.2772, pruned_loss=0.04363, over 16759.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2818, pruned_loss=0.05471, over 3178656.10 frames. ], batch size: 83, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:07:50,122 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:08:19,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.948e+02 2.386e+02 2.716e+02 6.913e+02, threshold=4.771e+02, percent-clipped=1.0 2023-04-29 19:08:36,877 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4719, 4.2273, 4.2941, 2.7440, 3.6390, 4.1359, 3.6851, 2.3116], device='cuda:5'), covar=tensor([0.0419, 0.0024, 0.0024, 0.0324, 0.0075, 0.0091, 0.0069, 0.0382], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0070, 0.0071, 0.0126, 0.0083, 0.0091, 0.0082, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 19:08:59,252 INFO [train.py:904] (5/8) Epoch 13, batch 4900, loss[loss=0.1943, simple_loss=0.2883, pruned_loss=0.0502, over 15367.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.281, pruned_loss=0.05335, over 3180242.71 frames. ], batch size: 190, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:08:59,642 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:09:49,787 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:10:12,994 INFO [train.py:904] (5/8) Epoch 13, batch 4950, loss[loss=0.198, simple_loss=0.2814, pruned_loss=0.05733, over 16569.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2806, pruned_loss=0.05278, over 3192673.07 frames. ], batch size: 75, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:10:39,195 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9990, 3.9434, 3.9286, 3.2446, 3.9369, 1.7131, 3.7101, 3.4488], device='cuda:5'), covar=tensor([0.0097, 0.0099, 0.0128, 0.0363, 0.0090, 0.2484, 0.0125, 0.0237], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0124, 0.0168, 0.0159, 0.0141, 0.0180, 0.0158, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:10:45,733 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.239e+02 2.710e+02 3.273e+02 4.681e+02, threshold=5.421e+02, percent-clipped=0.0 2023-04-29 19:11:26,267 INFO [train.py:904] (5/8) Epoch 13, batch 5000, loss[loss=0.1932, simple_loss=0.278, pruned_loss=0.05424, over 16572.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2822, pruned_loss=0.053, over 3207736.66 frames. ], batch size: 62, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:11:32,789 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:11:47,361 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126816.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:12:38,206 INFO [train.py:904] (5/8) Epoch 13, batch 5050, loss[loss=0.198, simple_loss=0.2947, pruned_loss=0.05069, over 16720.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2829, pruned_loss=0.05298, over 3213494.99 frames. ], batch size: 89, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:13:11,130 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.161e+02 2.561e+02 3.247e+02 4.827e+02, threshold=5.121e+02, percent-clipped=0.0 2023-04-29 19:13:14,621 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:13:51,473 INFO [train.py:904] (5/8) Epoch 13, batch 5100, loss[loss=0.1844, simple_loss=0.2677, pruned_loss=0.05051, over 16609.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2809, pruned_loss=0.05224, over 3206695.95 frames. ], batch size: 68, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:14:02,215 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-29 19:15:04,900 INFO [train.py:904] (5/8) Epoch 13, batch 5150, loss[loss=0.1845, simple_loss=0.2821, pruned_loss=0.04344, over 16894.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2811, pruned_loss=0.05172, over 3210532.51 frames. ], batch size: 116, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:15:14,518 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6115, 4.5247, 4.5934, 3.2012, 3.8828, 4.4634, 3.8899, 2.6520], device='cuda:5'), covar=tensor([0.0444, 0.0023, 0.0022, 0.0265, 0.0074, 0.0091, 0.0065, 0.0334], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0071, 0.0072, 0.0128, 0.0085, 0.0093, 0.0083, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 19:15:16,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3222, 2.0188, 2.7325, 3.2238, 3.0752, 3.6143, 1.8306, 3.5349], device='cuda:5'), covar=tensor([0.0097, 0.0399, 0.0222, 0.0158, 0.0170, 0.0085, 0.0500, 0.0067], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0176, 0.0161, 0.0164, 0.0175, 0.0131, 0.0175, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 19:15:37,791 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.071e+02 2.483e+02 2.841e+02 7.629e+02, threshold=4.965e+02, percent-clipped=2.0 2023-04-29 19:16:11,923 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7296, 3.9874, 2.9349, 2.3159, 2.6223, 2.4014, 4.0476, 3.4775], device='cuda:5'), covar=tensor([0.2475, 0.0562, 0.1619, 0.2440, 0.2409, 0.1766, 0.0421, 0.1049], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0259, 0.0287, 0.0285, 0.0281, 0.0228, 0.0271, 0.0305], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:16:17,746 INFO [train.py:904] (5/8) Epoch 13, batch 5200, loss[loss=0.1663, simple_loss=0.2536, pruned_loss=0.0395, over 16568.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2805, pruned_loss=0.05121, over 3204872.68 frames. ], batch size: 68, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:16:51,353 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 19:17:08,370 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:17:31,272 INFO [train.py:904] (5/8) Epoch 13, batch 5250, loss[loss=0.17, simple_loss=0.264, pruned_loss=0.03799, over 16802.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2776, pruned_loss=0.05052, over 3213314.53 frames. ], batch size: 83, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:04,193 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.082e+02 2.366e+02 2.761e+02 5.233e+02, threshold=4.733e+02, percent-clipped=1.0 2023-04-29 19:18:11,867 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-29 19:18:18,851 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:18:20,553 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 19:18:44,231 INFO [train.py:904] (5/8) Epoch 13, batch 5300, loss[loss=0.1726, simple_loss=0.2627, pruned_loss=0.04128, over 16658.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2737, pruned_loss=0.04913, over 3211858.35 frames. ], batch size: 134, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:49,502 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9393, 5.2586, 5.0003, 5.0398, 4.6817, 4.7241, 4.6575, 5.3439], device='cuda:5'), covar=tensor([0.1076, 0.0793, 0.0995, 0.0684, 0.0835, 0.0790, 0.1010, 0.0852], device='cuda:5'), in_proj_covar=tensor([0.0562, 0.0705, 0.0571, 0.0496, 0.0446, 0.0450, 0.0585, 0.0542], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:18:50,751 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:19:13,340 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 19:19:26,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7793, 3.8496, 2.0010, 4.4648, 2.7949, 4.3338, 2.4312, 3.0064], device='cuda:5'), covar=tensor([0.0217, 0.0309, 0.1797, 0.0101, 0.0808, 0.0325, 0.1409, 0.0696], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0164, 0.0187, 0.0135, 0.0166, 0.0205, 0.0194, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 19:19:58,323 INFO [train.py:904] (5/8) Epoch 13, batch 5350, loss[loss=0.1976, simple_loss=0.2897, pruned_loss=0.05276, over 15350.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2718, pruned_loss=0.04829, over 3208986.65 frames. ], batch size: 190, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:20:01,845 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127154.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:27,671 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:32,664 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.449e+02 2.172e+02 2.538e+02 2.960e+02 5.660e+02, threshold=5.076e+02, percent-clipped=1.0 2023-04-29 19:20:42,159 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0900, 3.4998, 3.6898, 1.4525, 3.7407, 3.8973, 2.8381, 2.6514], device='cuda:5'), covar=tensor([0.1221, 0.0174, 0.0143, 0.1521, 0.0098, 0.0107, 0.0422, 0.0643], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0103, 0.0089, 0.0138, 0.0070, 0.0109, 0.0122, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 19:21:01,680 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7656, 4.9992, 5.3650, 5.2721, 5.2749, 4.9037, 4.4802, 4.5811], device='cuda:5'), covar=tensor([0.0504, 0.0655, 0.0416, 0.0584, 0.0718, 0.0509, 0.1551, 0.0531], device='cuda:5'), in_proj_covar=tensor([0.0345, 0.0361, 0.0360, 0.0351, 0.0411, 0.0385, 0.0482, 0.0309], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 19:21:10,972 INFO [train.py:904] (5/8) Epoch 13, batch 5400, loss[loss=0.2295, simple_loss=0.3078, pruned_loss=0.07558, over 12273.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2749, pruned_loss=0.04941, over 3210485.30 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:22:03,752 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6837, 3.7357, 2.9374, 2.3322, 2.7234, 2.4443, 3.9669, 3.5876], device='cuda:5'), covar=tensor([0.2556, 0.0752, 0.1586, 0.2341, 0.2156, 0.1696, 0.0449, 0.0957], device='cuda:5'), in_proj_covar=tensor([0.0306, 0.0259, 0.0286, 0.0285, 0.0281, 0.0227, 0.0272, 0.0304], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:22:20,958 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:22:26,314 INFO [train.py:904] (5/8) Epoch 13, batch 5450, loss[loss=0.2565, simple_loss=0.3175, pruned_loss=0.09772, over 11735.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2784, pruned_loss=0.05138, over 3199904.31 frames. ], batch size: 247, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:22:29,849 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2162, 4.1918, 4.6086, 4.5880, 4.5731, 4.2864, 4.2641, 4.1899], device='cuda:5'), covar=tensor([0.0300, 0.0562, 0.0357, 0.0418, 0.0471, 0.0385, 0.0950, 0.0467], device='cuda:5'), in_proj_covar=tensor([0.0347, 0.0364, 0.0363, 0.0352, 0.0414, 0.0387, 0.0484, 0.0310], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 19:23:02,326 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.258e+02 2.675e+02 3.738e+02 1.100e+03, threshold=5.349e+02, percent-clipped=10.0 2023-04-29 19:23:43,868 INFO [train.py:904] (5/8) Epoch 13, batch 5500, loss[loss=0.2119, simple_loss=0.3004, pruned_loss=0.06171, over 16824.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2863, pruned_loss=0.05666, over 3155162.99 frames. ], batch size: 102, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:23:56,258 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 19:23:57,136 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:25:00,722 INFO [train.py:904] (5/8) Epoch 13, batch 5550, loss[loss=0.2527, simple_loss=0.3297, pruned_loss=0.08779, over 15332.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2947, pruned_loss=0.0626, over 3125758.08 frames. ], batch size: 191, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:25:26,700 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5357, 4.6541, 4.8463, 4.6552, 4.7011, 5.2293, 4.7345, 4.4819], device='cuda:5'), covar=tensor([0.1154, 0.1929, 0.1966, 0.1844, 0.2453, 0.1021, 0.1568, 0.2480], device='cuda:5'), in_proj_covar=tensor([0.0360, 0.0504, 0.0541, 0.0427, 0.0583, 0.0572, 0.0433, 0.0583], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 19:25:32,515 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9793, 2.7718, 2.7340, 2.0885, 2.5871, 2.1484, 2.7037, 2.9089], device='cuda:5'), covar=tensor([0.0332, 0.0652, 0.0500, 0.1610, 0.0768, 0.0884, 0.0606, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0149, 0.0160, 0.0146, 0.0139, 0.0126, 0.0139, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 19:25:38,576 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.215e+02 3.416e+02 4.230e+02 5.078e+02 9.045e+02, threshold=8.460e+02, percent-clipped=18.0 2023-04-29 19:26:20,883 INFO [train.py:904] (5/8) Epoch 13, batch 5600, loss[loss=0.3057, simple_loss=0.3573, pruned_loss=0.127, over 11073.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3003, pruned_loss=0.068, over 3070282.55 frames. ], batch size: 247, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:27:02,211 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 19:27:41,570 INFO [train.py:904] (5/8) Epoch 13, batch 5650, loss[loss=0.2173, simple_loss=0.2942, pruned_loss=0.07021, over 16452.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3054, pruned_loss=0.0722, over 3039995.27 frames. ], batch size: 68, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:27:55,542 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9357, 2.8715, 2.7034, 4.7094, 3.6479, 4.2539, 1.5961, 3.0285], device='cuda:5'), covar=tensor([0.1138, 0.0687, 0.1088, 0.0178, 0.0326, 0.0343, 0.1452, 0.0762], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0163, 0.0184, 0.0158, 0.0201, 0.0208, 0.0185, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 19:28:12,815 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:28:17,741 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 3.489e+02 4.637e+02 5.886e+02 1.352e+03, threshold=9.273e+02, percent-clipped=4.0 2023-04-29 19:28:21,276 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 19:28:31,896 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0035, 4.7422, 4.9884, 5.2033, 5.3842, 4.7195, 5.3629, 5.3344], device='cuda:5'), covar=tensor([0.1575, 0.1272, 0.1469, 0.0572, 0.0441, 0.0848, 0.0449, 0.0610], device='cuda:5'), in_proj_covar=tensor([0.0543, 0.0682, 0.0813, 0.0686, 0.0519, 0.0538, 0.0549, 0.0631], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:28:36,980 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 19:28:58,357 INFO [train.py:904] (5/8) Epoch 13, batch 5700, loss[loss=0.2917, simple_loss=0.3416, pruned_loss=0.1209, over 11639.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3066, pruned_loss=0.07358, over 3032874.39 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:29:27,678 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:29:42,122 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:30:18,045 INFO [train.py:904] (5/8) Epoch 13, batch 5750, loss[loss=0.2495, simple_loss=0.3149, pruned_loss=0.09207, over 11350.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3085, pruned_loss=0.07429, over 3025643.01 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:30:56,353 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.152e+02 3.042e+02 3.668e+02 4.519e+02 8.025e+02, threshold=7.337e+02, percent-clipped=0.0 2023-04-29 19:31:21,789 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:31:39,677 INFO [train.py:904] (5/8) Epoch 13, batch 5800, loss[loss=0.2261, simple_loss=0.3174, pruned_loss=0.06741, over 16790.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3082, pruned_loss=0.07277, over 3042535.81 frames. ], batch size: 116, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:31:40,299 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7970, 2.5122, 2.4291, 3.3010, 2.4317, 3.6197, 1.4670, 2.8504], device='cuda:5'), covar=tensor([0.1246, 0.0656, 0.1057, 0.0165, 0.0149, 0.0365, 0.1563, 0.0705], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0163, 0.0184, 0.0157, 0.0200, 0.0208, 0.0184, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 19:31:44,432 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:31:49,493 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5074, 1.6754, 2.0845, 2.4145, 2.5251, 2.8236, 1.7818, 2.7647], device='cuda:5'), covar=tensor([0.0182, 0.0448, 0.0279, 0.0280, 0.0233, 0.0153, 0.0423, 0.0126], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0163, 0.0174, 0.0130, 0.0175, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 19:32:57,629 INFO [train.py:904] (5/8) Epoch 13, batch 5850, loss[loss=0.2189, simple_loss=0.2978, pruned_loss=0.07, over 16580.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3058, pruned_loss=0.07102, over 3047130.76 frames. ], batch size: 62, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:33:15,137 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127662.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:33:36,797 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.121e+02 3.118e+02 3.715e+02 4.396e+02 1.301e+03, threshold=7.431e+02, percent-clipped=3.0 2023-04-29 19:34:19,531 INFO [train.py:904] (5/8) Epoch 13, batch 5900, loss[loss=0.2571, simple_loss=0.3209, pruned_loss=0.0967, over 11486.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3052, pruned_loss=0.07059, over 3054088.65 frames. ], batch size: 246, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:34:30,657 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 19:34:57,908 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:35:36,306 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0215, 3.9902, 3.9552, 3.3269, 4.0010, 1.7973, 3.7637, 3.6285], device='cuda:5'), covar=tensor([0.0108, 0.0105, 0.0150, 0.0308, 0.0084, 0.2357, 0.0119, 0.0209], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0124, 0.0170, 0.0160, 0.0141, 0.0182, 0.0157, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:35:42,727 INFO [train.py:904] (5/8) Epoch 13, batch 5950, loss[loss=0.2108, simple_loss=0.305, pruned_loss=0.05827, over 16837.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3059, pruned_loss=0.06919, over 3054179.94 frames. ], batch size: 102, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:36:16,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8232, 5.2083, 5.4373, 5.2113, 5.2539, 5.7706, 5.2177, 5.0221], device='cuda:5'), covar=tensor([0.0983, 0.1648, 0.1665, 0.1738, 0.2144, 0.0910, 0.1577, 0.2399], device='cuda:5'), in_proj_covar=tensor([0.0367, 0.0512, 0.0556, 0.0437, 0.0593, 0.0581, 0.0444, 0.0592], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 19:36:21,829 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.141e+02 2.934e+02 3.631e+02 4.257e+02 8.531e+02, threshold=7.262e+02, percent-clipped=1.0 2023-04-29 19:36:26,906 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:36:27,137 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 19:36:32,793 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:36:41,787 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 19:37:04,082 INFO [train.py:904] (5/8) Epoch 13, batch 6000, loss[loss=0.2044, simple_loss=0.283, pruned_loss=0.0629, over 16657.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3044, pruned_loss=0.06844, over 3068640.68 frames. ], batch size: 62, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:37:04,083 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 19:37:14,217 INFO [train.py:938] (5/8) Epoch 13, validation: loss=0.1599, simple_loss=0.2726, pruned_loss=0.02359, over 944034.00 frames. 2023-04-29 19:37:14,218 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 19:38:12,170 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5822, 2.6496, 2.3071, 4.0717, 3.0426, 3.9793, 1.3348, 2.8435], device='cuda:5'), covar=tensor([0.1284, 0.0714, 0.1251, 0.0192, 0.0304, 0.0400, 0.1619, 0.0780], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0163, 0.0184, 0.0157, 0.0202, 0.0209, 0.0184, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 19:38:13,857 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:38:27,283 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6319, 2.2777, 1.7355, 1.9917, 2.6152, 2.2587, 2.5572, 2.7804], device='cuda:5'), covar=tensor([0.0153, 0.0349, 0.0469, 0.0395, 0.0211, 0.0332, 0.0226, 0.0233], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0210, 0.0204, 0.0206, 0.0209, 0.0208, 0.0215, 0.0201], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:38:32,251 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127850.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:38:35,161 INFO [train.py:904] (5/8) Epoch 13, batch 6050, loss[loss=0.2065, simple_loss=0.2909, pruned_loss=0.06106, over 17049.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3024, pruned_loss=0.06694, over 3112854.44 frames. ], batch size: 41, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:38:36,664 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8954, 4.0743, 3.2402, 2.3640, 3.0057, 2.5511, 4.2589, 3.8486], device='cuda:5'), covar=tensor([0.2535, 0.0669, 0.1433, 0.2367, 0.2158, 0.1689, 0.0444, 0.0996], device='cuda:5'), in_proj_covar=tensor([0.0308, 0.0259, 0.0288, 0.0285, 0.0283, 0.0227, 0.0273, 0.0305], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:39:00,247 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5781, 3.6638, 2.8476, 2.1250, 2.4817, 2.2837, 3.7543, 3.4177], device='cuda:5'), covar=tensor([0.2715, 0.0643, 0.1514, 0.2482, 0.2361, 0.1790, 0.0498, 0.1070], device='cuda:5'), in_proj_covar=tensor([0.0308, 0.0259, 0.0287, 0.0285, 0.0283, 0.0226, 0.0272, 0.0305], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:39:07,508 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:14,942 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.979e+02 3.574e+02 4.345e+02 1.015e+03, threshold=7.148e+02, percent-clipped=2.0 2023-04-29 19:39:28,471 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:55,369 INFO [train.py:904] (5/8) Epoch 13, batch 6100, loss[loss=0.2146, simple_loss=0.2997, pruned_loss=0.06471, over 16406.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.3015, pruned_loss=0.06565, over 3130165.55 frames. ], batch size: 146, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:40:01,653 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:12,652 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:47,066 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:41:10,133 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-04-29 19:41:15,209 INFO [train.py:904] (5/8) Epoch 13, batch 6150, loss[loss=0.2315, simple_loss=0.3063, pruned_loss=0.07832, over 11904.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2998, pruned_loss=0.06525, over 3129975.44 frames. ], batch size: 248, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:41:17,426 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127953.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:41:39,469 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0249, 5.3402, 5.0847, 5.0884, 4.8061, 4.7406, 4.8005, 5.4409], device='cuda:5'), covar=tensor([0.0990, 0.0743, 0.0898, 0.0732, 0.0797, 0.0823, 0.0985, 0.0730], device='cuda:5'), in_proj_covar=tensor([0.0565, 0.0700, 0.0576, 0.0502, 0.0442, 0.0452, 0.0588, 0.0542], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:41:39,653 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9884, 3.3888, 3.3555, 1.9900, 2.9630, 2.2616, 3.5584, 3.5309], device='cuda:5'), covar=tensor([0.0251, 0.0736, 0.0570, 0.1848, 0.0752, 0.0913, 0.0610, 0.0888], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0150, 0.0160, 0.0146, 0.0140, 0.0126, 0.0139, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 19:41:56,819 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 3.037e+02 3.583e+02 4.334e+02 1.010e+03, threshold=7.167e+02, percent-clipped=2.0 2023-04-29 19:42:39,415 INFO [train.py:904] (5/8) Epoch 13, batch 6200, loss[loss=0.1961, simple_loss=0.2825, pruned_loss=0.05491, over 16461.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2977, pruned_loss=0.06459, over 3132639.24 frames. ], batch size: 68, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:43:05,199 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:43:26,202 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 19:43:57,685 INFO [train.py:904] (5/8) Epoch 13, batch 6250, loss[loss=0.215, simple_loss=0.2985, pruned_loss=0.06573, over 15454.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2966, pruned_loss=0.06403, over 3135048.47 frames. ], batch size: 190, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:44:37,190 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.817e+02 3.684e+02 4.282e+02 1.153e+03, threshold=7.367e+02, percent-clipped=4.0 2023-04-29 19:44:42,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1414, 3.1943, 3.4737, 1.7532, 3.6164, 3.6898, 2.7459, 2.6627], device='cuda:5'), covar=tensor([0.0799, 0.0199, 0.0142, 0.1088, 0.0054, 0.0117, 0.0393, 0.0451], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0101, 0.0088, 0.0137, 0.0068, 0.0108, 0.0120, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 19:44:45,189 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 19:45:13,045 INFO [train.py:904] (5/8) Epoch 13, batch 6300, loss[loss=0.2001, simple_loss=0.2866, pruned_loss=0.05684, over 16667.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2969, pruned_loss=0.06411, over 3127100.33 frames. ], batch size: 134, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:45:34,464 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7313, 2.4065, 2.3400, 3.1148, 2.3560, 3.5626, 1.4067, 2.7332], device='cuda:5'), covar=tensor([0.1311, 0.0717, 0.1146, 0.0163, 0.0191, 0.0423, 0.1647, 0.0791], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0163, 0.0183, 0.0157, 0.0202, 0.0208, 0.0184, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 19:45:38,858 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4394, 2.9061, 2.6392, 2.2293, 2.2661, 2.1998, 2.8484, 2.8638], device='cuda:5'), covar=tensor([0.2165, 0.0710, 0.1413, 0.2106, 0.2263, 0.1940, 0.0473, 0.1093], device='cuda:5'), in_proj_covar=tensor([0.0308, 0.0260, 0.0289, 0.0287, 0.0284, 0.0228, 0.0273, 0.0306], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:46:02,726 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 19:46:08,206 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:46:33,873 INFO [train.py:904] (5/8) Epoch 13, batch 6350, loss[loss=0.2375, simple_loss=0.3163, pruned_loss=0.0793, over 16843.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2979, pruned_loss=0.06515, over 3124712.12 frames. ], batch size: 116, lr: 5.22e-03, grad_scale: 4.0 2023-04-29 19:47:13,679 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 3.177e+02 3.833e+02 4.854e+02 7.993e+02, threshold=7.666e+02, percent-clipped=3.0 2023-04-29 19:47:25,250 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128185.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:47:44,596 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1330, 3.1849, 1.7492, 3.4677, 2.3304, 3.4558, 2.0125, 2.5275], device='cuda:5'), covar=tensor([0.0266, 0.0419, 0.1823, 0.0204, 0.0870, 0.0607, 0.1557, 0.0789], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0169, 0.0192, 0.0139, 0.0169, 0.0209, 0.0199, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 19:47:49,930 INFO [train.py:904] (5/8) Epoch 13, batch 6400, loss[loss=0.2673, simple_loss=0.3328, pruned_loss=0.1009, over 11299.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2991, pruned_loss=0.06672, over 3097726.94 frames. ], batch size: 248, lr: 5.22e-03, grad_scale: 8.0 2023-04-29 19:47:56,274 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:30,015 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:30,136 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0991, 3.1981, 3.4770, 1.7877, 3.6224, 3.6748, 2.8215, 2.7363], device='cuda:5'), covar=tensor([0.0887, 0.0218, 0.0168, 0.1207, 0.0060, 0.0142, 0.0379, 0.0481], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0101, 0.0088, 0.0137, 0.0069, 0.0109, 0.0121, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 19:48:36,414 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:49:04,035 INFO [train.py:904] (5/8) Epoch 13, batch 6450, loss[loss=0.2135, simple_loss=0.301, pruned_loss=0.06306, over 16915.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2997, pruned_loss=0.06688, over 3072012.33 frames. ], batch size: 116, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:49:48,891 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.912e+02 3.353e+02 4.052e+02 7.402e+02, threshold=6.705e+02, percent-clipped=0.0 2023-04-29 19:50:21,569 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 19:50:21,895 INFO [train.py:904] (5/8) Epoch 13, batch 6500, loss[loss=0.2075, simple_loss=0.2863, pruned_loss=0.06435, over 16641.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2975, pruned_loss=0.06607, over 3079144.04 frames. ], batch size: 62, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:50:45,362 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:50:48,321 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128320.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:51:39,461 INFO [train.py:904] (5/8) Epoch 13, batch 6550, loss[loss=0.2021, simple_loss=0.3032, pruned_loss=0.05049, over 16300.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2994, pruned_loss=0.06634, over 3093866.04 frames. ], batch size: 146, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:51:50,583 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-29 19:52:01,658 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:22,964 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.847e+02 3.449e+02 4.250e+02 8.173e+02, threshold=6.898e+02, percent-clipped=4.0 2023-04-29 19:52:25,025 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:56,150 INFO [train.py:904] (5/8) Epoch 13, batch 6600, loss[loss=0.2276, simple_loss=0.3109, pruned_loss=0.07213, over 16919.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.302, pruned_loss=0.06723, over 3078638.43 frames. ], batch size: 96, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:53:14,134 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:53:48,834 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:54:01,568 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0076, 4.9560, 4.7675, 4.1211, 4.8732, 1.8030, 4.6095, 4.5743], device='cuda:5'), covar=tensor([0.0066, 0.0058, 0.0152, 0.0345, 0.0070, 0.2512, 0.0109, 0.0184], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0125, 0.0170, 0.0161, 0.0142, 0.0184, 0.0158, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 19:54:14,495 INFO [train.py:904] (5/8) Epoch 13, batch 6650, loss[loss=0.2657, simple_loss=0.329, pruned_loss=0.1012, over 11492.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3025, pruned_loss=0.06864, over 3063609.47 frames. ], batch size: 248, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:54:47,954 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:54:56,952 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 3.058e+02 3.713e+02 4.648e+02 8.477e+02, threshold=7.425e+02, percent-clipped=4.0 2023-04-29 19:55:02,267 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:55:30,395 INFO [train.py:904] (5/8) Epoch 13, batch 6700, loss[loss=0.2441, simple_loss=0.3058, pruned_loss=0.09119, over 11360.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3016, pruned_loss=0.06893, over 3047428.08 frames. ], batch size: 247, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:55:36,463 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:10,004 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:46,052 INFO [train.py:904] (5/8) Epoch 13, batch 6750, loss[loss=0.227, simple_loss=0.3023, pruned_loss=0.07587, over 16655.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2998, pruned_loss=0.0683, over 3069261.34 frames. ], batch size: 62, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:56:49,450 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:57:22,974 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:57:28,179 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.101e+02 3.923e+02 4.752e+02 1.055e+03, threshold=7.847e+02, percent-clipped=2.0 2023-04-29 19:58:01,822 INFO [train.py:904] (5/8) Epoch 13, batch 6800, loss[loss=0.212, simple_loss=0.3034, pruned_loss=0.06028, over 16745.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2996, pruned_loss=0.06776, over 3080370.88 frames. ], batch size: 124, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:58:13,943 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 19:58:48,817 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-29 19:58:52,258 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6970, 3.6961, 2.1420, 4.2655, 2.7965, 4.1455, 2.2611, 2.9328], device='cuda:5'), covar=tensor([0.0199, 0.0333, 0.1570, 0.0158, 0.0717, 0.0482, 0.1504, 0.0683], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0167, 0.0189, 0.0137, 0.0168, 0.0207, 0.0196, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 19:59:18,070 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-04-29 19:59:19,108 INFO [train.py:904] (5/8) Epoch 13, batch 6850, loss[loss=0.2669, simple_loss=0.3243, pruned_loss=0.1047, over 11560.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3005, pruned_loss=0.06824, over 3061909.36 frames. ], batch size: 247, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:59:46,279 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7744, 5.0343, 5.2574, 4.9651, 4.9935, 5.5944, 5.1210, 4.8762], device='cuda:5'), covar=tensor([0.1020, 0.1763, 0.2197, 0.2291, 0.2513, 0.0935, 0.1359, 0.2285], device='cuda:5'), in_proj_covar=tensor([0.0366, 0.0517, 0.0561, 0.0439, 0.0596, 0.0583, 0.0449, 0.0595], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 19:59:55,667 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:00:00,832 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.804e+02 2.799e+02 3.352e+02 4.100e+02 8.801e+02, threshold=6.704e+02, percent-clipped=1.0 2023-04-29 20:00:26,358 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 20:00:34,562 INFO [train.py:904] (5/8) Epoch 13, batch 6900, loss[loss=0.2336, simple_loss=0.3135, pruned_loss=0.07682, over 16427.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3027, pruned_loss=0.06765, over 3080286.61 frames. ], batch size: 146, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:01:52,397 INFO [train.py:904] (5/8) Epoch 13, batch 6950, loss[loss=0.2165, simple_loss=0.3018, pruned_loss=0.06566, over 16797.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3048, pruned_loss=0.06955, over 3070633.82 frames. ], batch size: 124, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:02:20,025 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:02:20,471 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 20:02:36,170 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.242e+02 3.339e+02 3.931e+02 4.735e+02 7.907e+02, threshold=7.862e+02, percent-clipped=3.0 2023-04-29 20:02:38,468 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:03:08,760 INFO [train.py:904] (5/8) Epoch 13, batch 7000, loss[loss=0.2764, simple_loss=0.3324, pruned_loss=0.1102, over 11632.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3046, pruned_loss=0.06834, over 3077981.17 frames. ], batch size: 247, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:04:07,305 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:04:20,572 INFO [train.py:904] (5/8) Epoch 13, batch 7050, loss[loss=0.2745, simple_loss=0.3336, pruned_loss=0.1077, over 11667.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3053, pruned_loss=0.06801, over 3093081.12 frames. ], batch size: 247, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:05:02,605 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 2.845e+02 3.584e+02 4.333e+02 9.135e+02, threshold=7.167e+02, percent-clipped=3.0 2023-04-29 20:05:19,907 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128890.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:05:38,306 INFO [train.py:904] (5/8) Epoch 13, batch 7100, loss[loss=0.1949, simple_loss=0.2788, pruned_loss=0.05555, over 17137.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3038, pruned_loss=0.06771, over 3091817.10 frames. ], batch size: 48, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:06:04,518 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 20:06:31,854 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-29 20:06:56,864 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:06:57,472 INFO [train.py:904] (5/8) Epoch 13, batch 7150, loss[loss=0.2017, simple_loss=0.2908, pruned_loss=0.05628, over 16395.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3017, pruned_loss=0.06708, over 3095487.80 frames. ], batch size: 68, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:07:34,077 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128976.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:07:39,436 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 3.228e+02 3.806e+02 5.170e+02 8.621e+02, threshold=7.612e+02, percent-clipped=3.0 2023-04-29 20:07:45,276 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5498, 2.5247, 2.9345, 4.0532, 3.1989, 4.0177, 1.1782, 3.2390], device='cuda:5'), covar=tensor([0.1398, 0.0738, 0.0896, 0.0131, 0.0244, 0.0322, 0.1710, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0164, 0.0184, 0.0157, 0.0204, 0.0209, 0.0186, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 20:07:46,926 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2831, 4.2167, 4.1716, 2.9357, 4.1582, 1.4808, 3.8630, 3.6691], device='cuda:5'), covar=tensor([0.0145, 0.0140, 0.0201, 0.0712, 0.0150, 0.3405, 0.0223, 0.0419], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0124, 0.0170, 0.0160, 0.0142, 0.0185, 0.0158, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:08:12,092 INFO [train.py:904] (5/8) Epoch 13, batch 7200, loss[loss=0.1997, simple_loss=0.2897, pruned_loss=0.05487, over 16421.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2996, pruned_loss=0.06531, over 3100264.81 frames. ], batch size: 146, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:08:45,130 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:09:08,258 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6673, 2.5878, 2.2508, 4.2162, 2.9475, 3.8966, 1.3474, 2.7169], device='cuda:5'), covar=tensor([0.1377, 0.0719, 0.1301, 0.0165, 0.0261, 0.0408, 0.1698, 0.0847], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0164, 0.0184, 0.0157, 0.0205, 0.0209, 0.0187, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 20:09:29,940 INFO [train.py:904] (5/8) Epoch 13, batch 7250, loss[loss=0.1905, simple_loss=0.2716, pruned_loss=0.05465, over 16582.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2969, pruned_loss=0.06378, over 3103064.59 frames. ], batch size: 68, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:09:52,920 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 20:09:56,959 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:10:12,186 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.885e+02 3.444e+02 4.352e+02 7.294e+02, threshold=6.888e+02, percent-clipped=0.0 2023-04-29 20:10:45,720 INFO [train.py:904] (5/8) Epoch 13, batch 7300, loss[loss=0.2287, simple_loss=0.3161, pruned_loss=0.07066, over 16443.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2972, pruned_loss=0.06421, over 3088159.60 frames. ], batch size: 146, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:11:09,715 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:11:40,207 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129137.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:12:02,357 INFO [train.py:904] (5/8) Epoch 13, batch 7350, loss[loss=0.2515, simple_loss=0.3145, pruned_loss=0.09425, over 11300.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2978, pruned_loss=0.06547, over 3067328.19 frames. ], batch size: 247, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:12:46,225 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.924e+02 3.417e+02 4.096e+02 1.600e+03, threshold=6.834e+02, percent-clipped=4.0 2023-04-29 20:13:21,046 INFO [train.py:904] (5/8) Epoch 13, batch 7400, loss[loss=0.2112, simple_loss=0.2992, pruned_loss=0.06161, over 16709.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2993, pruned_loss=0.06629, over 3075174.99 frames. ], batch size: 124, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:14:15,547 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6895, 4.7322, 4.5449, 4.2265, 4.1393, 4.5891, 4.4824, 4.2988], device='cuda:5'), covar=tensor([0.0662, 0.0469, 0.0301, 0.0317, 0.1005, 0.0458, 0.0485, 0.0762], device='cuda:5'), in_proj_covar=tensor([0.0242, 0.0330, 0.0291, 0.0271, 0.0308, 0.0313, 0.0198, 0.0338], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:14:32,927 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:14:41,695 INFO [train.py:904] (5/8) Epoch 13, batch 7450, loss[loss=0.1914, simple_loss=0.2893, pruned_loss=0.04675, over 16803.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.299, pruned_loss=0.06572, over 3105008.15 frames. ], batch size: 83, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:15:30,918 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 3.121e+02 3.842e+02 4.432e+02 7.351e+02, threshold=7.685e+02, percent-clipped=1.0 2023-04-29 20:16:03,378 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 20:16:05,639 INFO [train.py:904] (5/8) Epoch 13, batch 7500, loss[loss=0.2102, simple_loss=0.2931, pruned_loss=0.06371, over 16882.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2999, pruned_loss=0.06555, over 3098505.34 frames. ], batch size: 109, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:16:19,467 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 20:17:24,568 INFO [train.py:904] (5/8) Epoch 13, batch 7550, loss[loss=0.2165, simple_loss=0.3017, pruned_loss=0.06565, over 16204.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2994, pruned_loss=0.0662, over 3075233.55 frames. ], batch size: 165, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:17:43,651 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3827, 5.7073, 5.3990, 5.4337, 5.1597, 5.0142, 5.1385, 5.7774], device='cuda:5'), covar=tensor([0.1072, 0.0783, 0.1033, 0.0807, 0.0790, 0.0759, 0.1017, 0.0879], device='cuda:5'), in_proj_covar=tensor([0.0568, 0.0702, 0.0576, 0.0500, 0.0442, 0.0455, 0.0587, 0.0538], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:17:58,064 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 20:18:07,708 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.820e+02 3.717e+02 4.924e+02 9.310e+02, threshold=7.434e+02, percent-clipped=3.0 2023-04-29 20:18:41,443 INFO [train.py:904] (5/8) Epoch 13, batch 7600, loss[loss=0.1986, simple_loss=0.2889, pruned_loss=0.05413, over 16226.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2989, pruned_loss=0.0668, over 3071985.19 frames. ], batch size: 165, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:18:59,703 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:19:37,559 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:19:41,501 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-29 20:19:43,626 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.41 vs. limit=5.0 2023-04-29 20:20:00,027 INFO [train.py:904] (5/8) Epoch 13, batch 7650, loss[loss=0.2064, simple_loss=0.2947, pruned_loss=0.05907, over 16435.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2992, pruned_loss=0.0674, over 3073227.59 frames. ], batch size: 146, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:20:12,416 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 20:20:36,420 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:20:44,166 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.211e+02 3.428e+02 4.160e+02 4.831e+02 1.059e+03, threshold=8.320e+02, percent-clipped=3.0 2023-04-29 20:20:53,101 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:21:06,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8640, 4.1288, 3.9404, 3.9675, 3.6685, 3.8026, 3.8383, 4.0963], device='cuda:5'), covar=tensor([0.1017, 0.0886, 0.1008, 0.0732, 0.0793, 0.1545, 0.0844, 0.1049], device='cuda:5'), in_proj_covar=tensor([0.0568, 0.0702, 0.0576, 0.0499, 0.0441, 0.0455, 0.0587, 0.0536], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:21:18,584 INFO [train.py:904] (5/8) Epoch 13, batch 7700, loss[loss=0.1989, simple_loss=0.2871, pruned_loss=0.05528, over 16738.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2996, pruned_loss=0.06831, over 3052123.28 frames. ], batch size: 89, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:21:54,407 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5863, 1.7101, 2.2087, 2.6083, 2.5462, 2.9796, 1.7258, 2.9086], device='cuda:5'), covar=tensor([0.0196, 0.0418, 0.0264, 0.0238, 0.0245, 0.0146, 0.0445, 0.0110], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0162, 0.0173, 0.0128, 0.0175, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 20:22:26,949 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:22:35,891 INFO [train.py:904] (5/8) Epoch 13, batch 7750, loss[loss=0.2635, simple_loss=0.3184, pruned_loss=0.1043, over 11362.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2998, pruned_loss=0.06813, over 3054765.81 frames. ], batch size: 248, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:23:20,360 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 3.197e+02 3.787e+02 4.587e+02 8.661e+02, threshold=7.574e+02, percent-clipped=1.0 2023-04-29 20:23:40,230 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:23:52,184 INFO [train.py:904] (5/8) Epoch 13, batch 7800, loss[loss=0.2079, simple_loss=0.299, pruned_loss=0.05844, over 16302.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3014, pruned_loss=0.06934, over 3051852.10 frames. ], batch size: 165, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:24:09,348 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129612.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:24:48,758 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2096, 4.2346, 4.4287, 4.2029, 4.3150, 4.7964, 4.4203, 4.1635], device='cuda:5'), covar=tensor([0.1698, 0.3683, 0.2536, 0.3225, 0.3013, 0.1277, 0.1882, 0.3392], device='cuda:5'), in_proj_covar=tensor([0.0366, 0.0519, 0.0567, 0.0441, 0.0597, 0.0584, 0.0450, 0.0593], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 20:25:09,861 INFO [train.py:904] (5/8) Epoch 13, batch 7850, loss[loss=0.2022, simple_loss=0.2927, pruned_loss=0.05587, over 16298.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3014, pruned_loss=0.06849, over 3060030.96 frames. ], batch size: 146, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:25:26,668 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7612, 3.6097, 3.8586, 3.6614, 3.7807, 4.2007, 3.8827, 3.5844], device='cuda:5'), covar=tensor([0.2065, 0.2612, 0.2612, 0.2451, 0.3109, 0.1752, 0.1614, 0.2657], device='cuda:5'), in_proj_covar=tensor([0.0366, 0.0518, 0.0566, 0.0441, 0.0597, 0.0584, 0.0450, 0.0593], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 20:25:40,341 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:25:52,272 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 3.020e+02 3.492e+02 4.287e+02 1.158e+03, threshold=6.983e+02, percent-clipped=4.0 2023-04-29 20:26:02,605 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5203, 5.8467, 5.5490, 5.6486, 5.2318, 5.1375, 5.3126, 5.9354], device='cuda:5'), covar=tensor([0.1056, 0.0769, 0.0943, 0.0751, 0.0844, 0.0664, 0.1029, 0.0805], device='cuda:5'), in_proj_covar=tensor([0.0570, 0.0703, 0.0580, 0.0502, 0.0442, 0.0456, 0.0588, 0.0538], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:26:22,708 INFO [train.py:904] (5/8) Epoch 13, batch 7900, loss[loss=0.25, simple_loss=0.3182, pruned_loss=0.0909, over 11454.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3, pruned_loss=0.06767, over 3061559.88 frames. ], batch size: 247, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:27:36,781 INFO [train.py:904] (5/8) Epoch 13, batch 7950, loss[loss=0.2359, simple_loss=0.3107, pruned_loss=0.08061, over 15286.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3005, pruned_loss=0.06796, over 3062852.77 frames. ], batch size: 190, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:28:02,256 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:28:18,211 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.009e+02 2.840e+02 3.431e+02 4.053e+02 8.396e+02, threshold=6.863e+02, percent-clipped=2.0 2023-04-29 20:28:33,887 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129791.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:28:49,447 INFO [train.py:904] (5/8) Epoch 13, batch 8000, loss[loss=0.1864, simple_loss=0.2826, pruned_loss=0.04509, over 16877.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3001, pruned_loss=0.06781, over 3064158.19 frames. ], batch size: 102, lr: 5.19e-03, grad_scale: 8.0 2023-04-29 20:30:02,343 INFO [train.py:904] (5/8) Epoch 13, batch 8050, loss[loss=0.2135, simple_loss=0.3025, pruned_loss=0.06222, over 16716.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.3007, pruned_loss=0.06807, over 3056289.76 frames. ], batch size: 124, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:30:02,913 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:30:25,174 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 20:30:37,803 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 20:30:45,833 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.353e+02 3.065e+02 3.748e+02 4.862e+02 1.232e+03, threshold=7.497e+02, percent-clipped=5.0 2023-04-29 20:30:53,712 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3115, 3.2231, 3.4815, 1.7428, 3.7003, 3.7440, 2.7723, 2.7572], device='cuda:5'), covar=tensor([0.0786, 0.0227, 0.0190, 0.1181, 0.0057, 0.0141, 0.0434, 0.0459], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0138, 0.0069, 0.0109, 0.0121, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 20:31:15,198 INFO [train.py:904] (5/8) Epoch 13, batch 8100, loss[loss=0.1966, simple_loss=0.275, pruned_loss=0.05913, over 16302.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2995, pruned_loss=0.06713, over 3067113.04 frames. ], batch size: 35, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:29,507 INFO [train.py:904] (5/8) Epoch 13, batch 8150, loss[loss=0.1699, simple_loss=0.2542, pruned_loss=0.04276, over 16712.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.298, pruned_loss=0.06665, over 3070470.51 frames. ], batch size: 83, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:53,005 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129968.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:14,450 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.827e+02 2.878e+02 3.437e+02 4.165e+02 9.532e+02, threshold=6.873e+02, percent-clipped=2.0 2023-04-29 20:33:46,563 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130000.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:48,631 INFO [train.py:904] (5/8) Epoch 13, batch 8200, loss[loss=0.2206, simple_loss=0.2921, pruned_loss=0.07459, over 11608.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2952, pruned_loss=0.0658, over 3080346.18 frames. ], batch size: 247, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:33:49,162 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:34:58,959 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2980, 4.3008, 4.1429, 3.8552, 3.7690, 4.2178, 3.9913, 3.9195], device='cuda:5'), covar=tensor([0.0542, 0.0446, 0.0312, 0.0303, 0.1005, 0.0410, 0.0651, 0.0691], device='cuda:5'), in_proj_covar=tensor([0.0245, 0.0338, 0.0295, 0.0273, 0.0310, 0.0316, 0.0201, 0.0342], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:35:09,259 INFO [train.py:904] (5/8) Epoch 13, batch 8250, loss[loss=0.187, simple_loss=0.2699, pruned_loss=0.05212, over 12180.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2942, pruned_loss=0.06379, over 3057173.32 frames. ], batch size: 248, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:35:24,097 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:35:27,856 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:35:37,699 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130069.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:35:57,082 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.705e+02 3.349e+02 4.037e+02 8.257e+02, threshold=6.697e+02, percent-clipped=3.0 2023-04-29 20:36:29,992 INFO [train.py:904] (5/8) Epoch 13, batch 8300, loss[loss=0.1982, simple_loss=0.2927, pruned_loss=0.05179, over 16317.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2915, pruned_loss=0.06025, over 3067357.62 frames. ], batch size: 146, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:36:55,890 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:37:43,814 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:37:51,488 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5266, 2.1038, 2.0866, 4.0418, 1.9953, 2.6834, 2.2058, 2.2681], device='cuda:5'), covar=tensor([0.0825, 0.3492, 0.2591, 0.0369, 0.4116, 0.2084, 0.3358, 0.3121], device='cuda:5'), in_proj_covar=tensor([0.0362, 0.0398, 0.0332, 0.0312, 0.0411, 0.0454, 0.0361, 0.0462], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:37:51,999 INFO [train.py:904] (5/8) Epoch 13, batch 8350, loss[loss=0.1998, simple_loss=0.2972, pruned_loss=0.05118, over 16734.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2911, pruned_loss=0.05887, over 3048158.16 frames. ], batch size: 124, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:38:39,967 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.345e+02 2.793e+02 3.218e+02 5.548e+02, threshold=5.587e+02, percent-clipped=0.0 2023-04-29 20:39:12,434 INFO [train.py:904] (5/8) Epoch 13, batch 8400, loss[loss=0.1803, simple_loss=0.2724, pruned_loss=0.04414, over 16422.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2879, pruned_loss=0.05593, over 3050768.00 frames. ], batch size: 146, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:29,232 INFO [train.py:904] (5/8) Epoch 13, batch 8450, loss[loss=0.1749, simple_loss=0.2703, pruned_loss=0.03974, over 15302.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2859, pruned_loss=0.05435, over 3053820.44 frames. ], batch size: 190, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:55,921 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:40:59,903 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5806, 2.1903, 2.2538, 4.3899, 2.1160, 2.7316, 2.2981, 2.4102], device='cuda:5'), covar=tensor([0.0925, 0.3740, 0.2584, 0.0344, 0.4203, 0.2308, 0.3409, 0.3319], device='cuda:5'), in_proj_covar=tensor([0.0359, 0.0395, 0.0329, 0.0309, 0.0408, 0.0450, 0.0359, 0.0458], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:41:17,436 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.422e+02 2.811e+02 3.625e+02 1.309e+03, threshold=5.622e+02, percent-clipped=6.0 2023-04-29 20:41:49,398 INFO [train.py:904] (5/8) Epoch 13, batch 8500, loss[loss=0.1736, simple_loss=0.2491, pruned_loss=0.04899, over 11581.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2814, pruned_loss=0.05161, over 3043691.09 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:42:13,190 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:43:10,849 INFO [train.py:904] (5/8) Epoch 13, batch 8550, loss[loss=0.2211, simple_loss=0.3085, pruned_loss=0.06687, over 16767.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2793, pruned_loss=0.05072, over 3041099.25 frames. ], batch size: 124, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:43:19,919 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:43:22,007 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0944, 3.3537, 3.3624, 2.2575, 3.1257, 3.3953, 3.2498, 1.8519], device='cuda:5'), covar=tensor([0.0448, 0.0038, 0.0039, 0.0336, 0.0073, 0.0063, 0.0064, 0.0447], device='cuda:5'), in_proj_covar=tensor([0.0128, 0.0068, 0.0070, 0.0125, 0.0082, 0.0090, 0.0080, 0.0119], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 20:43:23,663 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130358.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:44:07,712 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.204e+02 2.928e+02 3.716e+02 7.366e+02, threshold=5.857e+02, percent-clipped=6.0 2023-04-29 20:44:12,826 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:44:43,262 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6125, 3.1246, 3.1982, 1.8947, 2.7623, 2.1610, 3.0234, 3.3040], device='cuda:5'), covar=tensor([0.0398, 0.0814, 0.0508, 0.1901, 0.0807, 0.0975, 0.0894, 0.0881], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0144, 0.0155, 0.0142, 0.0135, 0.0123, 0.0134, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 20:44:50,019 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0984, 2.0423, 2.1694, 3.5580, 2.0229, 2.4106, 2.1556, 2.2104], device='cuda:5'), covar=tensor([0.1089, 0.3719, 0.2577, 0.0516, 0.4186, 0.2383, 0.3391, 0.3333], device='cuda:5'), in_proj_covar=tensor([0.0360, 0.0396, 0.0330, 0.0309, 0.0410, 0.0452, 0.0360, 0.0459], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:44:50,605 INFO [train.py:904] (5/8) Epoch 13, batch 8600, loss[loss=0.2007, simple_loss=0.2888, pruned_loss=0.05633, over 15114.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2803, pruned_loss=0.04995, over 3053956.56 frames. ], batch size: 190, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:45:44,267 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2782, 1.5920, 1.9799, 2.3215, 2.3091, 2.5548, 1.7294, 2.4913], device='cuda:5'), covar=tensor([0.0174, 0.0434, 0.0263, 0.0241, 0.0254, 0.0170, 0.0406, 0.0121], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0171, 0.0155, 0.0157, 0.0168, 0.0126, 0.0171, 0.0116], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 20:45:51,721 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:04,172 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:04,458 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-04-29 20:46:16,957 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:20,657 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130447.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:30,004 INFO [train.py:904] (5/8) Epoch 13, batch 8650, loss[loss=0.1909, simple_loss=0.2855, pruned_loss=0.04821, over 16331.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2781, pruned_loss=0.04797, over 3070608.26 frames. ], batch size: 146, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:47:40,991 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.301e+02 2.678e+02 3.280e+02 8.282e+02, threshold=5.356e+02, percent-clipped=3.0 2023-04-29 20:47:45,113 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8386, 3.2099, 2.5744, 4.5146, 3.1752, 4.1647, 1.6912, 2.8622], device='cuda:5'), covar=tensor([0.1191, 0.0508, 0.1037, 0.0173, 0.0112, 0.0358, 0.1384, 0.0694], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0162, 0.0182, 0.0154, 0.0198, 0.0206, 0.0185, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 20:47:46,557 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1587, 3.9710, 4.2056, 4.3077, 4.4807, 4.0304, 4.4609, 4.4508], device='cuda:5'), covar=tensor([0.1508, 0.1183, 0.1334, 0.0689, 0.0520, 0.1223, 0.0535, 0.0630], device='cuda:5'), in_proj_covar=tensor([0.0524, 0.0648, 0.0772, 0.0664, 0.0505, 0.0512, 0.0528, 0.0617], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:48:01,205 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:05,354 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:12,929 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130499.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:17,496 INFO [train.py:904] (5/8) Epoch 13, batch 8700, loss[loss=0.1693, simple_loss=0.268, pruned_loss=0.03533, over 16738.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2753, pruned_loss=0.04649, over 3073773.52 frames. ], batch size: 76, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:48:39,605 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.65 vs. limit=5.0 2023-04-29 20:49:01,326 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7079, 1.9885, 1.7252, 1.8774, 2.3617, 2.0629, 2.4094, 2.5692], device='cuda:5'), covar=tensor([0.0134, 0.0394, 0.0470, 0.0420, 0.0269, 0.0345, 0.0166, 0.0226], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0203, 0.0197, 0.0198, 0.0202, 0.0201, 0.0202, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:49:47,064 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130548.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:49:54,316 INFO [train.py:904] (5/8) Epoch 13, batch 8750, loss[loss=0.1663, simple_loss=0.2529, pruned_loss=0.03982, over 12005.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2746, pruned_loss=0.04575, over 3061598.49 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:50:30,270 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:51:07,897 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.544e+02 3.252e+02 4.020e+02 9.168e+02, threshold=6.504e+02, percent-clipped=9.0 2023-04-29 20:51:37,544 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4022, 3.0608, 2.6709, 2.1707, 2.1982, 2.1803, 3.0741, 2.8617], device='cuda:5'), covar=tensor([0.2491, 0.0725, 0.1491, 0.2231, 0.2218, 0.1897, 0.0454, 0.1090], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0253, 0.0284, 0.0280, 0.0270, 0.0225, 0.0267, 0.0294], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:51:39,437 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2803, 2.3682, 2.0245, 2.1970, 2.7571, 2.4696, 2.9819, 2.9828], device='cuda:5'), covar=tensor([0.0100, 0.0338, 0.0407, 0.0344, 0.0237, 0.0300, 0.0172, 0.0207], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0204, 0.0198, 0.0198, 0.0202, 0.0202, 0.0202, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:51:48,217 INFO [train.py:904] (5/8) Epoch 13, batch 8800, loss[loss=0.1721, simple_loss=0.266, pruned_loss=0.03914, over 16299.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2726, pruned_loss=0.04459, over 3048862.34 frames. ], batch size: 146, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:51:55,882 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8087, 1.2967, 1.6927, 1.7336, 1.7795, 1.8376, 1.6419, 1.8060], device='cuda:5'), covar=tensor([0.0235, 0.0362, 0.0193, 0.0225, 0.0285, 0.0181, 0.0327, 0.0101], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0170, 0.0154, 0.0156, 0.0167, 0.0125, 0.0170, 0.0115], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 20:52:02,504 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:52:39,868 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130626.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:52:58,221 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 20:53:32,472 INFO [train.py:904] (5/8) Epoch 13, batch 8850, loss[loss=0.1738, simple_loss=0.2755, pruned_loss=0.03609, over 15310.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2744, pruned_loss=0.04393, over 3035472.69 frames. ], batch size: 191, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:53:41,566 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:53:45,886 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:54:37,532 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.288e+02 2.852e+02 3.513e+02 7.607e+02, threshold=5.705e+02, percent-clipped=2.0 2023-04-29 20:55:17,119 INFO [train.py:904] (5/8) Epoch 13, batch 8900, loss[loss=0.1887, simple_loss=0.2847, pruned_loss=0.04633, over 12661.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2751, pruned_loss=0.04354, over 3041717.51 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:55:19,549 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 20:55:22,780 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:55:26,793 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:56:03,909 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2075, 3.3126, 1.8839, 3.5027, 2.4361, 3.4568, 2.0541, 2.6612], device='cuda:5'), covar=tensor([0.0228, 0.0295, 0.1547, 0.0166, 0.0749, 0.0592, 0.1490, 0.0663], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0159, 0.0184, 0.0129, 0.0163, 0.0198, 0.0191, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 20:56:42,458 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2128, 4.0662, 4.2801, 4.3750, 4.5442, 4.1008, 4.5435, 4.5529], device='cuda:5'), covar=tensor([0.1544, 0.1122, 0.1511, 0.0731, 0.0546, 0.1087, 0.0485, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0522, 0.0646, 0.0770, 0.0662, 0.0502, 0.0508, 0.0527, 0.0609], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:56:52,528 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-29 20:56:54,229 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:57:07,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5652, 1.9503, 1.6329, 1.7945, 2.3718, 2.0226, 2.2708, 2.4596], device='cuda:5'), covar=tensor([0.0119, 0.0406, 0.0459, 0.0403, 0.0222, 0.0352, 0.0158, 0.0240], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0206, 0.0200, 0.0199, 0.0203, 0.0203, 0.0203, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:57:21,729 INFO [train.py:904] (5/8) Epoch 13, batch 8950, loss[loss=0.1759, simple_loss=0.2642, pruned_loss=0.04382, over 17239.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2752, pruned_loss=0.04408, over 3055113.60 frames. ], batch size: 52, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:57:31,869 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5274, 1.9600, 1.6018, 1.7841, 2.3398, 1.9979, 2.1707, 2.4609], device='cuda:5'), covar=tensor([0.0129, 0.0388, 0.0511, 0.0447, 0.0247, 0.0357, 0.0180, 0.0230], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0206, 0.0200, 0.0199, 0.0204, 0.0204, 0.0203, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 20:58:29,463 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.201e+02 2.684e+02 3.167e+02 7.959e+02, threshold=5.368e+02, percent-clipped=1.0 2023-04-29 20:58:41,707 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:58:54,906 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:59:03,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2908, 3.3980, 3.6498, 3.6238, 3.6339, 3.4404, 3.5016, 3.5210], device='cuda:5'), covar=tensor([0.0379, 0.0690, 0.0466, 0.0473, 0.0519, 0.0501, 0.0719, 0.0417], device='cuda:5'), in_proj_covar=tensor([0.0338, 0.0351, 0.0354, 0.0340, 0.0398, 0.0377, 0.0463, 0.0301], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 20:59:11,372 INFO [train.py:904] (5/8) Epoch 13, batch 9000, loss[loss=0.1759, simple_loss=0.2582, pruned_loss=0.04684, over 11949.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.272, pruned_loss=0.04272, over 3057568.83 frames. ], batch size: 250, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:59:11,373 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 20:59:22,057 INFO [train.py:938] (5/8) Epoch 13, validation: loss=0.1517, simple_loss=0.2561, pruned_loss=0.02371, over 944034.00 frames. 2023-04-29 20:59:22,058 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 20:59:52,662 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 21:01:06,018 INFO [train.py:904] (5/8) Epoch 13, batch 9050, loss[loss=0.162, simple_loss=0.2569, pruned_loss=0.03353, over 16677.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2736, pruned_loss=0.04358, over 3060531.83 frames. ], batch size: 89, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:01:48,416 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8299, 3.8353, 3.9829, 3.7926, 3.8976, 4.3062, 3.9762, 3.6786], device='cuda:5'), covar=tensor([0.2061, 0.2100, 0.1799, 0.2402, 0.2684, 0.1479, 0.1515, 0.2690], device='cuda:5'), in_proj_covar=tensor([0.0347, 0.0491, 0.0534, 0.0420, 0.0565, 0.0564, 0.0430, 0.0562], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 21:02:07,080 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.403e+02 2.798e+02 3.395e+02 5.131e+02, threshold=5.596e+02, percent-clipped=0.0 2023-04-29 21:02:46,032 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 21:02:52,487 INFO [train.py:904] (5/8) Epoch 13, batch 9100, loss[loss=0.1578, simple_loss=0.247, pruned_loss=0.0343, over 12471.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.273, pruned_loss=0.04387, over 3073037.70 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:58,110 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:03:33,391 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130921.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:04:49,579 INFO [train.py:904] (5/8) Epoch 13, batch 9150, loss[loss=0.1594, simple_loss=0.2474, pruned_loss=0.03566, over 11983.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2725, pruned_loss=0.0432, over 3049448.36 frames. ], batch size: 250, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:05:52,921 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.726e+02 2.558e+02 2.960e+02 3.606e+02 6.781e+02, threshold=5.920e+02, percent-clipped=4.0 2023-04-29 21:06:27,257 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5512, 4.5921, 4.4112, 4.0978, 4.1131, 4.5243, 4.3291, 4.1471], device='cuda:5'), covar=tensor([0.0492, 0.0420, 0.0256, 0.0239, 0.0741, 0.0373, 0.0415, 0.0596], device='cuda:5'), in_proj_covar=tensor([0.0239, 0.0326, 0.0289, 0.0267, 0.0299, 0.0309, 0.0197, 0.0334], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:06:31,938 INFO [train.py:904] (5/8) Epoch 13, batch 9200, loss[loss=0.1833, simple_loss=0.2736, pruned_loss=0.04656, over 15189.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2684, pruned_loss=0.04252, over 3057248.41 frames. ], batch size: 191, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:06:47,121 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7001, 2.7321, 2.2966, 2.5110, 3.1204, 2.8234, 3.4248, 3.3134], device='cuda:5'), covar=tensor([0.0070, 0.0310, 0.0387, 0.0361, 0.0215, 0.0293, 0.0162, 0.0187], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0207, 0.0201, 0.0201, 0.0204, 0.0204, 0.0202, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:07:31,185 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0992, 2.0176, 2.1907, 3.5780, 2.0343, 2.3410, 2.2176, 2.1476], device='cuda:5'), covar=tensor([0.0985, 0.3586, 0.2529, 0.0479, 0.4051, 0.2421, 0.3175, 0.3391], device='cuda:5'), in_proj_covar=tensor([0.0359, 0.0395, 0.0331, 0.0309, 0.0409, 0.0449, 0.0360, 0.0458], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:07:43,486 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:08:07,059 INFO [train.py:904] (5/8) Epoch 13, batch 9250, loss[loss=0.1619, simple_loss=0.2426, pruned_loss=0.04055, over 12114.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2678, pruned_loss=0.04224, over 3063534.69 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:08:34,752 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6988, 4.8592, 4.6147, 4.2347, 4.0532, 4.7554, 4.6427, 4.3461], device='cuda:5'), covar=tensor([0.0593, 0.0458, 0.0372, 0.0313, 0.1139, 0.0420, 0.0368, 0.0655], device='cuda:5'), in_proj_covar=tensor([0.0237, 0.0325, 0.0288, 0.0266, 0.0297, 0.0309, 0.0196, 0.0332], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:09:12,851 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 2.375e+02 2.887e+02 3.471e+02 9.563e+02, threshold=5.774e+02, percent-clipped=3.0 2023-04-29 21:09:24,456 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:24,499 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:40,200 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:55,055 INFO [train.py:904] (5/8) Epoch 13, batch 9300, loss[loss=0.158, simple_loss=0.2421, pruned_loss=0.03691, over 12142.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2664, pruned_loss=0.04143, over 3075462.65 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:10:50,618 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7958, 3.6859, 3.8770, 3.9650, 4.0679, 3.6526, 4.0364, 4.0925], device='cuda:5'), covar=tensor([0.1468, 0.1024, 0.1178, 0.0660, 0.0533, 0.1564, 0.0610, 0.0562], device='cuda:5'), in_proj_covar=tensor([0.0522, 0.0640, 0.0765, 0.0659, 0.0498, 0.0505, 0.0525, 0.0607], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:11:11,219 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:24,002 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:35,098 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3969, 3.3164, 3.5232, 1.7776, 3.7086, 3.8263, 2.8451, 2.8141], device='cuda:5'), covar=tensor([0.0798, 0.0222, 0.0196, 0.1213, 0.0063, 0.0115, 0.0415, 0.0446], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0097, 0.0083, 0.0132, 0.0065, 0.0103, 0.0116, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 21:11:40,568 INFO [train.py:904] (5/8) Epoch 13, batch 9350, loss[loss=0.1709, simple_loss=0.2548, pruned_loss=0.0435, over 12317.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2664, pruned_loss=0.04126, over 3095116.96 frames. ], batch size: 248, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:12:12,186 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 21:12:25,583 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131174.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:12:41,637 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.118e+02 2.389e+02 2.842e+02 4.349e+02, threshold=4.777e+02, percent-clipped=0.0 2023-04-29 21:13:20,701 INFO [train.py:904] (5/8) Epoch 13, batch 9400, loss[loss=0.1884, simple_loss=0.2831, pruned_loss=0.0469, over 16307.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2662, pruned_loss=0.04107, over 3077948.59 frames. ], batch size: 146, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:13:25,712 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 21:13:59,478 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:14:26,223 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 21:14:27,973 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:14:59,710 INFO [train.py:904] (5/8) Epoch 13, batch 9450, loss[loss=0.1777, simple_loss=0.2674, pruned_loss=0.04404, over 12496.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2679, pruned_loss=0.04149, over 3065550.26 frames. ], batch size: 250, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:15:00,864 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131252.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:15:34,990 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:16:03,424 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.351e+02 2.704e+02 3.479e+02 5.453e+02, threshold=5.408e+02, percent-clipped=2.0 2023-04-29 21:16:16,084 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3488, 4.3978, 4.7970, 4.7507, 4.7431, 4.4845, 4.4594, 4.3791], device='cuda:5'), covar=tensor([0.0317, 0.0493, 0.0349, 0.0462, 0.0448, 0.0339, 0.0751, 0.0377], device='cuda:5'), in_proj_covar=tensor([0.0328, 0.0341, 0.0343, 0.0330, 0.0387, 0.0368, 0.0445, 0.0292], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-29 21:16:36,331 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2985, 3.3564, 3.4809, 1.7548, 3.6425, 3.7382, 2.8090, 2.7304], device='cuda:5'), covar=tensor([0.0744, 0.0168, 0.0163, 0.1127, 0.0053, 0.0113, 0.0378, 0.0419], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0097, 0.0083, 0.0132, 0.0065, 0.0103, 0.0116, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 21:16:40,124 INFO [train.py:904] (5/8) Epoch 13, batch 9500, loss[loss=0.1704, simple_loss=0.2694, pruned_loss=0.03567, over 16895.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2674, pruned_loss=0.04119, over 3071656.12 frames. ], batch size: 96, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:17:00,424 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5694, 3.6081, 3.4342, 3.1428, 3.2640, 3.5515, 3.3417, 3.2990], device='cuda:5'), covar=tensor([0.0512, 0.0451, 0.0258, 0.0220, 0.0465, 0.0410, 0.0991, 0.0407], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0321, 0.0286, 0.0263, 0.0294, 0.0306, 0.0194, 0.0328], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-29 21:17:21,369 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9003, 4.6974, 4.9529, 5.1080, 5.3077, 4.7082, 5.2693, 5.3091], device='cuda:5'), covar=tensor([0.1592, 0.1101, 0.1449, 0.0683, 0.0468, 0.0638, 0.0455, 0.0581], device='cuda:5'), in_proj_covar=tensor([0.0518, 0.0634, 0.0754, 0.0650, 0.0490, 0.0500, 0.0520, 0.0601], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:17:30,145 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 21:18:25,261 INFO [train.py:904] (5/8) Epoch 13, batch 9550, loss[loss=0.1744, simple_loss=0.2655, pruned_loss=0.04171, over 12711.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.267, pruned_loss=0.04161, over 3045906.48 frames. ], batch size: 247, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:18:40,903 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5994, 3.6132, 3.4382, 3.1476, 3.2550, 3.5529, 3.3375, 3.3399], device='cuda:5'), covar=tensor([0.0488, 0.0558, 0.0271, 0.0231, 0.0512, 0.0410, 0.1236, 0.0456], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0323, 0.0287, 0.0264, 0.0295, 0.0307, 0.0195, 0.0330], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-29 21:19:19,721 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4706, 4.5294, 4.3233, 4.0466, 4.0371, 4.4297, 4.2129, 4.1151], device='cuda:5'), covar=tensor([0.0522, 0.0503, 0.0296, 0.0256, 0.0749, 0.0478, 0.0492, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0323, 0.0287, 0.0264, 0.0295, 0.0307, 0.0195, 0.0329], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-29 21:19:26,688 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-29 21:19:29,584 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.317e+02 2.711e+02 3.253e+02 6.482e+02, threshold=5.422e+02, percent-clipped=4.0 2023-04-29 21:20:03,679 INFO [train.py:904] (5/8) Epoch 13, batch 9600, loss[loss=0.1779, simple_loss=0.2813, pruned_loss=0.03724, over 16798.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2683, pruned_loss=0.04241, over 3016348.44 frames. ], batch size: 83, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:20:18,007 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:20:30,437 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:21:49,849 INFO [train.py:904] (5/8) Epoch 13, batch 9650, loss[loss=0.1719, simple_loss=0.2664, pruned_loss=0.03877, over 16969.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2698, pruned_loss=0.04258, over 3003092.33 frames. ], batch size: 109, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:22:34,310 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:22:48,246 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131477.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:22:58,034 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.434e+02 2.934e+02 3.737e+02 7.668e+02, threshold=5.867e+02, percent-clipped=4.0 2023-04-29 21:23:18,252 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131492.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:23:38,322 INFO [train.py:904] (5/8) Epoch 13, batch 9700, loss[loss=0.1785, simple_loss=0.2673, pruned_loss=0.04487, over 16669.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.269, pruned_loss=0.04226, over 3024724.79 frames. ], batch size: 134, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:24:36,250 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131530.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:25:20,244 INFO [train.py:904] (5/8) Epoch 13, batch 9750, loss[loss=0.1504, simple_loss=0.2457, pruned_loss=0.02753, over 16549.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.268, pruned_loss=0.0421, over 3041038.55 frames. ], batch size: 62, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:25:22,338 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:26:22,853 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.333e+02 2.960e+02 3.674e+02 6.318e+02, threshold=5.921e+02, percent-clipped=2.0 2023-04-29 21:26:31,008 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6753, 2.6716, 1.8573, 2.8544, 2.1810, 2.8255, 2.1053, 2.4110], device='cuda:5'), covar=tensor([0.0242, 0.0322, 0.1209, 0.0208, 0.0631, 0.0454, 0.1098, 0.0558], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0160, 0.0186, 0.0130, 0.0165, 0.0197, 0.0194, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-29 21:26:57,824 INFO [train.py:904] (5/8) Epoch 13, batch 9800, loss[loss=0.1597, simple_loss=0.2667, pruned_loss=0.02633, over 16893.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2679, pruned_loss=0.0413, over 3049162.37 frames. ], batch size: 96, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:27:30,278 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2966, 3.3937, 3.6491, 3.6111, 3.6363, 3.4312, 3.4748, 3.5053], device='cuda:5'), covar=tensor([0.0371, 0.0551, 0.0411, 0.0464, 0.0441, 0.0500, 0.0720, 0.0420], device='cuda:5'), in_proj_covar=tensor([0.0332, 0.0345, 0.0345, 0.0336, 0.0391, 0.0373, 0.0453, 0.0296], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 21:28:12,825 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 21:28:39,807 INFO [train.py:904] (5/8) Epoch 13, batch 9850, loss[loss=0.1597, simple_loss=0.261, pruned_loss=0.02917, over 16834.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2696, pruned_loss=0.04144, over 3055009.70 frames. ], batch size: 102, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:29:46,612 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.305e+02 2.925e+02 3.439e+02 6.228e+02, threshold=5.850e+02, percent-clipped=2.0 2023-04-29 21:30:32,565 INFO [train.py:904] (5/8) Epoch 13, batch 9900, loss[loss=0.1667, simple_loss=0.253, pruned_loss=0.04017, over 12516.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2705, pruned_loss=0.0419, over 3052519.41 frames. ], batch size: 248, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:32:30,680 INFO [train.py:904] (5/8) Epoch 13, batch 9950, loss[loss=0.1827, simple_loss=0.2795, pruned_loss=0.043, over 16681.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2727, pruned_loss=0.04221, over 3059978.67 frames. ], batch size: 134, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:33:02,018 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:04,634 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5363, 4.8448, 4.6207, 4.6306, 4.3084, 4.3190, 4.2813, 4.8729], device='cuda:5'), covar=tensor([0.1123, 0.0862, 0.0936, 0.0730, 0.0820, 0.1162, 0.1162, 0.0918], device='cuda:5'), in_proj_covar=tensor([0.0548, 0.0689, 0.0552, 0.0487, 0.0431, 0.0442, 0.0570, 0.0525], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:33:20,266 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:47,206 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.347e+02 2.831e+02 3.741e+02 5.862e+02, threshold=5.662e+02, percent-clipped=1.0 2023-04-29 21:33:56,013 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1124, 4.1080, 4.6087, 4.5510, 4.5358, 4.2318, 4.2542, 4.1978], device='cuda:5'), covar=tensor([0.0317, 0.0787, 0.0379, 0.0451, 0.0472, 0.0444, 0.0830, 0.0405], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0344, 0.0342, 0.0334, 0.0390, 0.0370, 0.0450, 0.0296], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 21:34:16,931 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0382, 3.2151, 2.6122, 5.2405, 4.1071, 4.6020, 1.5912, 3.4442], device='cuda:5'), covar=tensor([0.1205, 0.0639, 0.1190, 0.0089, 0.0197, 0.0285, 0.1468, 0.0577], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0160, 0.0180, 0.0150, 0.0188, 0.0202, 0.0186, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 21:34:31,157 INFO [train.py:904] (5/8) Epoch 13, batch 10000, loss[loss=0.1874, simple_loss=0.2672, pruned_loss=0.05381, over 12456.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2715, pruned_loss=0.04191, over 3073573.00 frames. ], batch size: 250, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:35:27,714 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:06,340 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:11,649 INFO [train.py:904] (5/8) Epoch 13, batch 10050, loss[loss=0.1946, simple_loss=0.2912, pruned_loss=0.04898, over 16899.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2707, pruned_loss=0.04145, over 3063190.79 frames. ], batch size: 116, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:36:23,353 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:04,200 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:14,093 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.136e+02 2.458e+02 2.977e+02 5.348e+02, threshold=4.916e+02, percent-clipped=0.0 2023-04-29 21:37:45,832 INFO [train.py:904] (5/8) Epoch 13, batch 10100, loss[loss=0.1889, simple_loss=0.2676, pruned_loss=0.0551, over 12322.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2707, pruned_loss=0.04161, over 3063656.32 frames. ], batch size: 247, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:38:16,839 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:39:29,408 INFO [train.py:904] (5/8) Epoch 14, batch 0, loss[loss=0.1789, simple_loss=0.2656, pruned_loss=0.04604, over 17199.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2656, pruned_loss=0.04604, over 17199.00 frames. ], batch size: 44, lr: 4.96e-03, grad_scale: 8.0 2023-04-29 21:39:29,408 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 21:39:36,901 INFO [train.py:938] (5/8) Epoch 14, validation: loss=0.1515, simple_loss=0.2551, pruned_loss=0.024, over 944034.00 frames. 2023-04-29 21:39:36,902 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 21:40:04,655 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7232, 3.6277, 3.9204, 1.9671, 4.0413, 4.0353, 3.0407, 2.8574], device='cuda:5'), covar=tensor([0.0695, 0.0194, 0.0128, 0.1112, 0.0054, 0.0112, 0.0346, 0.0412], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0097, 0.0083, 0.0132, 0.0065, 0.0103, 0.0116, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 21:40:22,376 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.510e+02 3.083e+02 4.151e+02 9.991e+02, threshold=6.165e+02, percent-clipped=13.0 2023-04-29 21:40:46,692 INFO [train.py:904] (5/8) Epoch 14, batch 50, loss[loss=0.1901, simple_loss=0.2631, pruned_loss=0.05854, over 16920.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.281, pruned_loss=0.06094, over 756164.96 frames. ], batch size: 109, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:02,754 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5966, 2.6063, 1.8023, 2.7465, 2.0815, 2.7780, 2.0089, 2.3446], device='cuda:5'), covar=tensor([0.0238, 0.0346, 0.1337, 0.0205, 0.0696, 0.0546, 0.1220, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0161, 0.0186, 0.0132, 0.0165, 0.0198, 0.0195, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 21:41:24,173 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7676, 4.0041, 3.0182, 2.2930, 2.6839, 2.3571, 4.1248, 3.5457], device='cuda:5'), covar=tensor([0.2603, 0.0602, 0.1564, 0.2606, 0.2506, 0.1892, 0.0462, 0.1353], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0251, 0.0281, 0.0277, 0.0261, 0.0223, 0.0263, 0.0292], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:41:27,186 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132031.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:41:29,015 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7406, 4.1943, 4.3010, 3.1964, 3.6663, 4.3099, 4.0057, 2.5690], device='cuda:5'), covar=tensor([0.0402, 0.0056, 0.0032, 0.0277, 0.0100, 0.0067, 0.0062, 0.0389], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0071, 0.0071, 0.0129, 0.0084, 0.0091, 0.0081, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 21:41:29,261 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 21:41:55,086 INFO [train.py:904] (5/8) Epoch 14, batch 100, loss[loss=0.2137, simple_loss=0.2907, pruned_loss=0.0683, over 16707.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2782, pruned_loss=0.05769, over 1319671.09 frames. ], batch size: 134, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:58,527 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:14,076 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132065.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:23,840 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:44,832 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.362e+02 2.775e+02 3.378e+02 5.633e+02, threshold=5.550e+02, percent-clipped=0.0 2023-04-29 21:42:51,113 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132092.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:04,255 INFO [train.py:904] (5/8) Epoch 14, batch 150, loss[loss=0.1811, simple_loss=0.2789, pruned_loss=0.04162, over 17031.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2743, pruned_loss=0.05429, over 1761221.17 frames. ], batch size: 55, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:43:20,864 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:23,976 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 21:43:27,858 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 21:43:30,575 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:44:04,732 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 21:44:09,376 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:44:15,128 INFO [train.py:904] (5/8) Epoch 14, batch 200, loss[loss=0.2112, simple_loss=0.2783, pruned_loss=0.07208, over 16904.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2731, pruned_loss=0.05355, over 2108780.41 frames. ], batch size: 109, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:44:41,252 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6470, 3.5794, 2.8460, 2.1918, 2.4333, 2.1937, 3.7057, 3.2394], device='cuda:5'), covar=tensor([0.2416, 0.0682, 0.1467, 0.2646, 0.2428, 0.1954, 0.0496, 0.1313], device='cuda:5'), in_proj_covar=tensor([0.0301, 0.0253, 0.0284, 0.0281, 0.0266, 0.0225, 0.0267, 0.0297], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:44:43,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7126, 2.6572, 2.2340, 2.4393, 3.0390, 2.8278, 3.5262, 3.2654], device='cuda:5'), covar=tensor([0.0084, 0.0365, 0.0444, 0.0369, 0.0227, 0.0333, 0.0209, 0.0211], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0215, 0.0208, 0.0207, 0.0213, 0.0213, 0.0215, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:45:02,146 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.381e+02 2.781e+02 3.514e+02 5.258e+02, threshold=5.562e+02, percent-clipped=0.0 2023-04-29 21:45:15,516 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:45:21,426 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 21:45:23,096 INFO [train.py:904] (5/8) Epoch 14, batch 250, loss[loss=0.1971, simple_loss=0.2694, pruned_loss=0.06235, over 16851.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2718, pruned_loss=0.05416, over 2375279.04 frames. ], batch size: 96, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:45:38,812 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:46:33,341 INFO [train.py:904] (5/8) Epoch 14, batch 300, loss[loss=0.1622, simple_loss=0.25, pruned_loss=0.03717, over 17236.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2685, pruned_loss=0.052, over 2585546.77 frames. ], batch size: 45, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:47:16,171 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6891, 1.7604, 1.5365, 1.4314, 1.9099, 1.5946, 1.6780, 1.8794], device='cuda:5'), covar=tensor([0.0188, 0.0260, 0.0397, 0.0359, 0.0182, 0.0306, 0.0179, 0.0197], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0216, 0.0209, 0.0208, 0.0214, 0.0215, 0.0216, 0.0203], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:47:22,676 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.346e+02 2.777e+02 3.394e+02 6.651e+02, threshold=5.554e+02, percent-clipped=2.0 2023-04-29 21:47:43,455 INFO [train.py:904] (5/8) Epoch 14, batch 350, loss[loss=0.1951, simple_loss=0.2665, pruned_loss=0.06187, over 16778.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2661, pruned_loss=0.05065, over 2750871.49 frames. ], batch size: 134, lr: 4.95e-03, grad_scale: 1.0 2023-04-29 21:48:00,512 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0286, 2.0928, 2.5511, 3.0428, 2.7647, 3.3619, 2.2174, 3.3702], device='cuda:5'), covar=tensor([0.0176, 0.0354, 0.0233, 0.0204, 0.0230, 0.0157, 0.0353, 0.0135], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0176, 0.0159, 0.0162, 0.0173, 0.0129, 0.0177, 0.0120], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-29 21:48:51,154 INFO [train.py:904] (5/8) Epoch 14, batch 400, loss[loss=0.1939, simple_loss=0.2671, pruned_loss=0.0604, over 16763.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2648, pruned_loss=0.05072, over 2875521.34 frames. ], batch size: 83, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:49:38,426 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.217e+02 2.653e+02 3.223e+02 5.439e+02, threshold=5.306e+02, percent-clipped=0.0 2023-04-29 21:49:39,472 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132387.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:50:00,140 INFO [train.py:904] (5/8) Epoch 14, batch 450, loss[loss=0.1666, simple_loss=0.2467, pruned_loss=0.04324, over 16812.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2639, pruned_loss=0.04966, over 2978941.16 frames. ], batch size: 102, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:50:12,155 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:51:08,002 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0467, 4.7519, 5.0696, 5.2369, 5.4532, 4.8177, 5.4615, 5.4048], device='cuda:5'), covar=tensor([0.1668, 0.1288, 0.1550, 0.0708, 0.0482, 0.0775, 0.0420, 0.0562], device='cuda:5'), in_proj_covar=tensor([0.0572, 0.0701, 0.0842, 0.0718, 0.0543, 0.0549, 0.0568, 0.0661], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:51:11,610 INFO [train.py:904] (5/8) Epoch 14, batch 500, loss[loss=0.1862, simple_loss=0.2556, pruned_loss=0.05845, over 16690.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2626, pruned_loss=0.04853, over 3060812.71 frames. ], batch size: 134, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:51:58,585 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.377e+02 2.811e+02 3.534e+02 1.282e+03, threshold=5.622e+02, percent-clipped=6.0 2023-04-29 21:52:19,306 INFO [train.py:904] (5/8) Epoch 14, batch 550, loss[loss=0.1885, simple_loss=0.2627, pruned_loss=0.0571, over 15648.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2607, pruned_loss=0.04759, over 3113023.16 frames. ], batch size: 191, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:52:31,344 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 21:52:34,339 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:53:28,147 INFO [train.py:904] (5/8) Epoch 14, batch 600, loss[loss=0.1551, simple_loss=0.2389, pruned_loss=0.03568, over 17265.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2592, pruned_loss=0.0476, over 3157627.02 frames. ], batch size: 43, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:53:41,855 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:54:17,821 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 2.200e+02 2.653e+02 3.173e+02 6.384e+02, threshold=5.306e+02, percent-clipped=1.0 2023-04-29 21:54:39,153 INFO [train.py:904] (5/8) Epoch 14, batch 650, loss[loss=0.1736, simple_loss=0.2408, pruned_loss=0.05321, over 12167.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2585, pruned_loss=0.04727, over 3175837.91 frames. ], batch size: 247, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:55:49,519 INFO [train.py:904] (5/8) Epoch 14, batch 700, loss[loss=0.1872, simple_loss=0.2636, pruned_loss=0.05542, over 17008.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2582, pruned_loss=0.04662, over 3212123.05 frames. ], batch size: 41, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:56:01,409 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:56:38,038 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.258e+02 2.690e+02 3.168e+02 6.172e+02, threshold=5.381e+02, percent-clipped=1.0 2023-04-29 21:56:39,028 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132687.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:56:58,552 INFO [train.py:904] (5/8) Epoch 14, batch 750, loss[loss=0.1826, simple_loss=0.2733, pruned_loss=0.04601, over 16680.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2585, pruned_loss=0.04663, over 3242028.62 frames. ], batch size: 57, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:57:05,629 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 21:57:09,297 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 21:57:24,383 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:43,963 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:59,799 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132746.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:07,426 INFO [train.py:904] (5/8) Epoch 14, batch 800, loss[loss=0.1997, simple_loss=0.2703, pruned_loss=0.06449, over 16879.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2584, pruned_loss=0.04633, over 3264856.08 frames. ], batch size: 109, lr: 4.95e-03, grad_scale: 4.0 2023-04-29 21:58:16,067 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:18,680 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.45 vs. limit=5.0 2023-04-29 21:58:54,914 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.382e+02 2.716e+02 3.142e+02 6.467e+02, threshold=5.432e+02, percent-clipped=2.0 2023-04-29 21:59:08,494 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3944, 5.8243, 5.5964, 5.6856, 5.1523, 5.2127, 5.2821, 5.9683], device='cuda:5'), covar=tensor([0.1304, 0.1069, 0.1166, 0.0731, 0.1001, 0.0788, 0.1188, 0.0922], device='cuda:5'), in_proj_covar=tensor([0.0596, 0.0744, 0.0603, 0.0529, 0.0474, 0.0474, 0.0623, 0.0568], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 21:59:15,865 INFO [train.py:904] (5/8) Epoch 14, batch 850, loss[loss=0.1671, simple_loss=0.263, pruned_loss=0.03563, over 17233.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2577, pruned_loss=0.0457, over 3274427.67 frames. ], batch size: 52, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 21:59:23,474 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:59:58,622 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5470, 2.3915, 1.8868, 2.1952, 2.7648, 2.6065, 2.8172, 2.8676], device='cuda:5'), covar=tensor([0.0188, 0.0312, 0.0435, 0.0350, 0.0176, 0.0250, 0.0201, 0.0228], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0217, 0.0211, 0.0210, 0.0217, 0.0218, 0.0222, 0.0209], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:00:24,480 INFO [train.py:904] (5/8) Epoch 14, batch 900, loss[loss=0.1489, simple_loss=0.2291, pruned_loss=0.03436, over 15861.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2566, pruned_loss=0.04491, over 3283129.66 frames. ], batch size: 35, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:01:14,374 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.299e+02 2.623e+02 3.116e+02 6.405e+02, threshold=5.245e+02, percent-clipped=2.0 2023-04-29 22:01:15,135 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 22:01:34,128 INFO [train.py:904] (5/8) Epoch 14, batch 950, loss[loss=0.1725, simple_loss=0.2485, pruned_loss=0.04827, over 15584.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2578, pruned_loss=0.04505, over 3295752.98 frames. ], batch size: 190, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:01:52,530 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 2023-04-29 22:02:36,341 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6545, 4.3276, 4.3522, 2.9625, 3.7393, 4.3143, 3.9939, 2.4326], device='cuda:5'), covar=tensor([0.0449, 0.0051, 0.0038, 0.0347, 0.0087, 0.0082, 0.0068, 0.0396], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0073, 0.0074, 0.0129, 0.0085, 0.0094, 0.0083, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:02:41,762 INFO [train.py:904] (5/8) Epoch 14, batch 1000, loss[loss=0.152, simple_loss=0.2293, pruned_loss=0.03729, over 16741.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2571, pruned_loss=0.04501, over 3301282.37 frames. ], batch size: 134, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:03:29,375 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.118e+02 2.524e+02 3.195e+02 6.689e+02, threshold=5.047e+02, percent-clipped=1.0 2023-04-29 22:03:50,009 INFO [train.py:904] (5/8) Epoch 14, batch 1050, loss[loss=0.179, simple_loss=0.2719, pruned_loss=0.04307, over 17092.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2569, pruned_loss=0.04513, over 3300940.78 frames. ], batch size: 53, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:04:10,801 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:04:59,804 INFO [train.py:904] (5/8) Epoch 14, batch 1100, loss[loss=0.1761, simple_loss=0.2676, pruned_loss=0.04229, over 16734.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2563, pruned_loss=0.0449, over 3308428.32 frames. ], batch size: 57, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:05:39,971 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1482, 3.2680, 3.2032, 2.1597, 2.7423, 2.3650, 3.6070, 3.5749], device='cuda:5'), covar=tensor([0.0219, 0.0819, 0.0602, 0.1734, 0.0871, 0.0969, 0.0502, 0.0855], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0147, 0.0157, 0.0145, 0.0137, 0.0125, 0.0135, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 22:05:47,408 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.252e+02 2.587e+02 3.047e+02 6.829e+02, threshold=5.174e+02, percent-clipped=2.0 2023-04-29 22:06:05,400 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7593, 3.8979, 2.2891, 4.5040, 2.8601, 4.5061, 2.5955, 3.2570], device='cuda:5'), covar=tensor([0.0275, 0.0397, 0.1512, 0.0253, 0.0836, 0.0416, 0.1397, 0.0653], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0170, 0.0192, 0.0145, 0.0170, 0.0211, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 22:06:08,350 INFO [train.py:904] (5/8) Epoch 14, batch 1150, loss[loss=0.1907, simple_loss=0.2626, pruned_loss=0.05939, over 15349.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2552, pruned_loss=0.04465, over 3304185.74 frames. ], batch size: 190, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:06:08,674 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:06:35,910 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8661, 4.9843, 5.1021, 4.9483, 4.9453, 5.5872, 5.1077, 4.8496], device='cuda:5'), covar=tensor([0.1268, 0.2003, 0.2248, 0.2197, 0.3079, 0.1092, 0.1532, 0.2541], device='cuda:5'), in_proj_covar=tensor([0.0372, 0.0528, 0.0580, 0.0450, 0.0611, 0.0603, 0.0459, 0.0600], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:07:16,500 INFO [train.py:904] (5/8) Epoch 14, batch 1200, loss[loss=0.164, simple_loss=0.2659, pruned_loss=0.03105, over 17247.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2551, pruned_loss=0.04451, over 3302195.76 frames. ], batch size: 45, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:07:35,603 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3291, 3.4427, 1.9634, 3.5581, 2.6129, 3.5650, 2.1655, 2.7471], device='cuda:5'), covar=tensor([0.0253, 0.0339, 0.1479, 0.0308, 0.0764, 0.0694, 0.1375, 0.0662], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0169, 0.0191, 0.0145, 0.0170, 0.0210, 0.0199, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 22:07:48,168 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7192, 2.6903, 2.2864, 2.5966, 3.0442, 2.9230, 3.5367, 3.3040], device='cuda:5'), covar=tensor([0.0120, 0.0331, 0.0428, 0.0372, 0.0226, 0.0314, 0.0192, 0.0226], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0219, 0.0212, 0.0211, 0.0218, 0.0220, 0.0224, 0.0211], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:08:05,980 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.506e+02 2.966e+02 3.654e+02 5.614e+02, threshold=5.932e+02, percent-clipped=4.0 2023-04-29 22:08:28,191 INFO [train.py:904] (5/8) Epoch 14, batch 1250, loss[loss=0.1795, simple_loss=0.2722, pruned_loss=0.04338, over 16739.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2553, pruned_loss=0.04519, over 3309848.41 frames. ], batch size: 57, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:09:37,987 INFO [train.py:904] (5/8) Epoch 14, batch 1300, loss[loss=0.1821, simple_loss=0.26, pruned_loss=0.05215, over 16196.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2549, pruned_loss=0.04553, over 3311679.72 frames. ], batch size: 164, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:10:10,886 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1129, 5.5158, 5.7200, 5.4673, 5.5171, 6.1267, 5.6956, 5.3879], device='cuda:5'), covar=tensor([0.0915, 0.1967, 0.1885, 0.2321, 0.2951, 0.1043, 0.1390, 0.2554], device='cuda:5'), in_proj_covar=tensor([0.0371, 0.0527, 0.0580, 0.0448, 0.0611, 0.0604, 0.0459, 0.0600], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:10:13,954 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9857, 3.7841, 3.7676, 4.1468, 4.2539, 3.9277, 4.0718, 4.2597], device='cuda:5'), covar=tensor([0.1476, 0.1128, 0.2079, 0.0968, 0.0794, 0.1706, 0.1919, 0.0905], device='cuda:5'), in_proj_covar=tensor([0.0580, 0.0716, 0.0859, 0.0732, 0.0552, 0.0561, 0.0576, 0.0674], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:10:27,126 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 2.160e+02 2.631e+02 3.090e+02 7.401e+02, threshold=5.263e+02, percent-clipped=1.0 2023-04-29 22:10:42,228 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4731, 5.4707, 5.2724, 4.5174, 5.2963, 2.0132, 4.9837, 5.2225], device='cuda:5'), covar=tensor([0.0063, 0.0058, 0.0150, 0.0365, 0.0077, 0.2349, 0.0117, 0.0146], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0133, 0.0179, 0.0165, 0.0150, 0.0194, 0.0169, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:10:48,355 INFO [train.py:904] (5/8) Epoch 14, batch 1350, loss[loss=0.1634, simple_loss=0.2573, pruned_loss=0.03476, over 17087.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2551, pruned_loss=0.04533, over 3309714.36 frames. ], batch size: 47, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:11:06,820 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:11:56,452 INFO [train.py:904] (5/8) Epoch 14, batch 1400, loss[loss=0.1598, simple_loss=0.2376, pruned_loss=0.04105, over 15633.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2548, pruned_loss=0.0452, over 3314323.11 frames. ], batch size: 190, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:12:13,482 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:12:26,554 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:12:30,114 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:12:44,813 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.235e+02 2.580e+02 3.131e+02 5.456e+02, threshold=5.161e+02, percent-clipped=1.0 2023-04-29 22:13:06,252 INFO [train.py:904] (5/8) Epoch 14, batch 1450, loss[loss=0.175, simple_loss=0.2441, pruned_loss=0.05295, over 16622.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2541, pruned_loss=0.04491, over 3315396.42 frames. ], batch size: 134, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:13:06,595 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:13:17,364 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5368, 3.7564, 3.9757, 2.8408, 3.6508, 3.9525, 3.8229, 2.3191], device='cuda:5'), covar=tensor([0.0423, 0.0272, 0.0050, 0.0309, 0.0091, 0.0117, 0.0078, 0.0413], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0072, 0.0073, 0.0128, 0.0084, 0.0094, 0.0082, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:13:52,513 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 22:13:55,207 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:14:12,935 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:14:14,826 INFO [train.py:904] (5/8) Epoch 14, batch 1500, loss[loss=0.1781, simple_loss=0.2637, pruned_loss=0.04627, over 16814.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2544, pruned_loss=0.04539, over 3327550.24 frames. ], batch size: 42, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:14:18,466 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-04-29 22:14:36,204 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3863, 3.3379, 3.3958, 3.5249, 3.5395, 3.2743, 3.4802, 3.6038], device='cuda:5'), covar=tensor([0.1157, 0.0915, 0.1051, 0.0619, 0.0598, 0.2445, 0.1197, 0.0694], device='cuda:5'), in_proj_covar=tensor([0.0588, 0.0727, 0.0873, 0.0743, 0.0560, 0.0572, 0.0586, 0.0687], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:14:46,626 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:15:03,757 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.267e+02 2.734e+02 3.162e+02 6.055e+02, threshold=5.467e+02, percent-clipped=3.0 2023-04-29 22:15:21,281 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9928, 5.5085, 5.6580, 5.3703, 5.5645, 6.0987, 5.6806, 5.4063], device='cuda:5'), covar=tensor([0.0965, 0.1973, 0.2212, 0.2382, 0.2656, 0.0986, 0.1336, 0.2416], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0537, 0.0589, 0.0456, 0.0619, 0.0613, 0.0469, 0.0612], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:15:23,303 INFO [train.py:904] (5/8) Epoch 14, batch 1550, loss[loss=0.154, simple_loss=0.2389, pruned_loss=0.03458, over 17246.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2553, pruned_loss=0.04561, over 3323455.20 frames. ], batch size: 43, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:16:10,868 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:16:33,568 INFO [train.py:904] (5/8) Epoch 14, batch 1600, loss[loss=0.1888, simple_loss=0.2676, pruned_loss=0.055, over 16836.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2571, pruned_loss=0.04625, over 3317236.18 frames. ], batch size: 83, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:16:39,283 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4060, 5.8105, 5.5723, 5.6379, 5.2449, 5.1039, 5.1736, 5.8967], device='cuda:5'), covar=tensor([0.1261, 0.0871, 0.0965, 0.0709, 0.0746, 0.0727, 0.1118, 0.0880], device='cuda:5'), in_proj_covar=tensor([0.0603, 0.0756, 0.0610, 0.0539, 0.0480, 0.0483, 0.0633, 0.0578], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:17:21,877 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.312e+02 2.830e+02 3.166e+02 6.146e+02, threshold=5.659e+02, percent-clipped=1.0 2023-04-29 22:17:43,039 INFO [train.py:904] (5/8) Epoch 14, batch 1650, loss[loss=0.1901, simple_loss=0.2587, pruned_loss=0.06074, over 16848.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2588, pruned_loss=0.04654, over 3316239.45 frames. ], batch size: 109, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:37,865 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 22:18:52,521 INFO [train.py:904] (5/8) Epoch 14, batch 1700, loss[loss=0.1764, simple_loss=0.2673, pruned_loss=0.04275, over 17036.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2606, pruned_loss=0.04693, over 3322806.81 frames. ], batch size: 55, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:53,065 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133652.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:19:25,757 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9409, 4.9315, 5.3868, 5.3610, 5.3931, 4.9947, 4.9907, 4.6741], device='cuda:5'), covar=tensor([0.0301, 0.0519, 0.0387, 0.0418, 0.0410, 0.0357, 0.0875, 0.0438], device='cuda:5'), in_proj_covar=tensor([0.0371, 0.0389, 0.0389, 0.0372, 0.0430, 0.0416, 0.0508, 0.0330], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 22:19:42,414 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.547e+02 3.320e+02 4.199e+02 1.315e+03, threshold=6.641e+02, percent-clipped=10.0 2023-04-29 22:19:56,954 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7187, 5.0230, 4.8019, 4.7968, 4.5888, 4.4237, 4.5193, 5.0817], device='cuda:5'), covar=tensor([0.1121, 0.0916, 0.1079, 0.0782, 0.0767, 0.1148, 0.1113, 0.0933], device='cuda:5'), in_proj_covar=tensor([0.0599, 0.0750, 0.0606, 0.0535, 0.0477, 0.0478, 0.0629, 0.0574], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:20:03,033 INFO [train.py:904] (5/8) Epoch 14, batch 1750, loss[loss=0.1632, simple_loss=0.2574, pruned_loss=0.03452, over 17205.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2617, pruned_loss=0.04715, over 3315826.04 frames. ], batch size: 45, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:20:19,373 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133713.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:20:42,466 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 22:20:45,496 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133732.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:21:12,616 INFO [train.py:904] (5/8) Epoch 14, batch 1800, loss[loss=0.2013, simple_loss=0.2797, pruned_loss=0.06142, over 16234.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2628, pruned_loss=0.04752, over 3312082.38 frames. ], batch size: 165, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:22:00,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.275e+02 2.807e+02 3.369e+02 7.697e+02, threshold=5.615e+02, percent-clipped=1.0 2023-04-29 22:22:22,389 INFO [train.py:904] (5/8) Epoch 14, batch 1850, loss[loss=0.1761, simple_loss=0.2781, pruned_loss=0.03706, over 17108.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.264, pruned_loss=0.04812, over 3313719.76 frames. ], batch size: 47, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:22:54,271 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1156, 4.8498, 5.1215, 5.3563, 5.5758, 4.7968, 5.5077, 5.5273], device='cuda:5'), covar=tensor([0.1640, 0.1267, 0.1825, 0.0781, 0.0522, 0.0759, 0.0533, 0.0561], device='cuda:5'), in_proj_covar=tensor([0.0601, 0.0742, 0.0891, 0.0759, 0.0571, 0.0586, 0.0598, 0.0704], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:23:03,056 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133830.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:23:32,967 INFO [train.py:904] (5/8) Epoch 14, batch 1900, loss[loss=0.177, simple_loss=0.2696, pruned_loss=0.04223, over 17029.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2624, pruned_loss=0.04677, over 3320623.47 frames. ], batch size: 50, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:23:45,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5042, 4.5327, 4.9051, 4.8893, 4.9275, 4.5880, 4.5893, 4.4128], device='cuda:5'), covar=tensor([0.0331, 0.0526, 0.0399, 0.0455, 0.0481, 0.0438, 0.0910, 0.0564], device='cuda:5'), in_proj_covar=tensor([0.0375, 0.0391, 0.0394, 0.0375, 0.0435, 0.0418, 0.0512, 0.0332], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 22:24:23,484 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 2.126e+02 2.452e+02 2.920e+02 9.757e+02, threshold=4.904e+02, percent-clipped=2.0 2023-04-29 22:24:30,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4032, 5.7730, 5.5152, 5.6164, 5.1274, 5.1380, 5.2610, 5.8630], device='cuda:5'), covar=tensor([0.1188, 0.0938, 0.1126, 0.0681, 0.0825, 0.0680, 0.1095, 0.0953], device='cuda:5'), in_proj_covar=tensor([0.0598, 0.0748, 0.0606, 0.0533, 0.0475, 0.0476, 0.0623, 0.0571], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:24:32,329 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8677, 4.3233, 3.2073, 2.3038, 2.8880, 2.5154, 4.6612, 3.8449], device='cuda:5'), covar=tensor([0.2426, 0.0568, 0.1509, 0.2409, 0.2417, 0.1793, 0.0322, 0.0957], device='cuda:5'), in_proj_covar=tensor([0.0309, 0.0263, 0.0290, 0.0288, 0.0282, 0.0232, 0.0275, 0.0311], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:24:42,152 INFO [train.py:904] (5/8) Epoch 14, batch 1950, loss[loss=0.1832, simple_loss=0.2768, pruned_loss=0.04476, over 16664.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.263, pruned_loss=0.04673, over 3320510.10 frames. ], batch size: 62, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:25:52,918 INFO [train.py:904] (5/8) Epoch 14, batch 2000, loss[loss=0.1765, simple_loss=0.2711, pruned_loss=0.04099, over 17259.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2635, pruned_loss=0.04724, over 3316666.25 frames. ], batch size: 45, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:26:18,954 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6529, 3.7828, 2.9119, 2.2159, 2.4650, 2.2891, 3.8190, 3.3501], device='cuda:5'), covar=tensor([0.2560, 0.0616, 0.1531, 0.2726, 0.2654, 0.1915, 0.0486, 0.1272], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0262, 0.0290, 0.0289, 0.0283, 0.0233, 0.0276, 0.0312], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:26:43,622 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.429e+02 2.777e+02 3.533e+02 6.577e+02, threshold=5.554e+02, percent-clipped=5.0 2023-04-29 22:26:57,047 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133997.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:27:06,905 INFO [train.py:904] (5/8) Epoch 14, batch 2050, loss[loss=0.1666, simple_loss=0.2634, pruned_loss=0.0349, over 17038.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2632, pruned_loss=0.04738, over 3320166.34 frames. ], batch size: 50, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:27:09,266 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7412, 4.1233, 4.1169, 3.1110, 3.6464, 4.2253, 3.8629, 2.4468], device='cuda:5'), covar=tensor([0.0376, 0.0060, 0.0047, 0.0269, 0.0094, 0.0087, 0.0071, 0.0398], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0072, 0.0073, 0.0128, 0.0084, 0.0094, 0.0083, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:27:15,553 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:27:21,657 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5585, 4.4975, 4.4903, 3.9654, 4.4786, 1.8439, 4.2535, 4.1942], device='cuda:5'), covar=tensor([0.0117, 0.0102, 0.0154, 0.0330, 0.0094, 0.2378, 0.0151, 0.0185], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0136, 0.0184, 0.0170, 0.0154, 0.0196, 0.0173, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:27:46,424 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 22:27:49,330 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:28:16,838 INFO [train.py:904] (5/8) Epoch 14, batch 2100, loss[loss=0.2405, simple_loss=0.3134, pruned_loss=0.08386, over 11957.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.264, pruned_loss=0.04769, over 3312893.78 frames. ], batch size: 246, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:28:25,720 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:28:54,068 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 22:28:56,310 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:29:09,014 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.284e+02 2.721e+02 3.276e+02 7.722e+02, threshold=5.442e+02, percent-clipped=1.0 2023-04-29 22:29:26,684 INFO [train.py:904] (5/8) Epoch 14, batch 2150, loss[loss=0.1658, simple_loss=0.2537, pruned_loss=0.03895, over 16767.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2658, pruned_loss=0.04852, over 3317974.72 frames. ], batch size: 39, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:29:59,873 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 22:30:05,110 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8533, 3.0194, 2.6140, 4.5140, 3.7910, 4.2537, 1.7352, 3.0476], device='cuda:5'), covar=tensor([0.1285, 0.0647, 0.1086, 0.0188, 0.0244, 0.0392, 0.1395, 0.0760], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0163, 0.0181, 0.0161, 0.0198, 0.0210, 0.0186, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 22:30:06,158 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:30:30,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6168, 2.2350, 2.2554, 4.3420, 2.1748, 2.6089, 2.2775, 2.4282], device='cuda:5'), covar=tensor([0.1056, 0.3536, 0.2664, 0.0435, 0.3947, 0.2422, 0.3207, 0.3354], device='cuda:5'), in_proj_covar=tensor([0.0380, 0.0412, 0.0345, 0.0328, 0.0421, 0.0476, 0.0378, 0.0484], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:30:35,972 INFO [train.py:904] (5/8) Epoch 14, batch 2200, loss[loss=0.197, simple_loss=0.2701, pruned_loss=0.06196, over 16674.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2654, pruned_loss=0.04819, over 3323206.95 frames. ], batch size: 89, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:31:12,755 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134178.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:31:27,739 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.617e+02 2.496e+02 2.998e+02 3.803e+02 8.330e+02, threshold=5.996e+02, percent-clipped=6.0 2023-04-29 22:31:41,752 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 22:31:45,971 INFO [train.py:904] (5/8) Epoch 14, batch 2250, loss[loss=0.1821, simple_loss=0.2603, pruned_loss=0.05197, over 16697.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2666, pruned_loss=0.049, over 3324802.22 frames. ], batch size: 89, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:32:54,512 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2703, 5.0079, 5.2857, 5.4561, 5.6538, 4.9323, 5.6396, 5.6230], device='cuda:5'), covar=tensor([0.1517, 0.1229, 0.1555, 0.0704, 0.0474, 0.0773, 0.0429, 0.0490], device='cuda:5'), in_proj_covar=tensor([0.0601, 0.0743, 0.0890, 0.0764, 0.0572, 0.0588, 0.0602, 0.0702], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:32:56,461 INFO [train.py:904] (5/8) Epoch 14, batch 2300, loss[loss=0.2029, simple_loss=0.2805, pruned_loss=0.06261, over 16454.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2665, pruned_loss=0.04903, over 3329445.16 frames. ], batch size: 146, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:33:46,630 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7436, 3.6597, 4.1354, 2.0975, 4.2470, 4.2500, 3.1376, 3.2296], device='cuda:5'), covar=tensor([0.0731, 0.0219, 0.0175, 0.1086, 0.0073, 0.0171, 0.0394, 0.0374], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0102, 0.0089, 0.0136, 0.0070, 0.0112, 0.0121, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 22:33:48,102 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.438e+02 2.937e+02 3.292e+02 7.337e+02, threshold=5.875e+02, percent-clipped=1.0 2023-04-29 22:33:57,549 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-29 22:33:59,462 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 22:34:06,357 INFO [train.py:904] (5/8) Epoch 14, batch 2350, loss[loss=0.1567, simple_loss=0.2432, pruned_loss=0.03505, over 16751.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2678, pruned_loss=0.04927, over 3309707.56 frames. ], batch size: 39, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:34:09,722 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:15,119 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:16,981 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:38,289 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3747, 5.7217, 5.4232, 5.5487, 5.0762, 4.8744, 5.1198, 5.8215], device='cuda:5'), covar=tensor([0.1156, 0.0955, 0.1114, 0.0768, 0.0903, 0.0779, 0.1040, 0.0945], device='cuda:5'), in_proj_covar=tensor([0.0600, 0.0752, 0.0611, 0.0538, 0.0477, 0.0477, 0.0625, 0.0577], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:35:17,013 INFO [train.py:904] (5/8) Epoch 14, batch 2400, loss[loss=0.1747, simple_loss=0.2715, pruned_loss=0.03899, over 17142.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2677, pruned_loss=0.0486, over 3317370.01 frames. ], batch size: 49, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:35:19,249 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:23,358 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:35,666 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:42,586 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:46,766 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9772, 2.9791, 2.3902, 2.6218, 3.2603, 3.0349, 3.7329, 3.5405], device='cuda:5'), covar=tensor([0.0083, 0.0310, 0.0441, 0.0345, 0.0223, 0.0256, 0.0199, 0.0188], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0222, 0.0215, 0.0214, 0.0222, 0.0223, 0.0230, 0.0216], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:36:08,842 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.303e+02 2.657e+02 3.180e+02 6.909e+02, threshold=5.314e+02, percent-clipped=1.0 2023-04-29 22:36:25,723 INFO [train.py:904] (5/8) Epoch 14, batch 2450, loss[loss=0.1606, simple_loss=0.25, pruned_loss=0.03558, over 17226.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2678, pruned_loss=0.04813, over 3321801.17 frames. ], batch size: 45, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:37:29,795 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6749, 4.4627, 4.7001, 4.8796, 5.0434, 4.5336, 4.9377, 5.0044], device='cuda:5'), covar=tensor([0.1570, 0.1219, 0.1534, 0.0716, 0.0633, 0.0983, 0.1199, 0.0773], device='cuda:5'), in_proj_covar=tensor([0.0605, 0.0747, 0.0899, 0.0767, 0.0574, 0.0592, 0.0605, 0.0705], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:37:35,062 INFO [train.py:904] (5/8) Epoch 14, batch 2500, loss[loss=0.1853, simple_loss=0.2599, pruned_loss=0.05533, over 16867.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2677, pruned_loss=0.04821, over 3309667.55 frames. ], batch size: 96, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:37:46,246 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1076, 4.8477, 5.1226, 5.3225, 5.5246, 4.8291, 5.5035, 5.4677], device='cuda:5'), covar=tensor([0.1538, 0.1187, 0.1570, 0.0674, 0.0535, 0.0773, 0.0475, 0.0608], device='cuda:5'), in_proj_covar=tensor([0.0606, 0.0748, 0.0900, 0.0768, 0.0575, 0.0593, 0.0606, 0.0707], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:38:28,210 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.356e+02 2.745e+02 3.374e+02 4.761e+02, threshold=5.489e+02, percent-clipped=0.0 2023-04-29 22:38:30,034 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 22:38:41,318 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8876, 1.8830, 2.4268, 2.8187, 2.7556, 2.7837, 1.9760, 3.0430], device='cuda:5'), covar=tensor([0.0136, 0.0358, 0.0235, 0.0167, 0.0184, 0.0190, 0.0371, 0.0106], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0181, 0.0166, 0.0171, 0.0180, 0.0136, 0.0181, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:38:45,449 INFO [train.py:904] (5/8) Epoch 14, batch 2550, loss[loss=0.169, simple_loss=0.2483, pruned_loss=0.0449, over 16660.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2665, pruned_loss=0.04762, over 3317323.89 frames. ], batch size: 89, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:38:49,920 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9526, 4.6852, 4.9695, 5.1699, 5.3723, 4.6177, 5.3352, 5.3265], device='cuda:5'), covar=tensor([0.1707, 0.1390, 0.1696, 0.0719, 0.0549, 0.1018, 0.0595, 0.0643], device='cuda:5'), in_proj_covar=tensor([0.0607, 0.0749, 0.0902, 0.0770, 0.0576, 0.0595, 0.0608, 0.0708], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:39:06,176 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3370, 5.3117, 5.1711, 4.5448, 5.1791, 2.1551, 4.9335, 5.1894], device='cuda:5'), covar=tensor([0.0081, 0.0080, 0.0156, 0.0384, 0.0088, 0.2196, 0.0131, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0135, 0.0182, 0.0169, 0.0153, 0.0194, 0.0172, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:39:55,164 INFO [train.py:904] (5/8) Epoch 14, batch 2600, loss[loss=0.1842, simple_loss=0.2756, pruned_loss=0.04636, over 17001.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2668, pruned_loss=0.04768, over 3316341.19 frames. ], batch size: 50, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:40:19,780 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.33 vs. limit=5.0 2023-04-29 22:40:46,905 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.457e+02 2.364e+02 2.744e+02 3.247e+02 7.768e+02, threshold=5.488e+02, percent-clipped=4.0 2023-04-29 22:41:03,691 INFO [train.py:904] (5/8) Epoch 14, batch 2650, loss[loss=0.1852, simple_loss=0.2661, pruned_loss=0.05212, over 16304.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2673, pruned_loss=0.04777, over 3321367.19 frames. ], batch size: 164, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:42:12,180 INFO [train.py:904] (5/8) Epoch 14, batch 2700, loss[loss=0.1598, simple_loss=0.2542, pruned_loss=0.03265, over 15958.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.267, pruned_loss=0.04718, over 3324220.98 frames. ], batch size: 35, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:42:13,630 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:16,903 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0680, 3.3353, 3.2555, 2.0773, 2.8204, 2.3151, 3.5074, 3.5581], device='cuda:5'), covar=tensor([0.0248, 0.0764, 0.0597, 0.1684, 0.0768, 0.0930, 0.0578, 0.0785], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0152, 0.0160, 0.0147, 0.0139, 0.0126, 0.0138, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 22:42:23,637 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:31,301 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:43:04,249 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.219e+02 2.559e+02 3.067e+02 4.567e+02, threshold=5.119e+02, percent-clipped=0.0 2023-04-29 22:43:05,871 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9228, 4.0720, 2.4205, 4.7572, 3.0951, 4.7229, 2.7035, 3.4021], device='cuda:5'), covar=tensor([0.0262, 0.0358, 0.1590, 0.0216, 0.0851, 0.0376, 0.1344, 0.0636], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0172, 0.0191, 0.0148, 0.0170, 0.0214, 0.0200, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 22:43:20,436 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:43:21,291 INFO [train.py:904] (5/8) Epoch 14, batch 2750, loss[loss=0.2128, simple_loss=0.2893, pruned_loss=0.06817, over 12318.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2673, pruned_loss=0.04704, over 3322377.51 frames. ], batch size: 247, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:43:27,350 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8840, 4.3333, 2.9526, 2.2629, 2.9371, 2.4448, 4.6558, 3.8108], device='cuda:5'), covar=tensor([0.2591, 0.0559, 0.1723, 0.2510, 0.2453, 0.1868, 0.0374, 0.1114], device='cuda:5'), in_proj_covar=tensor([0.0312, 0.0265, 0.0293, 0.0292, 0.0287, 0.0234, 0.0279, 0.0317], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:43:48,092 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 22:44:05,239 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 22:44:29,021 INFO [train.py:904] (5/8) Epoch 14, batch 2800, loss[loss=0.1583, simple_loss=0.2483, pruned_loss=0.03413, over 17210.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2672, pruned_loss=0.04704, over 3313083.76 frames. ], batch size: 45, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:45:20,158 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.461e+02 2.848e+02 3.331e+02 6.185e+02, threshold=5.697e+02, percent-clipped=2.0 2023-04-29 22:45:37,178 INFO [train.py:904] (5/8) Epoch 14, batch 2850, loss[loss=0.1853, simple_loss=0.2735, pruned_loss=0.04858, over 16711.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2669, pruned_loss=0.04728, over 3312564.53 frames. ], batch size: 62, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:03,426 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-04-29 22:46:45,318 INFO [train.py:904] (5/8) Epoch 14, batch 2900, loss[loss=0.1633, simple_loss=0.2495, pruned_loss=0.03852, over 16844.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2649, pruned_loss=0.04705, over 3319983.40 frames. ], batch size: 42, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:58,788 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:47:29,344 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0584, 2.4549, 2.3966, 2.8040, 2.3494, 3.2454, 1.8620, 2.7132], device='cuda:5'), covar=tensor([0.1071, 0.0488, 0.0850, 0.0174, 0.0174, 0.0441, 0.1192, 0.0620], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0163, 0.0182, 0.0163, 0.0200, 0.0211, 0.0186, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 22:47:36,338 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.493e+02 2.912e+02 3.442e+02 5.810e+02, threshold=5.824e+02, percent-clipped=2.0 2023-04-29 22:47:54,152 INFO [train.py:904] (5/8) Epoch 14, batch 2950, loss[loss=0.193, simple_loss=0.281, pruned_loss=0.0525, over 17070.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2644, pruned_loss=0.04751, over 3315248.23 frames. ], batch size: 55, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:48:24,091 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:02,689 INFO [train.py:904] (5/8) Epoch 14, batch 3000, loss[loss=0.1554, simple_loss=0.2501, pruned_loss=0.0303, over 17161.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2652, pruned_loss=0.04818, over 3314671.04 frames. ], batch size: 46, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:49:02,689 INFO [train.py:929] (5/8) Computing validation loss 2023-04-29 22:49:12,425 INFO [train.py:938] (5/8) Epoch 14, validation: loss=0.1382, simple_loss=0.244, pruned_loss=0.01621, over 944034.00 frames. 2023-04-29 22:49:12,425 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-29 22:49:24,942 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:31,433 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:06,982 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.394e+02 2.911e+02 3.370e+02 7.406e+02, threshold=5.821e+02, percent-clipped=1.0 2023-04-29 22:50:24,234 INFO [train.py:904] (5/8) Epoch 14, batch 3050, loss[loss=0.1697, simple_loss=0.2699, pruned_loss=0.03475, over 17242.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2648, pruned_loss=0.04796, over 3323189.84 frames. ], batch size: 52, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:50:33,267 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:36,698 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-29 22:50:40,114 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:51:32,231 INFO [train.py:904] (5/8) Epoch 14, batch 3100, loss[loss=0.1848, simple_loss=0.2864, pruned_loss=0.04156, over 17029.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2646, pruned_loss=0.04808, over 3323160.80 frames. ], batch size: 50, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:52:24,973 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.254e+02 2.683e+02 3.212e+02 6.924e+02, threshold=5.366e+02, percent-clipped=1.0 2023-04-29 22:52:33,309 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 22:52:40,989 INFO [train.py:904] (5/8) Epoch 14, batch 3150, loss[loss=0.1828, simple_loss=0.2775, pruned_loss=0.04406, over 17013.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2637, pruned_loss=0.04763, over 3321045.96 frames. ], batch size: 50, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:53:10,228 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6948, 1.8501, 2.2988, 2.6578, 2.6547, 2.6797, 1.7647, 2.8470], device='cuda:5'), covar=tensor([0.0149, 0.0356, 0.0252, 0.0218, 0.0193, 0.0208, 0.0385, 0.0125], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0182, 0.0167, 0.0174, 0.0181, 0.0137, 0.0182, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 22:53:50,747 INFO [train.py:904] (5/8) Epoch 14, batch 3200, loss[loss=0.1696, simple_loss=0.2539, pruned_loss=0.0427, over 16566.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2632, pruned_loss=0.04722, over 3319826.09 frames. ], batch size: 68, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:54:17,383 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:54:31,273 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-29 22:54:41,148 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.733e+02 2.470e+02 2.854e+02 3.612e+02 7.577e+02, threshold=5.709e+02, percent-clipped=5.0 2023-04-29 22:54:59,028 INFO [train.py:904] (5/8) Epoch 14, batch 3250, loss[loss=0.1809, simple_loss=0.2709, pruned_loss=0.04545, over 17033.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2635, pruned_loss=0.04755, over 3315053.65 frames. ], batch size: 55, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:55:22,157 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:55:41,809 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135232.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:56:07,855 INFO [train.py:904] (5/8) Epoch 14, batch 3300, loss[loss=0.181, simple_loss=0.2596, pruned_loss=0.05118, over 16715.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2641, pruned_loss=0.04765, over 3315672.99 frames. ], batch size: 124, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:56:11,626 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 22:57:01,618 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 2.198e+02 2.720e+02 3.178e+02 6.678e+02, threshold=5.440e+02, percent-clipped=1.0 2023-04-29 22:57:17,008 INFO [train.py:904] (5/8) Epoch 14, batch 3350, loss[loss=0.1986, simple_loss=0.2773, pruned_loss=0.05997, over 16868.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2644, pruned_loss=0.04731, over 3321224.84 frames. ], batch size: 116, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:57:40,972 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2592, 3.8480, 4.2663, 2.0187, 4.4863, 4.5452, 3.3629, 3.4737], device='cuda:5'), covar=tensor([0.0553, 0.0219, 0.0180, 0.1157, 0.0057, 0.0140, 0.0339, 0.0362], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0103, 0.0090, 0.0136, 0.0072, 0.0114, 0.0121, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-29 22:57:48,082 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7186, 4.8058, 5.0261, 4.8296, 4.8068, 5.4165, 4.9405, 4.5923], device='cuda:5'), covar=tensor([0.1291, 0.1811, 0.1783, 0.2009, 0.2716, 0.1063, 0.1437, 0.2436], device='cuda:5'), in_proj_covar=tensor([0.0383, 0.0544, 0.0591, 0.0462, 0.0626, 0.0621, 0.0471, 0.0615], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 22:58:24,257 INFO [train.py:904] (5/8) Epoch 14, batch 3400, loss[loss=0.1878, simple_loss=0.2644, pruned_loss=0.0556, over 16800.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2641, pruned_loss=0.04738, over 3323725.28 frames. ], batch size: 124, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:59:16,985 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.205e+02 2.551e+02 3.042e+02 5.815e+02, threshold=5.103e+02, percent-clipped=2.0 2023-04-29 22:59:32,658 INFO [train.py:904] (5/8) Epoch 14, batch 3450, loss[loss=0.1588, simple_loss=0.2416, pruned_loss=0.03802, over 16153.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2622, pruned_loss=0.04661, over 3331723.86 frames. ], batch size: 36, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:59:38,170 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3933, 5.8224, 5.5269, 5.6049, 5.2259, 5.2288, 5.1674, 5.9015], device='cuda:5'), covar=tensor([0.1348, 0.0886, 0.1128, 0.0751, 0.0928, 0.0671, 0.1160, 0.1002], device='cuda:5'), in_proj_covar=tensor([0.0611, 0.0761, 0.0623, 0.0548, 0.0486, 0.0482, 0.0633, 0.0585], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:00:10,313 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5565, 2.4416, 1.9783, 2.2296, 2.7193, 2.5480, 2.7482, 2.8593], device='cuda:5'), covar=tensor([0.0135, 0.0274, 0.0370, 0.0361, 0.0172, 0.0250, 0.0195, 0.0204], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0220, 0.0215, 0.0215, 0.0222, 0.0222, 0.0232, 0.0216], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:00:41,089 INFO [train.py:904] (5/8) Epoch 14, batch 3500, loss[loss=0.173, simple_loss=0.2483, pruned_loss=0.0489, over 16751.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.262, pruned_loss=0.04693, over 3333993.48 frames. ], batch size: 89, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:00:51,741 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7757, 2.1278, 2.2649, 4.4905, 2.1132, 2.6056, 2.2839, 2.2836], device='cuda:5'), covar=tensor([0.0988, 0.4007, 0.2751, 0.0408, 0.4259, 0.2650, 0.3372, 0.3813], device='cuda:5'), in_proj_covar=tensor([0.0381, 0.0415, 0.0347, 0.0330, 0.0423, 0.0480, 0.0379, 0.0488], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:01:37,089 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.212e+02 2.743e+02 3.344e+02 6.080e+02, threshold=5.486e+02, percent-clipped=2.0 2023-04-29 23:01:39,983 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1195, 4.7585, 5.1086, 5.3171, 5.4935, 4.7670, 5.4841, 5.4385], device='cuda:5'), covar=tensor([0.1612, 0.1314, 0.1659, 0.0691, 0.0477, 0.0855, 0.0433, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0602, 0.0747, 0.0901, 0.0766, 0.0576, 0.0591, 0.0601, 0.0704], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:01:51,753 INFO [train.py:904] (5/8) Epoch 14, batch 3550, loss[loss=0.1602, simple_loss=0.2433, pruned_loss=0.03853, over 16650.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2612, pruned_loss=0.04615, over 3335009.34 frames. ], batch size: 76, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:02:04,693 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4740, 3.7408, 4.1208, 2.2107, 3.2279, 2.8370, 3.9089, 3.8803], device='cuda:5'), covar=tensor([0.0285, 0.0828, 0.0449, 0.1747, 0.0755, 0.0848, 0.0568, 0.0956], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0155, 0.0161, 0.0147, 0.0139, 0.0126, 0.0140, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:02:15,223 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:02:27,913 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:02:43,093 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3666, 5.3503, 5.2538, 4.8251, 4.8292, 5.3423, 5.2322, 4.9513], device='cuda:5'), covar=tensor([0.0562, 0.0439, 0.0246, 0.0281, 0.1110, 0.0342, 0.0263, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0277, 0.0382, 0.0335, 0.0314, 0.0352, 0.0365, 0.0227, 0.0393], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 23:03:02,261 INFO [train.py:904] (5/8) Epoch 14, batch 3600, loss[loss=0.1773, simple_loss=0.2704, pruned_loss=0.04209, over 17018.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2599, pruned_loss=0.04541, over 3327187.03 frames. ], batch size: 50, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:03:22,563 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:03:58,639 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.279e+02 2.683e+02 3.136e+02 7.106e+02, threshold=5.366e+02, percent-clipped=1.0 2023-04-29 23:04:14,268 INFO [train.py:904] (5/8) Epoch 14, batch 3650, loss[loss=0.1928, simple_loss=0.2548, pruned_loss=0.06538, over 16737.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2594, pruned_loss=0.04638, over 3325010.75 frames. ], batch size: 134, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:04:37,764 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2477, 3.4885, 3.5773, 3.5501, 3.5518, 3.4077, 3.4230, 3.4353], device='cuda:5'), covar=tensor([0.0431, 0.0579, 0.0406, 0.0439, 0.0566, 0.0480, 0.0753, 0.0525], device='cuda:5'), in_proj_covar=tensor([0.0373, 0.0394, 0.0389, 0.0371, 0.0437, 0.0416, 0.0510, 0.0329], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-29 23:04:57,349 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:05:08,066 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7230, 3.8285, 2.2500, 4.1339, 2.8791, 4.0720, 2.4583, 3.0661], device='cuda:5'), covar=tensor([0.0213, 0.0336, 0.1443, 0.0224, 0.0716, 0.0575, 0.1205, 0.0567], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0175, 0.0192, 0.0151, 0.0172, 0.0218, 0.0202, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:05:09,383 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 23:05:24,987 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 23:05:31,041 INFO [train.py:904] (5/8) Epoch 14, batch 3700, loss[loss=0.2008, simple_loss=0.2647, pruned_loss=0.06848, over 11643.00 frames. ], tot_loss[loss=0.177, simple_loss=0.258, pruned_loss=0.04799, over 3297958.05 frames. ], batch size: 246, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:05:40,380 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9042, 4.1751, 3.9545, 4.0523, 3.7021, 3.7868, 3.8274, 4.1149], device='cuda:5'), covar=tensor([0.1080, 0.0819, 0.1074, 0.0702, 0.0710, 0.1645, 0.0831, 0.1102], device='cuda:5'), in_proj_covar=tensor([0.0607, 0.0754, 0.0616, 0.0544, 0.0483, 0.0481, 0.0629, 0.0581], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:05:48,762 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6722, 3.7853, 2.2993, 4.0247, 2.9407, 3.9531, 2.3475, 3.0306], device='cuda:5'), covar=tensor([0.0210, 0.0316, 0.1392, 0.0230, 0.0669, 0.0655, 0.1271, 0.0541], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0174, 0.0192, 0.0151, 0.0172, 0.0218, 0.0202, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:06:30,667 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:06:32,147 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.199e+02 2.631e+02 3.126e+02 8.071e+02, threshold=5.262e+02, percent-clipped=3.0 2023-04-29 23:06:46,321 INFO [train.py:904] (5/8) Epoch 14, batch 3750, loss[loss=0.1916, simple_loss=0.2783, pruned_loss=0.05244, over 17246.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2591, pruned_loss=0.04968, over 3286141.44 frames. ], batch size: 45, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:07:30,490 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6312, 2.3650, 1.8031, 2.1510, 2.6875, 2.5530, 2.8087, 2.8183], device='cuda:5'), covar=tensor([0.0145, 0.0298, 0.0438, 0.0353, 0.0177, 0.0237, 0.0178, 0.0227], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0218, 0.0212, 0.0212, 0.0219, 0.0220, 0.0230, 0.0213], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:07:45,971 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 23:08:01,588 INFO [train.py:904] (5/8) Epoch 14, batch 3800, loss[loss=0.1946, simple_loss=0.2735, pruned_loss=0.05791, over 16735.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2609, pruned_loss=0.05122, over 3270143.46 frames. ], batch size: 76, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:08:19,085 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8089, 3.9661, 2.6051, 2.2599, 2.4703, 2.1649, 3.7539, 3.3300], device='cuda:5'), covar=tensor([0.2437, 0.0512, 0.1880, 0.2669, 0.2743, 0.2109, 0.0611, 0.1274], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0263, 0.0292, 0.0292, 0.0289, 0.0234, 0.0278, 0.0316], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 23:08:39,180 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8192, 3.7106, 3.9221, 3.6840, 3.8077, 4.2405, 3.9048, 3.5328], device='cuda:5'), covar=tensor([0.2171, 0.2139, 0.1862, 0.2431, 0.2905, 0.1700, 0.1366, 0.2545], device='cuda:5'), in_proj_covar=tensor([0.0374, 0.0536, 0.0579, 0.0454, 0.0614, 0.0608, 0.0463, 0.0602], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 23:09:00,855 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.313e+02 2.548e+02 3.011e+02 5.523e+02, threshold=5.096e+02, percent-clipped=2.0 2023-04-29 23:09:16,026 INFO [train.py:904] (5/8) Epoch 14, batch 3850, loss[loss=0.1845, simple_loss=0.2558, pruned_loss=0.05661, over 16671.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2604, pruned_loss=0.05191, over 3278633.74 frames. ], batch size: 89, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:09:53,690 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:10:30,042 INFO [train.py:904] (5/8) Epoch 14, batch 3900, loss[loss=0.1749, simple_loss=0.2485, pruned_loss=0.05066, over 16253.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2596, pruned_loss=0.05212, over 3287205.84 frames. ], batch size: 165, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:11:05,209 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:11:30,226 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.144e+02 2.448e+02 3.272e+02 6.091e+02, threshold=4.896e+02, percent-clipped=2.0 2023-04-29 23:11:40,165 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6119, 4.5810, 4.7437, 4.6869, 4.6502, 5.2360, 4.7757, 4.4574], device='cuda:5'), covar=tensor([0.1498, 0.2124, 0.2270, 0.2093, 0.2818, 0.1019, 0.1504, 0.2538], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0541, 0.0585, 0.0457, 0.0618, 0.0612, 0.0467, 0.0608], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 23:11:45,243 INFO [train.py:904] (5/8) Epoch 14, batch 3950, loss[loss=0.206, simple_loss=0.2804, pruned_loss=0.06576, over 15532.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2604, pruned_loss=0.05319, over 3283141.40 frames. ], batch size: 190, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:12:58,432 INFO [train.py:904] (5/8) Epoch 14, batch 4000, loss[loss=0.2029, simple_loss=0.2811, pruned_loss=0.06233, over 16460.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2604, pruned_loss=0.05365, over 3286622.80 frames. ], batch size: 146, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:13:13,434 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1939, 2.6712, 2.1214, 2.3271, 2.9953, 2.5850, 3.0685, 3.1457], device='cuda:5'), covar=tensor([0.0098, 0.0307, 0.0425, 0.0395, 0.0189, 0.0290, 0.0152, 0.0166], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0216, 0.0211, 0.0211, 0.0218, 0.0218, 0.0227, 0.0211], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:13:26,873 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 23:13:41,817 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:47,667 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:55,970 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.249e+02 2.823e+02 3.453e+02 5.778e+02, threshold=5.647e+02, percent-clipped=7.0 2023-04-29 23:13:58,943 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:14:13,538 INFO [train.py:904] (5/8) Epoch 14, batch 4050, loss[loss=0.1885, simple_loss=0.2657, pruned_loss=0.05567, over 16672.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2611, pruned_loss=0.05264, over 3290425.51 frames. ], batch size: 57, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:14:33,525 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2818, 3.2340, 2.5256, 2.0655, 2.1179, 2.1367, 3.2528, 3.0003], device='cuda:5'), covar=tensor([0.2860, 0.0697, 0.1809, 0.2535, 0.2630, 0.1932, 0.0594, 0.1023], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0264, 0.0293, 0.0293, 0.0291, 0.0235, 0.0279, 0.0317], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 23:15:12,513 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:15:24,941 INFO [train.py:904] (5/8) Epoch 14, batch 4100, loss[loss=0.1898, simple_loss=0.275, pruned_loss=0.05224, over 16675.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.262, pruned_loss=0.05165, over 3275484.82 frames. ], batch size: 134, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:30,431 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:16:24,203 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 1.983e+02 2.350e+02 2.941e+02 4.738e+02, threshold=4.699e+02, percent-clipped=0.0 2023-04-29 23:16:40,537 INFO [train.py:904] (5/8) Epoch 14, batch 4150, loss[loss=0.2684, simple_loss=0.3385, pruned_loss=0.09916, over 11535.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2692, pruned_loss=0.0543, over 3229703.74 frames. ], batch size: 248, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:16:45,274 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3790, 3.2775, 2.4554, 2.0302, 2.2625, 2.0943, 3.1911, 2.9316], device='cuda:5'), covar=tensor([0.2699, 0.0776, 0.1897, 0.2528, 0.2446, 0.2161, 0.0679, 0.1184], device='cuda:5'), in_proj_covar=tensor([0.0307, 0.0261, 0.0290, 0.0290, 0.0288, 0.0233, 0.0276, 0.0313], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-29 23:16:58,806 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1856, 3.4542, 3.6437, 2.1310, 3.0764, 2.5027, 3.5963, 3.6220], device='cuda:5'), covar=tensor([0.0238, 0.0707, 0.0498, 0.1767, 0.0742, 0.0842, 0.0560, 0.0893], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0154, 0.0160, 0.0147, 0.0139, 0.0126, 0.0138, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:17:41,305 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:17:55,772 INFO [train.py:904] (5/8) Epoch 14, batch 4200, loss[loss=0.2197, simple_loss=0.2995, pruned_loss=0.06996, over 11072.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2772, pruned_loss=0.05656, over 3198472.04 frames. ], batch size: 247, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:18:26,890 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3184, 2.1141, 2.1134, 4.0329, 2.0141, 2.5041, 2.1962, 2.3109], device='cuda:5'), covar=tensor([0.1060, 0.3836, 0.2652, 0.0401, 0.4276, 0.2294, 0.3164, 0.3199], device='cuda:5'), in_proj_covar=tensor([0.0377, 0.0415, 0.0344, 0.0325, 0.0418, 0.0478, 0.0377, 0.0486], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:18:55,506 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 2.476e+02 2.870e+02 3.638e+02 7.412e+02, threshold=5.739e+02, percent-clipped=7.0 2023-04-29 23:19:00,817 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.56 vs. limit=5.0 2023-04-29 23:19:10,104 INFO [train.py:904] (5/8) Epoch 14, batch 4250, loss[loss=0.1894, simple_loss=0.2882, pruned_loss=0.04527, over 17005.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2802, pruned_loss=0.05615, over 3188743.68 frames. ], batch size: 55, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:19:11,069 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-29 23:19:12,605 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136203.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:20:23,647 INFO [train.py:904] (5/8) Epoch 14, batch 4300, loss[loss=0.2302, simple_loss=0.3059, pruned_loss=0.0773, over 11564.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2802, pruned_loss=0.05485, over 3186638.23 frames. ], batch size: 246, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:20:39,022 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-04-29 23:21:09,060 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0097, 2.0550, 2.1764, 3.5151, 1.9874, 2.3550, 2.2347, 2.2251], device='cuda:5'), covar=tensor([0.1115, 0.3032, 0.2376, 0.0504, 0.3705, 0.2167, 0.2839, 0.3045], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0415, 0.0344, 0.0325, 0.0420, 0.0478, 0.0378, 0.0485], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:21:14,617 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:21:23,664 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.749e+02 2.283e+02 2.653e+02 3.073e+02 4.214e+02, threshold=5.307e+02, percent-clipped=0.0 2023-04-29 23:21:38,132 INFO [train.py:904] (5/8) Epoch 14, batch 4350, loss[loss=0.1958, simple_loss=0.2835, pruned_loss=0.05406, over 17127.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2836, pruned_loss=0.05602, over 3168920.02 frames. ], batch size: 47, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:21:51,131 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7281, 2.9134, 2.4211, 4.1596, 3.2574, 3.8764, 1.4312, 2.7028], device='cuda:5'), covar=tensor([0.1176, 0.0584, 0.1132, 0.0133, 0.0279, 0.0429, 0.1466, 0.0866], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0161, 0.0181, 0.0160, 0.0199, 0.0208, 0.0185, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-29 23:22:27,143 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136334.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:29,185 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:33,166 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:51,138 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:53,921 INFO [train.py:904] (5/8) Epoch 14, batch 4400, loss[loss=0.1807, simple_loss=0.2708, pruned_loss=0.04527, over 17272.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2857, pruned_loss=0.05691, over 3181389.59 frames. ], batch size: 52, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:23:32,654 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1430, 5.0965, 4.9844, 4.6795, 4.6693, 5.0522, 4.8906, 4.7459], device='cuda:5'), covar=tensor([0.0396, 0.0227, 0.0190, 0.0211, 0.0773, 0.0217, 0.0260, 0.0483], device='cuda:5'), in_proj_covar=tensor([0.0256, 0.0353, 0.0311, 0.0290, 0.0326, 0.0337, 0.0211, 0.0361], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:23:51,148 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.242e+02 2.686e+02 3.350e+02 5.764e+02, threshold=5.372e+02, percent-clipped=1.0 2023-04-29 23:23:57,433 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:24:06,509 INFO [train.py:904] (5/8) Epoch 14, batch 4450, loss[loss=0.2154, simple_loss=0.2862, pruned_loss=0.07229, over 11699.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2895, pruned_loss=0.0583, over 3190178.10 frames. ], batch size: 248, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:25:18,609 INFO [train.py:904] (5/8) Epoch 14, batch 4500, loss[loss=0.212, simple_loss=0.2992, pruned_loss=0.06238, over 16927.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2905, pruned_loss=0.05909, over 3192730.31 frames. ], batch size: 116, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:25:19,713 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9462, 2.7227, 2.6206, 1.9661, 2.5199, 2.7088, 2.5376, 1.8856], device='cuda:5'), covar=tensor([0.0342, 0.0064, 0.0055, 0.0298, 0.0093, 0.0092, 0.0097, 0.0335], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0073, 0.0074, 0.0130, 0.0086, 0.0095, 0.0084, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 23:26:04,707 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9861, 5.5617, 5.8380, 5.3983, 5.5019, 6.1762, 5.5992, 5.3026], device='cuda:5'), covar=tensor([0.0874, 0.1614, 0.1862, 0.1871, 0.2600, 0.0793, 0.1413, 0.2433], device='cuda:5'), in_proj_covar=tensor([0.0373, 0.0530, 0.0570, 0.0448, 0.0603, 0.0599, 0.0458, 0.0598], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-29 23:26:18,141 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.032e+02 2.340e+02 2.822e+02 5.483e+02, threshold=4.680e+02, percent-clipped=1.0 2023-04-29 23:26:27,082 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136498.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:26:31,831 INFO [train.py:904] (5/8) Epoch 14, batch 4550, loss[loss=0.2087, simple_loss=0.2928, pruned_loss=0.06234, over 17136.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2914, pruned_loss=0.05984, over 3198795.96 frames. ], batch size: 47, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:26:56,374 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-29 23:27:07,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9117, 4.6801, 4.6199, 5.1408, 5.2678, 4.6841, 5.2665, 5.3151], device='cuda:5'), covar=tensor([0.1488, 0.1257, 0.2156, 0.0730, 0.0646, 0.0918, 0.0715, 0.0695], device='cuda:5'), in_proj_covar=tensor([0.0567, 0.0701, 0.0841, 0.0718, 0.0542, 0.0555, 0.0560, 0.0661], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:27:45,704 INFO [train.py:904] (5/8) Epoch 14, batch 4600, loss[loss=0.1898, simple_loss=0.2789, pruned_loss=0.05039, over 16532.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2922, pruned_loss=0.05977, over 3209030.00 frames. ], batch size: 75, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:28:42,906 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 2.145e+02 2.405e+02 2.956e+02 4.611e+02, threshold=4.810e+02, percent-clipped=0.0 2023-04-29 23:28:57,107 INFO [train.py:904] (5/8) Epoch 14, batch 4650, loss[loss=0.2521, simple_loss=0.3134, pruned_loss=0.09536, over 11654.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2905, pruned_loss=0.05956, over 3213028.44 frames. ], batch size: 246, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:29:52,220 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:29:58,788 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:07,904 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:10,592 INFO [train.py:904] (5/8) Epoch 14, batch 4700, loss[loss=0.191, simple_loss=0.2771, pruned_loss=0.05251, over 16425.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2875, pruned_loss=0.0583, over 3205555.53 frames. ], batch size: 68, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:31:01,339 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:01,488 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6418, 4.6421, 4.5279, 3.8512, 4.6055, 1.7677, 4.3269, 4.3327], device='cuda:5'), covar=tensor([0.0109, 0.0102, 0.0143, 0.0429, 0.0104, 0.2465, 0.0135, 0.0236], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0132, 0.0177, 0.0167, 0.0150, 0.0190, 0.0168, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:31:09,105 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:09,908 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.115e+02 2.410e+02 3.015e+02 5.394e+02, threshold=4.821e+02, percent-clipped=2.0 2023-04-29 23:31:18,046 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136698.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:23,582 INFO [train.py:904] (5/8) Epoch 14, batch 4750, loss[loss=0.185, simple_loss=0.2727, pruned_loss=0.04868, over 16904.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2834, pruned_loss=0.05627, over 3209582.57 frames. ], batch size: 116, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:31:27,827 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:32:03,680 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-04-29 23:32:04,854 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 23:32:15,645 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8433, 1.9932, 2.4326, 2.8005, 2.7113, 3.2808, 1.9234, 3.0937], device='cuda:5'), covar=tensor([0.0180, 0.0396, 0.0272, 0.0266, 0.0245, 0.0130, 0.0455, 0.0104], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0177, 0.0161, 0.0170, 0.0177, 0.0134, 0.0179, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:32:36,929 INFO [train.py:904] (5/8) Epoch 14, batch 4800, loss[loss=0.168, simple_loss=0.2568, pruned_loss=0.03957, over 16623.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2793, pruned_loss=0.05377, over 3210567.87 frames. ], batch size: 62, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:36,436 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 1.970e+02 2.266e+02 2.773e+02 4.879e+02, threshold=4.532e+02, percent-clipped=1.0 2023-04-29 23:33:46,286 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136798.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:33:52,193 INFO [train.py:904] (5/8) Epoch 14, batch 4850, loss[loss=0.2235, simple_loss=0.2992, pruned_loss=0.07392, over 12128.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2802, pruned_loss=0.05357, over 3200085.38 frames. ], batch size: 247, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:59,195 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136806.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:34:56,904 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:35:00,325 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0841, 3.4014, 3.5136, 1.9225, 2.9368, 2.3622, 3.5124, 3.3986], device='cuda:5'), covar=tensor([0.0222, 0.0690, 0.0548, 0.1886, 0.0781, 0.0893, 0.0620, 0.0854], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0152, 0.0161, 0.0147, 0.0139, 0.0126, 0.0139, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:35:02,567 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6219, 4.4569, 4.7080, 4.8867, 5.0338, 4.5307, 5.0014, 5.0335], device='cuda:5'), covar=tensor([0.1544, 0.1259, 0.1432, 0.0634, 0.0489, 0.0785, 0.0510, 0.0562], device='cuda:5'), in_proj_covar=tensor([0.0560, 0.0692, 0.0826, 0.0708, 0.0533, 0.0546, 0.0553, 0.0653], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:35:05,046 INFO [train.py:904] (5/8) Epoch 14, batch 4900, loss[loss=0.1806, simple_loss=0.2742, pruned_loss=0.04348, over 16416.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2791, pruned_loss=0.05187, over 3208719.58 frames. ], batch size: 146, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:35:28,863 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136867.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:36:02,834 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.076e+02 2.406e+02 2.771e+02 5.710e+02, threshold=4.811e+02, percent-clipped=1.0 2023-04-29 23:36:18,498 INFO [train.py:904] (5/8) Epoch 14, batch 4950, loss[loss=0.1967, simple_loss=0.2843, pruned_loss=0.05457, over 16301.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2789, pruned_loss=0.05148, over 3186729.37 frames. ], batch size: 35, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:37:30,154 INFO [train.py:904] (5/8) Epoch 14, batch 5000, loss[loss=0.1883, simple_loss=0.2823, pruned_loss=0.04713, over 16780.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2805, pruned_loss=0.0516, over 3196114.15 frames. ], batch size: 83, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:37:33,425 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 23:38:16,746 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5257, 3.8810, 3.8630, 2.1338, 3.0413, 2.5341, 3.8253, 3.8743], device='cuda:5'), covar=tensor([0.0236, 0.0627, 0.0536, 0.1807, 0.0815, 0.0854, 0.0610, 0.0833], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0153, 0.0162, 0.0148, 0.0140, 0.0127, 0.0139, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:38:24,315 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136991.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:38:25,066 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.271e+02 2.523e+02 3.075e+02 5.093e+02, threshold=5.046e+02, percent-clipped=1.0 2023-04-29 23:38:32,179 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8548, 2.0317, 2.3286, 3.1039, 2.1076, 2.2422, 2.2426, 2.1870], device='cuda:5'), covar=tensor([0.1165, 0.3213, 0.2150, 0.0601, 0.3515, 0.2271, 0.2826, 0.2988], device='cuda:5'), in_proj_covar=tensor([0.0373, 0.0414, 0.0343, 0.0322, 0.0418, 0.0475, 0.0374, 0.0482], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:38:36,302 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136999.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:38:39,702 INFO [train.py:904] (5/8) Epoch 14, batch 5050, loss[loss=0.194, simple_loss=0.2847, pruned_loss=0.05169, over 16765.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2816, pruned_loss=0.05174, over 3208945.24 frames. ], batch size: 83, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:38:44,049 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-29 23:39:31,296 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:39:49,981 INFO [train.py:904] (5/8) Epoch 14, batch 5100, loss[loss=0.1824, simple_loss=0.2683, pruned_loss=0.04825, over 16730.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2797, pruned_loss=0.05089, over 3218874.54 frames. ], batch size: 57, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:40:45,952 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.206e+02 2.593e+02 3.193e+02 6.541e+02, threshold=5.187e+02, percent-clipped=2.0 2023-04-29 23:41:00,265 INFO [train.py:904] (5/8) Epoch 14, batch 5150, loss[loss=0.2009, simple_loss=0.2947, pruned_loss=0.05358, over 16688.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2799, pruned_loss=0.05028, over 3205585.31 frames. ], batch size: 124, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:41:13,200 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-29 23:42:08,572 INFO [train.py:904] (5/8) Epoch 14, batch 5200, loss[loss=0.1756, simple_loss=0.2614, pruned_loss=0.04492, over 16799.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2784, pruned_loss=0.04987, over 3195840.84 frames. ], batch size: 83, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:42:22,364 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:43:03,332 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.223e+02 2.609e+02 3.158e+02 5.537e+02, threshold=5.218e+02, percent-clipped=1.0 2023-04-29 23:43:18,146 INFO [train.py:904] (5/8) Epoch 14, batch 5250, loss[loss=0.1714, simple_loss=0.2597, pruned_loss=0.0416, over 16583.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2757, pruned_loss=0.04938, over 3205153.04 frames. ], batch size: 68, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:44:26,547 INFO [train.py:904] (5/8) Epoch 14, batch 5300, loss[loss=0.1627, simple_loss=0.2521, pruned_loss=0.03666, over 16645.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.272, pruned_loss=0.04787, over 3217099.03 frames. ], batch size: 62, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:44:32,795 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4144, 4.4604, 4.3053, 3.9724, 4.0110, 4.3821, 4.1488, 4.0828], device='cuda:5'), covar=tensor([0.0540, 0.0445, 0.0257, 0.0284, 0.0832, 0.0420, 0.0546, 0.0688], device='cuda:5'), in_proj_covar=tensor([0.0251, 0.0348, 0.0305, 0.0286, 0.0321, 0.0335, 0.0206, 0.0358], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:45:23,604 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.002e+02 2.371e+02 2.932e+02 5.680e+02, threshold=4.741e+02, percent-clipped=1.0 2023-04-29 23:45:33,751 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137299.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:45:36,785 INFO [train.py:904] (5/8) Epoch 14, batch 5350, loss[loss=0.1924, simple_loss=0.2831, pruned_loss=0.05087, over 16683.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2707, pruned_loss=0.04703, over 3228087.52 frames. ], batch size: 62, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:46:40,458 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137347.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:46:46,694 INFO [train.py:904] (5/8) Epoch 14, batch 5400, loss[loss=0.2019, simple_loss=0.297, pruned_loss=0.05342, over 16682.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2736, pruned_loss=0.04804, over 3225751.83 frames. ], batch size: 134, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:47:27,715 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-29 23:47:37,550 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7581, 2.8861, 2.8935, 4.9016, 3.8122, 4.3263, 1.6928, 3.1545], device='cuda:5'), covar=tensor([0.1254, 0.0701, 0.1007, 0.0108, 0.0274, 0.0329, 0.1440, 0.0719], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0163, 0.0185, 0.0160, 0.0200, 0.0208, 0.0187, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:47:46,870 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.156e+02 2.492e+02 3.031e+02 4.895e+02, threshold=4.984e+02, percent-clipped=1.0 2023-04-29 23:48:02,465 INFO [train.py:904] (5/8) Epoch 14, batch 5450, loss[loss=0.2485, simple_loss=0.3246, pruned_loss=0.08623, over 15285.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2765, pruned_loss=0.04988, over 3204341.60 frames. ], batch size: 191, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:49:19,972 INFO [train.py:904] (5/8) Epoch 14, batch 5500, loss[loss=0.2115, simple_loss=0.293, pruned_loss=0.06498, over 16547.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2837, pruned_loss=0.05435, over 3165651.24 frames. ], batch size: 68, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:49:32,450 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8941, 4.0261, 3.8005, 3.5703, 3.4425, 3.9491, 3.6428, 3.6211], device='cuda:5'), covar=tensor([0.0622, 0.0606, 0.0316, 0.0326, 0.0887, 0.0548, 0.1021, 0.0638], device='cuda:5'), in_proj_covar=tensor([0.0255, 0.0353, 0.0306, 0.0288, 0.0325, 0.0340, 0.0207, 0.0361], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-29 23:49:35,308 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:49:37,428 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5472, 3.0426, 3.0699, 1.8190, 2.6564, 2.1277, 3.1999, 3.3043], device='cuda:5'), covar=tensor([0.0259, 0.0687, 0.0580, 0.1986, 0.0858, 0.0952, 0.0606, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0152, 0.0162, 0.0147, 0.0139, 0.0127, 0.0139, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:50:23,840 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.042e+02 3.226e+02 3.760e+02 4.566e+02 8.180e+02, threshold=7.520e+02, percent-clipped=15.0 2023-04-29 23:50:31,362 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 23:50:37,000 INFO [train.py:904] (5/8) Epoch 14, batch 5550, loss[loss=0.2762, simple_loss=0.3423, pruned_loss=0.1051, over 15369.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2924, pruned_loss=0.061, over 3130737.54 frames. ], batch size: 191, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:50:46,776 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 23:50:51,277 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:51:55,251 INFO [train.py:904] (5/8) Epoch 14, batch 5600, loss[loss=0.2978, simple_loss=0.3505, pruned_loss=0.1226, over 11014.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2971, pruned_loss=0.0651, over 3119108.19 frames. ], batch size: 247, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:53:03,061 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.471e+02 3.513e+02 4.095e+02 5.435e+02 1.324e+03, threshold=8.190e+02, percent-clipped=8.0 2023-04-29 23:53:17,474 INFO [train.py:904] (5/8) Epoch 14, batch 5650, loss[loss=0.2238, simple_loss=0.3067, pruned_loss=0.07043, over 16759.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3032, pruned_loss=0.07031, over 3071882.05 frames. ], batch size: 83, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:54:38,734 INFO [train.py:904] (5/8) Epoch 14, batch 5700, loss[loss=0.2255, simple_loss=0.3101, pruned_loss=0.07048, over 16516.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3036, pruned_loss=0.07113, over 3061865.66 frames. ], batch size: 68, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:55:45,159 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.087e+02 3.198e+02 3.940e+02 4.824e+02 8.425e+02, threshold=7.880e+02, percent-clipped=1.0 2023-04-29 23:55:59,933 INFO [train.py:904] (5/8) Epoch 14, batch 5750, loss[loss=0.2583, simple_loss=0.3302, pruned_loss=0.09323, over 11537.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3066, pruned_loss=0.07243, over 3059346.90 frames. ], batch size: 247, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:57:21,152 INFO [train.py:904] (5/8) Epoch 14, batch 5800, loss[loss=0.2398, simple_loss=0.3084, pruned_loss=0.08564, over 12245.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.305, pruned_loss=0.07019, over 3070557.76 frames. ], batch size: 248, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:57:21,911 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.18 vs. limit=5.0 2023-04-29 23:58:15,919 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8151, 3.9658, 2.3756, 4.6699, 2.8314, 4.5542, 2.3325, 3.0072], device='cuda:5'), covar=tensor([0.0235, 0.0329, 0.1569, 0.0166, 0.0837, 0.0431, 0.1636, 0.0807], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0167, 0.0188, 0.0138, 0.0167, 0.0206, 0.0196, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:58:26,269 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.930e+02 3.076e+02 3.658e+02 4.339e+02 6.692e+02, threshold=7.316e+02, percent-clipped=0.0 2023-04-29 23:58:31,730 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7400, 2.7968, 2.5069, 4.6251, 3.4890, 4.2038, 1.5664, 3.0789], device='cuda:5'), covar=tensor([0.1327, 0.0720, 0.1206, 0.0157, 0.0387, 0.0380, 0.1516, 0.0779], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0162, 0.0184, 0.0159, 0.0201, 0.0208, 0.0186, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-29 23:58:41,175 INFO [train.py:904] (5/8) Epoch 14, batch 5850, loss[loss=0.2024, simple_loss=0.2927, pruned_loss=0.05609, over 16409.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3026, pruned_loss=0.06814, over 3079981.78 frames. ], batch size: 68, lr: 4.85e-03, grad_scale: 8.0 2023-04-29 23:59:01,276 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 00:00:03,144 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 00:00:03,520 INFO [train.py:904] (5/8) Epoch 14, batch 5900, loss[loss=0.198, simple_loss=0.2835, pruned_loss=0.0563, over 16150.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.3017, pruned_loss=0.0675, over 3094774.13 frames. ], batch size: 165, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:01:10,264 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137892.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:01:10,917 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.749e+02 3.330e+02 4.231e+02 8.888e+02, threshold=6.660e+02, percent-clipped=2.0 2023-04-30 00:01:15,141 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:01:24,308 INFO [train.py:904] (5/8) Epoch 14, batch 5950, loss[loss=0.1912, simple_loss=0.2879, pruned_loss=0.04722, over 17226.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.3024, pruned_loss=0.06609, over 3118334.77 frames. ], batch size: 45, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:01:44,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3193, 3.7677, 3.6029, 2.0790, 3.0931, 2.6585, 3.6775, 3.8968], device='cuda:5'), covar=tensor([0.0263, 0.0618, 0.0582, 0.1886, 0.0773, 0.0844, 0.0632, 0.0783], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0150, 0.0161, 0.0146, 0.0138, 0.0126, 0.0138, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 00:01:52,212 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0257, 5.0067, 4.7914, 4.2143, 4.9076, 1.9717, 4.6549, 4.6213], device='cuda:5'), covar=tensor([0.0077, 0.0065, 0.0144, 0.0323, 0.0079, 0.2260, 0.0105, 0.0171], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0129, 0.0174, 0.0164, 0.0147, 0.0187, 0.0163, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:01:56,499 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4517, 3.9120, 3.8984, 2.7565, 3.5534, 3.8940, 3.5970, 2.1420], device='cuda:5'), covar=tensor([0.0413, 0.0035, 0.0037, 0.0286, 0.0078, 0.0098, 0.0067, 0.0388], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0074, 0.0074, 0.0130, 0.0087, 0.0098, 0.0084, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 00:02:44,552 INFO [train.py:904] (5/8) Epoch 14, batch 6000, loss[loss=0.178, simple_loss=0.2682, pruned_loss=0.04389, over 16823.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.3011, pruned_loss=0.06522, over 3125051.36 frames. ], batch size: 102, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:02:44,552 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 00:02:55,352 INFO [train.py:938] (5/8) Epoch 14, validation: loss=0.1574, simple_loss=0.2706, pruned_loss=0.0221, over 944034.00 frames. 2023-04-30 00:02:55,353 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 00:02:57,574 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:01,587 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137956.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:59,085 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.967e+02 2.895e+02 3.490e+02 4.190e+02 7.002e+02, threshold=6.980e+02, percent-clipped=3.0 2023-04-30 00:04:18,781 INFO [train.py:904] (5/8) Epoch 14, batch 6050, loss[loss=0.199, simple_loss=0.2993, pruned_loss=0.0493, over 16695.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2993, pruned_loss=0.06423, over 3135968.77 frames. ], batch size: 134, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:05:40,156 INFO [train.py:904] (5/8) Epoch 14, batch 6100, loss[loss=0.2442, simple_loss=0.3136, pruned_loss=0.08736, over 11463.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2994, pruned_loss=0.06381, over 3123927.87 frames. ], batch size: 248, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:05:41,065 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5100, 5.8615, 5.5429, 5.6213, 5.2618, 5.1694, 5.2345, 5.9571], device='cuda:5'), covar=tensor([0.1251, 0.0808, 0.1028, 0.0754, 0.0892, 0.0701, 0.1077, 0.0798], device='cuda:5'), in_proj_covar=tensor([0.0581, 0.0716, 0.0593, 0.0523, 0.0456, 0.0465, 0.0601, 0.0553], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:06:45,101 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.919e+02 3.731e+02 4.841e+02 8.594e+02, threshold=7.462e+02, percent-clipped=5.0 2023-04-30 00:06:46,886 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0749, 1.5485, 1.9104, 1.9702, 2.1524, 2.3675, 1.5551, 2.2394], device='cuda:5'), covar=tensor([0.0180, 0.0361, 0.0219, 0.0266, 0.0240, 0.0138, 0.0398, 0.0096], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0176, 0.0161, 0.0167, 0.0176, 0.0133, 0.0180, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:06:59,058 INFO [train.py:904] (5/8) Epoch 14, batch 6150, loss[loss=0.2214, simple_loss=0.297, pruned_loss=0.07292, over 15470.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2974, pruned_loss=0.06351, over 3136335.43 frames. ], batch size: 190, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:08:02,232 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5537, 2.6497, 1.7192, 2.7381, 2.1457, 2.7684, 1.9092, 2.2676], device='cuda:5'), covar=tensor([0.0294, 0.0355, 0.1503, 0.0217, 0.0671, 0.0500, 0.1393, 0.0664], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0167, 0.0187, 0.0138, 0.0166, 0.0206, 0.0195, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 00:08:16,190 INFO [train.py:904] (5/8) Epoch 14, batch 6200, loss[loss=0.2021, simple_loss=0.2864, pruned_loss=0.05886, over 16211.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.296, pruned_loss=0.06355, over 3123147.01 frames. ], batch size: 165, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:09:13,947 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 00:09:17,892 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.947e+02 3.424e+02 4.172e+02 1.390e+03, threshold=6.848e+02, percent-clipped=3.0 2023-04-30 00:09:32,462 INFO [train.py:904] (5/8) Epoch 14, batch 6250, loss[loss=0.2433, simple_loss=0.324, pruned_loss=0.08132, over 15481.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2952, pruned_loss=0.06309, over 3130131.15 frames. ], batch size: 191, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:09:40,337 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0439, 4.2164, 2.3905, 4.8317, 3.0095, 4.7292, 2.6480, 3.1267], device='cuda:5'), covar=tensor([0.0223, 0.0286, 0.1687, 0.0195, 0.0826, 0.0497, 0.1484, 0.0780], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0167, 0.0187, 0.0138, 0.0166, 0.0207, 0.0195, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 00:10:42,331 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:10:46,668 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138251.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:10:47,496 INFO [train.py:904] (5/8) Epoch 14, batch 6300, loss[loss=0.2098, simple_loss=0.2946, pruned_loss=0.06252, over 16239.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.295, pruned_loss=0.06238, over 3127525.35 frames. ], batch size: 165, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:11:17,695 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:11:52,437 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.960e+02 2.966e+02 3.499e+02 4.315e+02 9.401e+02, threshold=6.998e+02, percent-clipped=7.0 2023-04-30 00:12:05,890 INFO [train.py:904] (5/8) Epoch 14, batch 6350, loss[loss=0.2671, simple_loss=0.3308, pruned_loss=0.1017, over 11222.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2961, pruned_loss=0.0634, over 3139650.90 frames. ], batch size: 247, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:12:37,567 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:12:39,955 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 00:12:50,087 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:13:05,071 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8042, 1.3495, 1.6396, 1.5938, 1.7486, 1.8002, 1.5998, 1.7175], device='cuda:5'), covar=tensor([0.0162, 0.0269, 0.0151, 0.0186, 0.0183, 0.0132, 0.0283, 0.0094], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0178, 0.0162, 0.0168, 0.0176, 0.0134, 0.0180, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:13:22,160 INFO [train.py:904] (5/8) Epoch 14, batch 6400, loss[loss=0.1749, simple_loss=0.2633, pruned_loss=0.04319, over 16784.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.297, pruned_loss=0.06509, over 3117587.71 frames. ], batch size: 83, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:14:09,362 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138383.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:14:23,619 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.151e+02 3.125e+02 3.740e+02 4.766e+02 1.292e+03, threshold=7.481e+02, percent-clipped=6.0 2023-04-30 00:14:25,954 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7761, 2.1732, 1.6641, 2.0116, 2.5447, 2.1824, 2.6334, 2.8129], device='cuda:5'), covar=tensor([0.0149, 0.0327, 0.0483, 0.0374, 0.0205, 0.0306, 0.0249, 0.0201], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0209, 0.0204, 0.0204, 0.0211, 0.0210, 0.0215, 0.0204], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:14:37,182 INFO [train.py:904] (5/8) Epoch 14, batch 6450, loss[loss=0.2239, simple_loss=0.3138, pruned_loss=0.06697, over 17033.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2969, pruned_loss=0.0648, over 3110918.65 frames. ], batch size: 41, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:15:54,750 INFO [train.py:904] (5/8) Epoch 14, batch 6500, loss[loss=0.21, simple_loss=0.2941, pruned_loss=0.063, over 16537.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2945, pruned_loss=0.06376, over 3122619.71 frames. ], batch size: 68, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:16:49,911 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9356, 3.4923, 3.6203, 1.5778, 3.6756, 3.9583, 3.1204, 2.5989], device='cuda:5'), covar=tensor([0.1246, 0.0155, 0.0144, 0.1453, 0.0075, 0.0110, 0.0317, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0105, 0.0090, 0.0139, 0.0073, 0.0114, 0.0124, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 00:16:59,452 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.094e+02 2.847e+02 3.325e+02 4.301e+02 7.980e+02, threshold=6.650e+02, percent-clipped=1.0 2023-04-30 00:17:12,097 INFO [train.py:904] (5/8) Epoch 14, batch 6550, loss[loss=0.2377, simple_loss=0.3261, pruned_loss=0.07464, over 17058.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2975, pruned_loss=0.06403, over 3133801.73 frames. ], batch size: 53, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:18:22,248 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:25,783 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:26,473 INFO [train.py:904] (5/8) Epoch 14, batch 6600, loss[loss=0.2005, simple_loss=0.2936, pruned_loss=0.05369, over 16744.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2997, pruned_loss=0.06445, over 3141457.34 frames. ], batch size: 124, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:19:29,571 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.801e+02 3.547e+02 4.436e+02 1.538e+03, threshold=7.095e+02, percent-clipped=5.0 2023-04-30 00:19:33,988 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:19:38,208 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138599.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:19:42,443 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-30 00:19:42,718 INFO [train.py:904] (5/8) Epoch 14, batch 6650, loss[loss=0.185, simple_loss=0.2707, pruned_loss=0.04968, over 16491.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.3006, pruned_loss=0.06608, over 3118286.99 frames. ], batch size: 68, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:20:19,708 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:20:58,176 INFO [train.py:904] (5/8) Epoch 14, batch 6700, loss[loss=0.2441, simple_loss=0.3202, pruned_loss=0.08402, over 16426.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2992, pruned_loss=0.06593, over 3125019.09 frames. ], batch size: 35, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:21:06,491 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9638, 2.5392, 2.6828, 1.8256, 2.7484, 2.8248, 2.4144, 2.3394], device='cuda:5'), covar=tensor([0.0788, 0.0231, 0.0198, 0.0948, 0.0098, 0.0210, 0.0453, 0.0441], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0105, 0.0089, 0.0139, 0.0073, 0.0113, 0.0124, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 00:21:09,613 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5718, 2.5635, 1.8960, 2.6761, 2.0923, 2.6933, 2.0703, 2.3412], device='cuda:5'), covar=tensor([0.0313, 0.0361, 0.1224, 0.0184, 0.0601, 0.0453, 0.1197, 0.0618], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0168, 0.0188, 0.0138, 0.0166, 0.0208, 0.0196, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 00:21:12,763 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5460, 1.6333, 2.0422, 2.4575, 2.5157, 2.8031, 1.5994, 2.6991], device='cuda:5'), covar=tensor([0.0152, 0.0417, 0.0273, 0.0232, 0.0229, 0.0141, 0.0477, 0.0112], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0174, 0.0159, 0.0165, 0.0172, 0.0131, 0.0177, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-30 00:21:40,472 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:44,962 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:52,996 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:22:03,785 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.354e+02 4.045e+02 5.286e+02 1.404e+03, threshold=8.090e+02, percent-clipped=8.0 2023-04-30 00:22:09,177 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:22:15,585 INFO [train.py:904] (5/8) Epoch 14, batch 6750, loss[loss=0.1884, simple_loss=0.27, pruned_loss=0.0534, over 17123.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2976, pruned_loss=0.06567, over 3126584.33 frames. ], batch size: 47, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:22:32,532 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-04-30 00:23:04,909 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9541, 3.2835, 3.4424, 1.9615, 2.7821, 2.1741, 3.4868, 3.4413], device='cuda:5'), covar=tensor([0.0250, 0.0774, 0.0562, 0.1996, 0.0864, 0.0994, 0.0590, 0.0936], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0150, 0.0160, 0.0145, 0.0137, 0.0125, 0.0138, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 00:23:15,033 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:22,975 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:29,053 INFO [train.py:904] (5/8) Epoch 14, batch 6800, loss[loss=0.2218, simple_loss=0.3092, pruned_loss=0.0672, over 16460.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2977, pruned_loss=0.06553, over 3133160.23 frames. ], batch size: 68, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:23:38,508 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138758.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:24:05,699 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 00:24:34,711 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.812e+02 3.433e+02 4.059e+02 1.030e+03, threshold=6.866e+02, percent-clipped=2.0 2023-04-30 00:24:46,920 INFO [train.py:904] (5/8) Epoch 14, batch 6850, loss[loss=0.2118, simple_loss=0.3036, pruned_loss=0.06, over 16303.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2983, pruned_loss=0.06538, over 3139178.46 frames. ], batch size: 165, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:25:04,826 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3876, 4.2064, 4.4494, 4.6276, 4.7720, 4.2987, 4.6730, 4.7271], device='cuda:5'), covar=tensor([0.1720, 0.1167, 0.1425, 0.0647, 0.0557, 0.1009, 0.0798, 0.0591], device='cuda:5'), in_proj_covar=tensor([0.0563, 0.0694, 0.0834, 0.0710, 0.0538, 0.0554, 0.0561, 0.0660], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:26:02,486 INFO [train.py:904] (5/8) Epoch 14, batch 6900, loss[loss=0.255, simple_loss=0.3288, pruned_loss=0.0906, over 17027.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.3004, pruned_loss=0.06459, over 3151620.01 frames. ], batch size: 53, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:26:09,718 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 00:27:10,350 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.725e+02 2.815e+02 3.844e+02 4.824e+02 1.140e+03, threshold=7.689e+02, percent-clipped=2.0 2023-04-30 00:27:22,087 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7447, 5.0386, 5.2886, 5.0516, 5.1185, 5.6487, 5.1427, 4.9322], device='cuda:5'), covar=tensor([0.1039, 0.1625, 0.2140, 0.1806, 0.2156, 0.0850, 0.1478, 0.2259], device='cuda:5'), in_proj_covar=tensor([0.0376, 0.0528, 0.0579, 0.0448, 0.0599, 0.0599, 0.0457, 0.0605], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 00:27:22,852 INFO [train.py:904] (5/8) Epoch 14, batch 6950, loss[loss=0.1955, simple_loss=0.2801, pruned_loss=0.0555, over 16909.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3024, pruned_loss=0.06656, over 3132857.40 frames. ], batch size: 90, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:27:33,824 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:28:00,035 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:28:38,224 INFO [train.py:904] (5/8) Epoch 14, batch 7000, loss[loss=0.2112, simple_loss=0.307, pruned_loss=0.05767, over 16237.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3025, pruned_loss=0.06559, over 3137647.77 frames. ], batch size: 165, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:29:06,869 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 00:29:12,849 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:18,528 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:22,356 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7670, 1.2562, 1.6410, 1.5797, 1.7986, 1.8602, 1.5668, 1.7761], device='cuda:5'), covar=tensor([0.0174, 0.0293, 0.0155, 0.0212, 0.0178, 0.0132, 0.0278, 0.0101], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0174, 0.0159, 0.0165, 0.0173, 0.0131, 0.0176, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-30 00:29:42,507 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.027e+02 2.958e+02 3.558e+02 4.270e+02 7.816e+02, threshold=7.117e+02, percent-clipped=1.0 2023-04-30 00:29:51,032 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8866, 2.6988, 2.6480, 1.8449, 2.5168, 2.7141, 2.5265, 1.8235], device='cuda:5'), covar=tensor([0.0377, 0.0064, 0.0070, 0.0306, 0.0111, 0.0101, 0.0084, 0.0358], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0073, 0.0074, 0.0131, 0.0087, 0.0097, 0.0083, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 00:29:55,157 INFO [train.py:904] (5/8) Epoch 14, batch 7050, loss[loss=0.2451, simple_loss=0.3134, pruned_loss=0.08843, over 11581.00 frames. ], tot_loss[loss=0.217, simple_loss=0.3029, pruned_loss=0.06553, over 3135421.51 frames. ], batch size: 248, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:30:33,408 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:50,686 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:58,268 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139042.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:31:07,627 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7279, 1.5579, 2.1587, 2.5749, 2.6358, 2.9502, 1.5800, 2.7487], device='cuda:5'), covar=tensor([0.0141, 0.0452, 0.0266, 0.0223, 0.0187, 0.0130, 0.0528, 0.0106], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0174, 0.0159, 0.0164, 0.0172, 0.0130, 0.0176, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-30 00:31:13,408 INFO [train.py:904] (5/8) Epoch 14, batch 7100, loss[loss=0.2413, simple_loss=0.2992, pruned_loss=0.0917, over 11257.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3024, pruned_loss=0.0662, over 3118221.73 frames. ], batch size: 246, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:31:15,067 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139053.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:31:15,635 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 00:32:18,350 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.051e+02 2.828e+02 3.507e+02 4.047e+02 5.820e+02, threshold=7.014e+02, percent-clipped=0.0 2023-04-30 00:32:31,532 INFO [train.py:904] (5/8) Epoch 14, batch 7150, loss[loss=0.2064, simple_loss=0.2948, pruned_loss=0.05906, over 16901.00 frames. ], tot_loss[loss=0.217, simple_loss=0.3009, pruned_loss=0.06654, over 3107733.13 frames. ], batch size: 90, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:33:34,452 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3998, 3.3508, 3.3589, 2.4499, 3.2673, 2.0494, 2.9831, 2.5276], device='cuda:5'), covar=tensor([0.0204, 0.0146, 0.0202, 0.0402, 0.0128, 0.2625, 0.0182, 0.0312], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0129, 0.0174, 0.0162, 0.0146, 0.0189, 0.0163, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:33:47,876 INFO [train.py:904] (5/8) Epoch 14, batch 7200, loss[loss=0.2233, simple_loss=0.3035, pruned_loss=0.07152, over 11512.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2985, pruned_loss=0.06518, over 3083997.68 frames. ], batch size: 247, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:34:55,353 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.878e+02 2.722e+02 3.151e+02 3.906e+02 1.100e+03, threshold=6.302e+02, percent-clipped=3.0 2023-04-30 00:35:07,127 INFO [train.py:904] (5/8) Epoch 14, batch 7250, loss[loss=0.2258, simple_loss=0.2918, pruned_loss=0.07992, over 11431.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2964, pruned_loss=0.06454, over 3062428.79 frames. ], batch size: 248, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:35:24,878 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5396, 4.1970, 4.2131, 2.8960, 3.6810, 4.2038, 3.7410, 2.5156], device='cuda:5'), covar=tensor([0.0445, 0.0032, 0.0037, 0.0299, 0.0085, 0.0093, 0.0062, 0.0346], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0072, 0.0073, 0.0129, 0.0086, 0.0096, 0.0083, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 00:36:01,902 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:36:22,194 INFO [train.py:904] (5/8) Epoch 14, batch 7300, loss[loss=0.2123, simple_loss=0.3011, pruned_loss=0.06181, over 16457.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2957, pruned_loss=0.06388, over 3068145.74 frames. ], batch size: 146, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:36:43,288 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 00:37:00,590 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:20,462 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139289.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:29,630 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.078e+02 2.984e+02 3.521e+02 4.495e+02 8.579e+02, threshold=7.042e+02, percent-clipped=5.0 2023-04-30 00:37:37,694 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:40,551 INFO [train.py:904] (5/8) Epoch 14, batch 7350, loss[loss=0.2149, simple_loss=0.3, pruned_loss=0.06494, over 16817.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2967, pruned_loss=0.06503, over 3060795.60 frames. ], batch size: 83, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:38:36,965 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:37,007 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 00:38:45,011 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:57,262 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:59,361 INFO [train.py:904] (5/8) Epoch 14, batch 7400, loss[loss=0.2164, simple_loss=0.3018, pruned_loss=0.06555, over 15261.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2979, pruned_loss=0.06591, over 3046663.58 frames. ], batch size: 190, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:39:01,649 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:39:53,001 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:00,778 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:08,331 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.794e+02 3.378e+02 4.039e+02 6.978e+02, threshold=6.757e+02, percent-clipped=0.0 2023-04-30 00:40:18,659 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:19,493 INFO [train.py:904] (5/8) Epoch 14, batch 7450, loss[loss=0.2558, simple_loss=0.3167, pruned_loss=0.09751, over 11337.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2998, pruned_loss=0.06728, over 3039196.89 frames. ], batch size: 246, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:41:43,395 INFO [train.py:904] (5/8) Epoch 14, batch 7500, loss[loss=0.2142, simple_loss=0.2955, pruned_loss=0.06644, over 16318.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2994, pruned_loss=0.06637, over 3041805.07 frames. ], batch size: 165, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:42:53,297 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 3.028e+02 3.537e+02 4.351e+02 9.459e+02, threshold=7.073e+02, percent-clipped=3.0 2023-04-30 00:43:02,955 INFO [train.py:904] (5/8) Epoch 14, batch 7550, loss[loss=0.2385, simple_loss=0.2987, pruned_loss=0.08911, over 11513.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2996, pruned_loss=0.0675, over 3016128.16 frames. ], batch size: 248, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:43:03,983 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3922, 4.6547, 4.4597, 4.4302, 4.1570, 4.1739, 4.1961, 4.7024], device='cuda:5'), covar=tensor([0.1061, 0.0818, 0.0887, 0.0811, 0.0796, 0.1328, 0.0938, 0.0893], device='cuda:5'), in_proj_covar=tensor([0.0585, 0.0718, 0.0594, 0.0520, 0.0454, 0.0469, 0.0601, 0.0552], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:44:19,041 INFO [train.py:904] (5/8) Epoch 14, batch 7600, loss[loss=0.2014, simple_loss=0.2939, pruned_loss=0.0545, over 16759.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2985, pruned_loss=0.06748, over 3013627.23 frames. ], batch size: 83, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:44:39,745 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:44:48,990 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:44:56,941 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7254, 3.8338, 2.2738, 4.4198, 2.7288, 4.3662, 2.3216, 2.9680], device='cuda:5'), covar=tensor([0.0261, 0.0340, 0.1624, 0.0174, 0.0764, 0.0422, 0.1577, 0.0759], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0169, 0.0190, 0.0139, 0.0166, 0.0208, 0.0197, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 00:45:06,323 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:22,774 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 00:45:23,473 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139594.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:24,243 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.745e+02 3.435e+02 4.599e+02 7.932e+02, threshold=6.869e+02, percent-clipped=2.0 2023-04-30 00:45:34,179 INFO [train.py:904] (5/8) Epoch 14, batch 7650, loss[loss=0.2232, simple_loss=0.3117, pruned_loss=0.06737, over 16728.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2981, pruned_loss=0.06724, over 3026936.38 frames. ], batch size: 134, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:45:50,717 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:19,161 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 00:46:19,276 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:37,208 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:39,486 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:49,455 INFO [train.py:904] (5/8) Epoch 14, batch 7700, loss[loss=0.2173, simple_loss=0.302, pruned_loss=0.06628, over 15337.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2979, pruned_loss=0.06745, over 3038970.75 frames. ], batch size: 190, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:47:16,553 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-30 00:47:57,378 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.201e+02 3.391e+02 3.997e+02 4.581e+02 8.153e+02, threshold=7.994e+02, percent-clipped=6.0 2023-04-30 00:48:06,811 INFO [train.py:904] (5/8) Epoch 14, batch 7750, loss[loss=0.2147, simple_loss=0.3049, pruned_loss=0.06219, over 16293.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2983, pruned_loss=0.06774, over 3016752.49 frames. ], batch size: 165, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:48:10,116 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1983, 2.1303, 2.2815, 3.9664, 2.0239, 2.5118, 2.2239, 2.2622], device='cuda:5'), covar=tensor([0.1145, 0.3368, 0.2462, 0.0488, 0.4099, 0.2470, 0.3360, 0.3305], device='cuda:5'), in_proj_covar=tensor([0.0370, 0.0410, 0.0341, 0.0320, 0.0420, 0.0470, 0.0374, 0.0477], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:49:14,986 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7878, 5.0174, 5.1994, 4.9928, 5.0385, 5.5850, 5.0388, 4.8206], device='cuda:5'), covar=tensor([0.0930, 0.1822, 0.2265, 0.1792, 0.2304, 0.0890, 0.1532, 0.2154], device='cuda:5'), in_proj_covar=tensor([0.0376, 0.0525, 0.0579, 0.0447, 0.0600, 0.0601, 0.0456, 0.0608], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 00:49:24,622 INFO [train.py:904] (5/8) Epoch 14, batch 7800, loss[loss=0.2049, simple_loss=0.2897, pruned_loss=0.0601, over 16920.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2986, pruned_loss=0.0675, over 3047148.20 frames. ], batch size: 109, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:51,621 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:18,963 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9040, 2.0391, 2.3603, 3.1135, 2.1322, 2.2508, 2.2888, 2.2097], device='cuda:5'), covar=tensor([0.1177, 0.3271, 0.2123, 0.0636, 0.3845, 0.2299, 0.2798, 0.2977], device='cuda:5'), in_proj_covar=tensor([0.0368, 0.0408, 0.0338, 0.0317, 0.0417, 0.0467, 0.0371, 0.0473], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:50:22,440 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139789.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:32,563 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.773e+02 2.870e+02 3.646e+02 4.608e+02 1.262e+03, threshold=7.291e+02, percent-clipped=3.0 2023-04-30 00:50:41,165 INFO [train.py:904] (5/8) Epoch 14, batch 7850, loss[loss=0.2311, simple_loss=0.3054, pruned_loss=0.07837, over 11568.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2993, pruned_loss=0.06767, over 3027893.63 frames. ], batch size: 247, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:50:46,026 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:48,420 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:24,816 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,681 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,702 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:56,202 INFO [train.py:904] (5/8) Epoch 14, batch 7900, loss[loss=0.2302, simple_loss=0.3133, pruned_loss=0.0735, over 15176.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2983, pruned_loss=0.06714, over 3019878.27 frames. ], batch size: 190, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:52:15,821 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:52:18,360 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:01,551 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139894.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:03,659 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.073e+02 2.924e+02 3.560e+02 4.104e+02 6.335e+02, threshold=7.120e+02, percent-clipped=0.0 2023-04-30 00:53:12,909 INFO [train.py:904] (5/8) Epoch 14, batch 7950, loss[loss=0.2444, simple_loss=0.3073, pruned_loss=0.0907, over 11803.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2992, pruned_loss=0.06799, over 3014435.33 frames. ], batch size: 248, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:53:19,716 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0806, 2.0106, 2.0787, 3.5515, 2.0152, 2.3384, 2.1506, 2.1606], device='cuda:5'), covar=tensor([0.1192, 0.3426, 0.2688, 0.0551, 0.4054, 0.2366, 0.3157, 0.3246], device='cuda:5'), in_proj_covar=tensor([0.0369, 0.0409, 0.0338, 0.0317, 0.0418, 0.0469, 0.0372, 0.0474], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:53:27,997 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:51,348 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:59,022 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:54:08,420 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:13,648 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:18,377 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:28,361 INFO [train.py:904] (5/8) Epoch 14, batch 8000, loss[loss=0.2074, simple_loss=0.2919, pruned_loss=0.06145, over 16779.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2998, pruned_loss=0.06869, over 3017840.13 frames. ], batch size: 124, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:55:10,564 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139980.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:30,359 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:30,509 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9507, 4.9133, 4.7409, 4.3861, 4.3403, 4.7838, 4.8029, 4.4729], device='cuda:5'), covar=tensor([0.0624, 0.0578, 0.0309, 0.0305, 0.1175, 0.0569, 0.0313, 0.0705], device='cuda:5'), in_proj_covar=tensor([0.0251, 0.0344, 0.0300, 0.0277, 0.0312, 0.0327, 0.0205, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:55:32,172 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 00:55:34,110 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 3.021e+02 3.514e+02 4.508e+02 9.326e+02, threshold=7.028e+02, percent-clipped=1.0 2023-04-30 00:55:46,033 INFO [train.py:904] (5/8) Epoch 14, batch 8050, loss[loss=0.2096, simple_loss=0.3008, pruned_loss=0.05924, over 16876.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2992, pruned_loss=0.06769, over 3043326.46 frames. ], batch size: 83, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:56:48,495 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 00:56:59,051 INFO [train.py:904] (5/8) Epoch 14, batch 8100, loss[loss=0.2207, simple_loss=0.3024, pruned_loss=0.06956, over 16932.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2986, pruned_loss=0.06699, over 3060460.14 frames. ], batch size: 109, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:57:57,852 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2929, 1.4217, 1.9620, 2.0921, 2.2353, 2.4104, 1.5738, 2.3269], device='cuda:5'), covar=tensor([0.0160, 0.0443, 0.0219, 0.0250, 0.0232, 0.0136, 0.0459, 0.0098], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0166, 0.0175, 0.0133, 0.0179, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-30 00:58:06,573 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.001e+02 3.052e+02 3.743e+02 5.058e+02 1.059e+03, threshold=7.487e+02, percent-clipped=6.0 2023-04-30 00:58:17,058 INFO [train.py:904] (5/8) Epoch 14, batch 8150, loss[loss=0.2379, simple_loss=0.3053, pruned_loss=0.08528, over 12022.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.296, pruned_loss=0.06571, over 3066612.11 frames. ], batch size: 247, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:58:18,908 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:29,476 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140110.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:50,717 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3308, 4.3061, 4.2047, 3.5168, 4.2679, 1.7338, 4.0294, 3.9298], device='cuda:5'), covar=tensor([0.0101, 0.0084, 0.0151, 0.0320, 0.0079, 0.2428, 0.0119, 0.0198], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0127, 0.0174, 0.0162, 0.0146, 0.0189, 0.0162, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 00:58:52,675 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:23,741 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:34,251 INFO [train.py:904] (5/8) Epoch 14, batch 8200, loss[loss=0.2219, simple_loss=0.3054, pruned_loss=0.06918, over 16895.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2934, pruned_loss=0.06454, over 3085193.56 frames. ], batch size: 116, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:59:47,496 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140160.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:50,572 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:53,819 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:05,670 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:44,854 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.531e+02 3.267e+02 4.112e+02 9.606e+02, threshold=6.535e+02, percent-clipped=4.0 2023-04-30 01:00:55,630 INFO [train.py:904] (5/8) Epoch 14, batch 8250, loss[loss=0.2102, simple_loss=0.299, pruned_loss=0.06065, over 16205.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.292, pruned_loss=0.06152, over 3085948.86 frames. ], batch size: 165, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:01:02,327 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140206.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:31,181 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3265, 2.0320, 1.9820, 3.8566, 2.0388, 2.4266, 2.1326, 2.2342], device='cuda:5'), covar=tensor([0.0988, 0.3671, 0.2909, 0.0455, 0.4189, 0.2493, 0.3660, 0.3425], device='cuda:5'), in_proj_covar=tensor([0.0367, 0.0406, 0.0337, 0.0314, 0.0415, 0.0464, 0.0370, 0.0472], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:01:37,222 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:50,648 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9752, 1.7938, 1.6308, 1.5097, 1.8899, 1.6027, 1.6479, 1.9307], device='cuda:5'), covar=tensor([0.0148, 0.0233, 0.0328, 0.0301, 0.0195, 0.0234, 0.0143, 0.0180], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0207, 0.0203, 0.0204, 0.0209, 0.0208, 0.0210, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:01:55,295 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:02:16,388 INFO [train.py:904] (5/8) Epoch 14, batch 8300, loss[loss=0.1874, simple_loss=0.2811, pruned_loss=0.0469, over 16827.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2896, pruned_loss=0.0589, over 3071546.84 frames. ], batch size: 116, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:02:29,789 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140259.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:02:42,541 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4640, 3.2514, 3.5624, 1.7710, 3.7293, 3.7598, 2.9997, 2.8714], device='cuda:5'), covar=tensor([0.0693, 0.0251, 0.0202, 0.1197, 0.0061, 0.0161, 0.0365, 0.0402], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0104, 0.0091, 0.0138, 0.0072, 0.0112, 0.0124, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 01:02:55,100 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:03:12,852 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:03:27,887 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.340e+02 2.728e+02 3.506e+02 7.480e+02, threshold=5.455e+02, percent-clipped=2.0 2023-04-30 01:03:34,346 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5033, 3.3638, 2.7980, 2.0560, 2.1419, 2.2909, 3.3634, 3.1686], device='cuda:5'), covar=tensor([0.2608, 0.0623, 0.1495, 0.2692, 0.2562, 0.2010, 0.0402, 0.1022], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0255, 0.0284, 0.0285, 0.0278, 0.0228, 0.0269, 0.0302], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:03:36,683 INFO [train.py:904] (5/8) Epoch 14, batch 8350, loss[loss=0.2057, simple_loss=0.2829, pruned_loss=0.06426, over 12116.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2892, pruned_loss=0.05727, over 3062688.47 frames. ], batch size: 247, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:03:39,777 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 01:04:07,970 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:04:57,951 INFO [train.py:904] (5/8) Epoch 14, batch 8400, loss[loss=0.1622, simple_loss=0.2615, pruned_loss=0.03139, over 16723.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2865, pruned_loss=0.05512, over 3065532.53 frames. ], batch size: 89, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:05:54,821 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140387.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:05:57,918 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2404, 4.2666, 4.3902, 4.2484, 4.3435, 4.8179, 4.3993, 4.0804], device='cuda:5'), covar=tensor([0.1425, 0.2045, 0.1944, 0.2120, 0.2527, 0.1018, 0.1459, 0.2447], device='cuda:5'), in_proj_covar=tensor([0.0363, 0.0509, 0.0563, 0.0435, 0.0581, 0.0585, 0.0443, 0.0587], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 01:06:04,289 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5513, 4.8282, 4.6254, 4.6287, 4.3106, 4.3445, 4.2553, 4.8894], device='cuda:5'), covar=tensor([0.1021, 0.0806, 0.0945, 0.0717, 0.0804, 0.1040, 0.1068, 0.0849], device='cuda:5'), in_proj_covar=tensor([0.0572, 0.0700, 0.0578, 0.0505, 0.0443, 0.0457, 0.0587, 0.0535], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:06:04,509 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1291, 2.0029, 2.1159, 3.6578, 1.9854, 2.4295, 2.1513, 2.2189], device='cuda:5'), covar=tensor([0.1116, 0.3913, 0.2764, 0.0493, 0.4360, 0.2537, 0.3669, 0.3609], device='cuda:5'), in_proj_covar=tensor([0.0364, 0.0402, 0.0336, 0.0310, 0.0411, 0.0459, 0.0366, 0.0467], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:06:04,758 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-04-30 01:06:08,149 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.349e+02 2.839e+02 3.572e+02 6.227e+02, threshold=5.678e+02, percent-clipped=3.0 2023-04-30 01:06:18,712 INFO [train.py:904] (5/8) Epoch 14, batch 8450, loss[loss=0.2003, simple_loss=0.2913, pruned_loss=0.05461, over 16376.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2848, pruned_loss=0.05315, over 3076556.16 frames. ], batch size: 146, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:06:55,882 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:27,309 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:32,367 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140448.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:37,971 INFO [train.py:904] (5/8) Epoch 14, batch 8500, loss[loss=0.1817, simple_loss=0.2548, pruned_loss=0.05431, over 11987.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2805, pruned_loss=0.05061, over 3050948.93 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:07:50,062 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:51,452 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:54,756 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:01,420 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:13,346 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:38,673 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140489.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:46,173 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:51,331 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.295e+02 2.729e+02 3.409e+02 8.118e+02, threshold=5.459e+02, percent-clipped=4.0 2023-04-30 01:09:01,487 INFO [train.py:904] (5/8) Epoch 14, batch 8550, loss[loss=0.176, simple_loss=0.2733, pruned_loss=0.03929, over 16846.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2782, pruned_loss=0.04971, over 3026795.31 frames. ], batch size: 83, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:09:11,920 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:15,779 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:17,284 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:19,130 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:39,287 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:41,528 INFO [train.py:904] (5/8) Epoch 14, batch 8600, loss[loss=0.1803, simple_loss=0.2824, pruned_loss=0.03912, over 16887.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2785, pruned_loss=0.04847, over 3037918.93 frames. ], batch size: 102, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:10:46,552 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:11:19,564 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:12:06,771 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.341e+02 2.953e+02 3.525e+02 9.042e+02, threshold=5.906e+02, percent-clipped=3.0 2023-04-30 01:12:18,464 INFO [train.py:904] (5/8) Epoch 14, batch 8650, loss[loss=0.1925, simple_loss=0.2836, pruned_loss=0.05075, over 16675.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2764, pruned_loss=0.04681, over 3024264.95 frames. ], batch size: 134, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:12:38,353 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5225, 2.9288, 3.1919, 1.9466, 2.7600, 2.1082, 3.0914, 3.1165], device='cuda:5'), covar=tensor([0.0261, 0.0750, 0.0514, 0.1854, 0.0764, 0.0990, 0.0585, 0.0745], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0146, 0.0155, 0.0142, 0.0134, 0.0123, 0.0134, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 01:12:52,308 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:14:05,330 INFO [train.py:904] (5/8) Epoch 14, batch 8700, loss[loss=0.1734, simple_loss=0.2618, pruned_loss=0.04255, over 12431.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2734, pruned_loss=0.04556, over 3023822.78 frames. ], batch size: 247, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:14:55,707 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1636, 2.5163, 2.6981, 1.8859, 2.8152, 2.8937, 2.5326, 2.4945], device='cuda:5'), covar=tensor([0.0643, 0.0203, 0.0199, 0.1026, 0.0092, 0.0195, 0.0416, 0.0400], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0100, 0.0087, 0.0135, 0.0069, 0.0109, 0.0120, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 01:15:17,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9452, 5.2975, 5.4586, 5.2427, 5.2659, 5.8042, 5.3160, 5.0364], device='cuda:5'), covar=tensor([0.0760, 0.1544, 0.1689, 0.1746, 0.2314, 0.0830, 0.1392, 0.2313], device='cuda:5'), in_proj_covar=tensor([0.0355, 0.0498, 0.0547, 0.0422, 0.0566, 0.0576, 0.0435, 0.0573], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 01:15:22,255 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 01:15:24,089 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.380e+02 2.814e+02 3.680e+02 6.881e+02, threshold=5.627e+02, percent-clipped=2.0 2023-04-30 01:15:27,770 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6637, 2.7221, 2.4804, 4.1543, 2.6892, 4.1215, 1.3762, 3.0617], device='cuda:5'), covar=tensor([0.1475, 0.0725, 0.1174, 0.0146, 0.0134, 0.0341, 0.1720, 0.0665], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0161, 0.0183, 0.0157, 0.0196, 0.0208, 0.0186, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 01:15:38,747 INFO [train.py:904] (5/8) Epoch 14, batch 8750, loss[loss=0.2049, simple_loss=0.303, pruned_loss=0.05341, over 16713.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2729, pruned_loss=0.045, over 3028480.60 frames. ], batch size: 134, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:16:31,562 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:16:31,813 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 01:16:43,716 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5425, 4.8456, 4.6828, 4.6615, 4.3471, 4.3699, 4.2586, 4.9141], device='cuda:5'), covar=tensor([0.1028, 0.0767, 0.0745, 0.0663, 0.0733, 0.1019, 0.0943, 0.0742], device='cuda:5'), in_proj_covar=tensor([0.0564, 0.0695, 0.0570, 0.0502, 0.0440, 0.0453, 0.0583, 0.0528], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:16:45,853 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5429, 3.8820, 4.1435, 1.9962, 3.3531, 2.7912, 3.9430, 3.8969], device='cuda:5'), covar=tensor([0.0237, 0.0719, 0.0417, 0.2025, 0.0682, 0.0801, 0.0589, 0.0921], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0144, 0.0154, 0.0141, 0.0133, 0.0122, 0.0133, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 01:16:58,648 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.57 vs. limit=5.0 2023-04-30 01:17:11,080 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:27,846 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4578, 4.7649, 4.4339, 4.1781, 3.7476, 4.6784, 4.5876, 4.2741], device='cuda:5'), covar=tensor([0.0908, 0.0507, 0.0464, 0.0410, 0.1908, 0.0475, 0.0347, 0.0762], device='cuda:5'), in_proj_covar=tensor([0.0247, 0.0336, 0.0296, 0.0274, 0.0306, 0.0321, 0.0204, 0.0345], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:17:28,584 INFO [train.py:904] (5/8) Epoch 14, batch 8800, loss[loss=0.1852, simple_loss=0.2825, pruned_loss=0.04392, over 16635.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2716, pruned_loss=0.04388, over 3048626.06 frames. ], batch size: 89, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:17:35,483 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 01:17:43,099 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:56,993 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:18:31,193 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6028, 3.5895, 3.5323, 2.8821, 3.4819, 1.9308, 3.2410, 2.9476], device='cuda:5'), covar=tensor([0.0095, 0.0082, 0.0136, 0.0168, 0.0078, 0.2186, 0.0091, 0.0198], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0124, 0.0170, 0.0154, 0.0142, 0.0186, 0.0157, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:18:35,813 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:00,061 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.211e+02 2.894e+02 3.689e+02 7.406e+02, threshold=5.789e+02, percent-clipped=2.0 2023-04-30 01:19:12,357 INFO [train.py:904] (5/8) Epoch 14, batch 8850, loss[loss=0.1784, simple_loss=0.2838, pruned_loss=0.03652, over 16622.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2744, pruned_loss=0.04322, over 3054584.08 frames. ], batch size: 134, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:19:23,527 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:27,231 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:37,860 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:20:27,816 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 01:20:40,357 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9376, 2.2601, 1.9904, 2.0165, 2.5923, 2.2824, 2.6211, 2.8553], device='cuda:5'), covar=tensor([0.0117, 0.0341, 0.0426, 0.0389, 0.0200, 0.0334, 0.0177, 0.0189], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0208, 0.0202, 0.0203, 0.0207, 0.0207, 0.0207, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:20:44,024 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140845.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:20:58,225 INFO [train.py:904] (5/8) Epoch 14, batch 8900, loss[loss=0.1802, simple_loss=0.2793, pruned_loss=0.04055, over 16816.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2745, pruned_loss=0.04242, over 3047091.65 frames. ], batch size: 83, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:21:23,735 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9508, 4.1148, 2.3584, 4.6290, 3.0087, 4.4971, 2.2904, 3.0479], device='cuda:5'), covar=tensor([0.0190, 0.0214, 0.1486, 0.0125, 0.0725, 0.0319, 0.1580, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0160, 0.0183, 0.0132, 0.0162, 0.0198, 0.0190, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 01:21:25,590 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:21:34,659 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:22:44,746 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 01:22:46,891 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.342e+02 2.775e+02 3.236e+02 6.448e+02, threshold=5.549e+02, percent-clipped=1.0 2023-04-30 01:23:01,545 INFO [train.py:904] (5/8) Epoch 14, batch 8950, loss[loss=0.1864, simple_loss=0.2724, pruned_loss=0.05023, over 16975.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2742, pruned_loss=0.04291, over 3058616.43 frames. ], batch size: 109, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:23:29,184 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:23:32,811 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:24:38,390 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:24:48,141 INFO [train.py:904] (5/8) Epoch 14, batch 9000, loss[loss=0.1547, simple_loss=0.2467, pruned_loss=0.03137, over 16855.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2709, pruned_loss=0.04144, over 3071212.93 frames. ], batch size: 124, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:24:48,142 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 01:24:58,092 INFO [train.py:938] (5/8) Epoch 14, validation: loss=0.1514, simple_loss=0.2553, pruned_loss=0.0237, over 944034.00 frames. 2023-04-30 01:24:58,094 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 01:25:21,319 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:25:52,483 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:26:29,791 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 2.199e+02 2.701e+02 3.272e+02 7.252e+02, threshold=5.402e+02, percent-clipped=4.0 2023-04-30 01:26:39,932 INFO [train.py:904] (5/8) Epoch 14, batch 9050, loss[loss=0.195, simple_loss=0.2799, pruned_loss=0.05507, over 15343.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2718, pruned_loss=0.04207, over 3077753.82 frames. ], batch size: 190, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:26:55,093 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141008.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:08,232 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:25,850 INFO [train.py:904] (5/8) Epoch 14, batch 9100, loss[loss=0.1811, simple_loss=0.2679, pruned_loss=0.04721, over 12267.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2716, pruned_loss=0.04258, over 3077192.33 frames. ], batch size: 246, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:29:30,809 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:29:58,972 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:30:13,182 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.366e+02 2.876e+02 3.628e+02 8.857e+02, threshold=5.752e+02, percent-clipped=2.0 2023-04-30 01:30:22,851 INFO [train.py:904] (5/8) Epoch 14, batch 9150, loss[loss=0.1763, simple_loss=0.2694, pruned_loss=0.04161, over 15318.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2723, pruned_loss=0.04249, over 3076237.08 frames. ], batch size: 191, lr: 4.80e-03, grad_scale: 4.0 2023-04-30 01:30:54,347 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 01:31:41,022 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4924, 3.4994, 3.4731, 2.6106, 3.3838, 1.9765, 3.1771, 2.7968], device='cuda:5'), covar=tensor([0.0221, 0.0181, 0.0234, 0.0388, 0.0162, 0.2634, 0.0181, 0.0310], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0125, 0.0169, 0.0154, 0.0143, 0.0186, 0.0157, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:31:51,313 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:31:54,441 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:04,846 INFO [train.py:904] (5/8) Epoch 14, batch 9200, loss[loss=0.1715, simple_loss=0.267, pruned_loss=0.03798, over 12047.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2676, pruned_loss=0.04156, over 3066780.71 frames. ], batch size: 247, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:32:26,103 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6424, 4.0802, 4.2203, 2.3973, 3.3077, 2.8184, 3.9162, 4.1963], device='cuda:5'), covar=tensor([0.0234, 0.0642, 0.0407, 0.1710, 0.0716, 0.0849, 0.0623, 0.0753], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0144, 0.0156, 0.0143, 0.0136, 0.0124, 0.0135, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 01:32:29,664 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:29,702 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:33:23,826 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:33:31,531 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.509e+02 2.979e+02 4.006e+02 1.027e+03, threshold=5.958e+02, percent-clipped=7.0 2023-04-30 01:33:40,999 INFO [train.py:904] (5/8) Epoch 14, batch 9250, loss[loss=0.1693, simple_loss=0.2604, pruned_loss=0.03912, over 16852.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.267, pruned_loss=0.04168, over 3055040.71 frames. ], batch size: 124, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:33:46,145 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141204.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:34:03,407 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:35:30,128 INFO [train.py:904] (5/8) Epoch 14, batch 9300, loss[loss=0.1687, simple_loss=0.2686, pruned_loss=0.03439, over 15236.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2657, pruned_loss=0.0408, over 3061034.69 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:35:38,110 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 01:36:20,432 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:37:05,693 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.032e+02 2.564e+02 3.502e+02 9.246e+02, threshold=5.129e+02, percent-clipped=2.0 2023-04-30 01:37:13,199 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5568, 4.6223, 4.4280, 4.1212, 4.0535, 4.4954, 4.2724, 4.1853], device='cuda:5'), covar=tensor([0.0542, 0.0542, 0.0316, 0.0292, 0.0939, 0.0632, 0.0465, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0245, 0.0333, 0.0293, 0.0272, 0.0304, 0.0320, 0.0202, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:37:14,533 INFO [train.py:904] (5/8) Epoch 14, batch 9350, loss[loss=0.179, simple_loss=0.2739, pruned_loss=0.04202, over 16922.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2661, pruned_loss=0.04122, over 3054600.91 frames. ], batch size: 116, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:37:17,097 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:08,682 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:26,788 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3130, 4.4199, 4.1962, 3.8918, 3.8943, 4.2962, 4.0611, 4.0479], device='cuda:5'), covar=tensor([0.0532, 0.0469, 0.0285, 0.0258, 0.0814, 0.0453, 0.0617, 0.0530], device='cuda:5'), in_proj_covar=tensor([0.0244, 0.0332, 0.0293, 0.0272, 0.0304, 0.0319, 0.0202, 0.0342], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:38:54,841 INFO [train.py:904] (5/8) Epoch 14, batch 9400, loss[loss=0.1751, simple_loss=0.2714, pruned_loss=0.03936, over 15367.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2655, pruned_loss=0.04092, over 3035216.35 frames. ], batch size: 190, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:39:17,898 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2674, 5.5615, 5.3792, 5.3530, 4.9667, 4.9856, 5.0415, 5.6590], device='cuda:5'), covar=tensor([0.1164, 0.0970, 0.0838, 0.0728, 0.0813, 0.0699, 0.1028, 0.0861], device='cuda:5'), in_proj_covar=tensor([0.0556, 0.0685, 0.0560, 0.0493, 0.0436, 0.0448, 0.0576, 0.0523], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:39:49,383 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:12,099 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:25,323 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.208e+02 2.590e+02 3.171e+02 6.862e+02, threshold=5.180e+02, percent-clipped=4.0 2023-04-30 01:40:33,417 INFO [train.py:904] (5/8) Epoch 14, batch 9450, loss[loss=0.1634, simple_loss=0.2591, pruned_loss=0.03389, over 17049.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2672, pruned_loss=0.04116, over 3032182.87 frames. ], batch size: 50, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:40:38,958 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0512, 3.1929, 3.0906, 2.1618, 2.8769, 3.2398, 3.1181, 1.9981], device='cuda:5'), covar=tensor([0.0422, 0.0040, 0.0050, 0.0340, 0.0106, 0.0066, 0.0068, 0.0391], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0070, 0.0072, 0.0128, 0.0084, 0.0093, 0.0082, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 01:41:03,215 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141417.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 01:41:24,366 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141427.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:41:29,323 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 01:41:40,276 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 01:41:54,109 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141441.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:42:14,339 INFO [train.py:904] (5/8) Epoch 14, batch 9500, loss[loss=0.1564, simple_loss=0.2422, pruned_loss=0.03535, over 12788.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2661, pruned_loss=0.04053, over 3037395.48 frames. ], batch size: 246, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:42:28,798 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 01:42:43,595 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141465.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:42:59,309 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 01:43:08,429 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141478.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 01:43:21,757 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2168, 3.2755, 1.9040, 3.5070, 2.3500, 3.4649, 2.0543, 2.6338], device='cuda:5'), covar=tensor([0.0272, 0.0334, 0.1551, 0.0234, 0.0851, 0.0596, 0.1501, 0.0738], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0161, 0.0184, 0.0132, 0.0163, 0.0198, 0.0191, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 01:43:47,718 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.502e+02 2.299e+02 2.699e+02 3.093e+02 4.944e+02, threshold=5.398e+02, percent-clipped=0.0 2023-04-30 01:43:55,191 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141499.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:43:59,705 INFO [train.py:904] (5/8) Epoch 14, batch 9550, loss[loss=0.1975, simple_loss=0.2931, pruned_loss=0.05095, over 16291.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2657, pruned_loss=0.04089, over 3047432.07 frames. ], batch size: 146, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:44:01,087 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141502.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:44:23,849 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:45:41,076 INFO [train.py:904] (5/8) Epoch 14, batch 9600, loss[loss=0.1777, simple_loss=0.277, pruned_loss=0.03915, over 16471.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2678, pruned_loss=0.04183, over 3052481.66 frames. ], batch size: 68, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:46:06,741 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2476, 4.2275, 4.6309, 4.6122, 4.6316, 4.3604, 4.3378, 4.2208], device='cuda:5'), covar=tensor([0.0300, 0.0515, 0.0429, 0.0425, 0.0504, 0.0357, 0.0876, 0.0465], device='cuda:5'), in_proj_covar=tensor([0.0333, 0.0350, 0.0351, 0.0335, 0.0395, 0.0373, 0.0450, 0.0297], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:5') 2023-04-30 01:46:22,434 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:47:19,268 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.296e+02 2.650e+02 3.286e+02 9.159e+02, threshold=5.299e+02, percent-clipped=3.0 2023-04-30 01:47:30,287 INFO [train.py:904] (5/8) Epoch 14, batch 9650, loss[loss=0.1782, simple_loss=0.272, pruned_loss=0.04226, over 16695.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2703, pruned_loss=0.04234, over 3058628.64 frames. ], batch size: 89, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:47:34,929 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141603.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:47:35,414 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-30 01:48:16,938 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:48:39,767 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5343, 2.1602, 2.1920, 4.2264, 2.1938, 2.5830, 2.3483, 2.4263], device='cuda:5'), covar=tensor([0.0931, 0.3498, 0.2778, 0.0392, 0.3952, 0.2411, 0.3339, 0.3194], device='cuda:5'), in_proj_covar=tensor([0.0359, 0.0397, 0.0334, 0.0309, 0.0408, 0.0451, 0.0362, 0.0460], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 01:49:17,049 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141651.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:49:17,852 INFO [train.py:904] (5/8) Epoch 14, batch 9700, loss[loss=0.1634, simple_loss=0.2506, pruned_loss=0.03808, over 12471.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2688, pruned_loss=0.04215, over 3042117.91 frames. ], batch size: 248, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:49:45,684 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 01:50:27,993 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141685.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:50:53,327 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.310e+02 2.782e+02 3.506e+02 1.114e+03, threshold=5.564e+02, percent-clipped=5.0 2023-04-30 01:51:00,301 INFO [train.py:904] (5/8) Epoch 14, batch 9750, loss[loss=0.1683, simple_loss=0.2508, pruned_loss=0.04292, over 12542.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2673, pruned_loss=0.04227, over 3027503.16 frames. ], batch size: 246, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:51:59,263 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3253, 3.2341, 3.4853, 1.7149, 3.6536, 3.7262, 2.8505, 2.8068], device='cuda:5'), covar=tensor([0.0769, 0.0206, 0.0178, 0.1231, 0.0058, 0.0136, 0.0456, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0098, 0.0084, 0.0133, 0.0068, 0.0106, 0.0119, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 01:52:38,206 INFO [train.py:904] (5/8) Epoch 14, batch 9800, loss[loss=0.1908, simple_loss=0.2972, pruned_loss=0.0422, over 16880.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2678, pruned_loss=0.04122, over 3057052.52 frames. ], batch size: 116, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:53:07,077 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-30 01:53:18,167 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141773.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 01:54:13,234 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141797.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:54:13,944 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.347e+02 2.793e+02 3.311e+02 9.229e+02, threshold=5.586e+02, percent-clipped=3.0 2023-04-30 01:54:17,316 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141799.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:54:21,455 INFO [train.py:904] (5/8) Epoch 14, batch 9850, loss[loss=0.1668, simple_loss=0.2553, pruned_loss=0.03911, over 12608.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2696, pruned_loss=0.04131, over 3058796.38 frames. ], batch size: 248, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:56:05,082 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141847.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:56:14,800 INFO [train.py:904] (5/8) Epoch 14, batch 9900, loss[loss=0.1658, simple_loss=0.2553, pruned_loss=0.03815, over 12700.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2704, pruned_loss=0.04149, over 3040929.62 frames. ], batch size: 248, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:56:20,131 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6208, 2.5754, 2.5833, 4.4634, 2.9356, 4.2076, 1.5045, 3.0119], device='cuda:5'), covar=tensor([0.1445, 0.0824, 0.1132, 0.0145, 0.0146, 0.0293, 0.1661, 0.0706], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0162, 0.0184, 0.0157, 0.0190, 0.0206, 0.0188, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 01:58:03,749 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.142e+02 2.825e+02 3.311e+02 8.255e+02, threshold=5.650e+02, percent-clipped=2.0 2023-04-30 01:58:05,855 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5870, 3.6535, 3.4549, 3.1849, 3.2456, 3.5211, 3.3350, 3.3791], device='cuda:5'), covar=tensor([0.0518, 0.0548, 0.0261, 0.0240, 0.0492, 0.0444, 0.1046, 0.0454], device='cuda:5'), in_proj_covar=tensor([0.0239, 0.0324, 0.0286, 0.0266, 0.0295, 0.0310, 0.0196, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-30 01:58:13,951 INFO [train.py:904] (5/8) Epoch 14, batch 9950, loss[loss=0.1702, simple_loss=0.2634, pruned_loss=0.03852, over 17176.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2725, pruned_loss=0.04154, over 3062257.05 frames. ], batch size: 44, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:58:48,442 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141916.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:59:21,083 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:00:16,392 INFO [train.py:904] (5/8) Epoch 14, batch 10000, loss[loss=0.1618, simple_loss=0.2556, pruned_loss=0.03398, over 16563.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2709, pruned_loss=0.04075, over 3073854.38 frames. ], batch size: 62, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:00:17,591 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9913, 3.8166, 4.0538, 4.1564, 4.2729, 3.8182, 4.2414, 4.2701], device='cuda:5'), covar=tensor([0.1483, 0.1130, 0.1207, 0.0760, 0.0488, 0.1395, 0.0696, 0.0557], device='cuda:5'), in_proj_covar=tensor([0.0533, 0.0657, 0.0776, 0.0676, 0.0506, 0.0524, 0.0533, 0.0629], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:00:43,596 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5268, 4.5302, 4.3547, 3.9360, 4.4482, 1.6494, 4.1762, 4.2451], device='cuda:5'), covar=tensor([0.0084, 0.0072, 0.0152, 0.0207, 0.0070, 0.2527, 0.0110, 0.0172], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0124, 0.0167, 0.0150, 0.0141, 0.0185, 0.0156, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:01:07,150 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:25,930 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:34,599 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141989.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:50,851 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.061e+02 2.551e+02 3.100e+02 8.241e+02, threshold=5.102e+02, percent-clipped=1.0 2023-04-30 02:02:01,159 INFO [train.py:904] (5/8) Epoch 14, batch 10050, loss[loss=0.191, simple_loss=0.2873, pruned_loss=0.04738, over 16222.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2703, pruned_loss=0.04039, over 3060600.89 frames. ], batch size: 165, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:02:32,141 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 02:03:01,096 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142033.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:03:34,650 INFO [train.py:904] (5/8) Epoch 14, batch 10100, loss[loss=0.1615, simple_loss=0.2534, pruned_loss=0.03476, over 16239.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2707, pruned_loss=0.0406, over 3055814.76 frames. ], batch size: 165, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:04:19,981 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142073.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:04:44,738 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9576, 2.0480, 2.3742, 3.2309, 2.1700, 2.2783, 2.2364, 2.1566], device='cuda:5'), covar=tensor([0.1098, 0.3659, 0.2327, 0.0594, 0.4041, 0.2420, 0.3033, 0.3544], device='cuda:5'), in_proj_covar=tensor([0.0359, 0.0396, 0.0333, 0.0307, 0.0407, 0.0450, 0.0362, 0.0458], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:04:49,690 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142097.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:04:50,458 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.492e+02 2.995e+02 3.648e+02 7.218e+02, threshold=5.990e+02, percent-clipped=8.0 2023-04-30 02:05:19,926 INFO [train.py:904] (5/8) Epoch 15, batch 0, loss[loss=0.2257, simple_loss=0.3157, pruned_loss=0.06779, over 17004.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3157, pruned_loss=0.06779, over 17004.00 frames. ], batch size: 55, lr: 4.62e-03, grad_scale: 8.0 2023-04-30 02:05:19,926 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 02:05:27,348 INFO [train.py:938] (5/8) Epoch 15, validation: loss=0.1501, simple_loss=0.2536, pruned_loss=0.02333, over 944034.00 frames. 2023-04-30 02:05:27,349 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 02:05:53,987 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142121.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:06:15,265 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 02:06:27,846 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:06:37,645 INFO [train.py:904] (5/8) Epoch 15, batch 50, loss[loss=0.1771, simple_loss=0.2746, pruned_loss=0.03976, over 16685.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2801, pruned_loss=0.06036, over 752681.47 frames. ], batch size: 57, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:07:02,102 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3554, 3.5381, 3.2915, 5.2757, 4.4362, 4.7417, 2.2148, 3.7009], device='cuda:5'), covar=tensor([0.1112, 0.0575, 0.0917, 0.0146, 0.0282, 0.0341, 0.1294, 0.0621], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0162, 0.0184, 0.0159, 0.0190, 0.0207, 0.0189, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:07:23,968 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3520, 2.1991, 2.3663, 4.1030, 2.2245, 2.5406, 2.2735, 2.4057], device='cuda:5'), covar=tensor([0.1122, 0.3261, 0.2508, 0.0552, 0.3786, 0.2245, 0.3350, 0.2810], device='cuda:5'), in_proj_covar=tensor([0.0364, 0.0402, 0.0337, 0.0312, 0.0411, 0.0456, 0.0366, 0.0464], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:07:44,551 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.683e+02 3.202e+02 3.967e+02 8.381e+02, threshold=6.403e+02, percent-clipped=5.0 2023-04-30 02:07:47,916 INFO [train.py:904] (5/8) Epoch 15, batch 100, loss[loss=0.193, simple_loss=0.275, pruned_loss=0.05555, over 16585.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2762, pruned_loss=0.05637, over 1319148.27 frames. ], batch size: 68, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:07:51,076 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 02:08:33,335 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 02:08:56,697 INFO [train.py:904] (5/8) Epoch 15, batch 150, loss[loss=0.1569, simple_loss=0.2553, pruned_loss=0.02924, over 17107.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2738, pruned_loss=0.05408, over 1765232.75 frames. ], batch size: 47, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:09:25,356 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:09:42,935 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:10:04,594 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.354e+02 2.662e+02 3.206e+02 6.929e+02, threshold=5.325e+02, percent-clipped=1.0 2023-04-30 02:10:07,781 INFO [train.py:904] (5/8) Epoch 15, batch 200, loss[loss=0.1877, simple_loss=0.2748, pruned_loss=0.05027, over 16022.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2724, pruned_loss=0.05319, over 2113813.23 frames. ], batch size: 35, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:11:16,795 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 02:11:17,116 INFO [train.py:904] (5/8) Epoch 15, batch 250, loss[loss=0.1635, simple_loss=0.2579, pruned_loss=0.03452, over 17091.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2701, pruned_loss=0.05257, over 2378823.44 frames. ], batch size: 50, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:12:22,523 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 2.346e+02 2.797e+02 3.590e+02 6.679e+02, threshold=5.594e+02, percent-clipped=6.0 2023-04-30 02:12:25,449 INFO [train.py:904] (5/8) Epoch 15, batch 300, loss[loss=0.1768, simple_loss=0.252, pruned_loss=0.05081, over 16476.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2688, pruned_loss=0.05168, over 2590062.27 frames. ], batch size: 75, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:12:54,923 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 02:13:36,232 INFO [train.py:904] (5/8) Epoch 15, batch 350, loss[loss=0.19, simple_loss=0.2619, pruned_loss=0.05901, over 16286.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2656, pruned_loss=0.05001, over 2758506.66 frames. ], batch size: 165, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:14:42,902 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.253e+02 2.722e+02 3.272e+02 5.602e+02, threshold=5.443e+02, percent-clipped=1.0 2023-04-30 02:14:45,198 INFO [train.py:904] (5/8) Epoch 15, batch 400, loss[loss=0.1674, simple_loss=0.2615, pruned_loss=0.0366, over 17022.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2641, pruned_loss=0.04975, over 2880656.75 frames. ], batch size: 50, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:15:54,203 INFO [train.py:904] (5/8) Epoch 15, batch 450, loss[loss=0.1474, simple_loss=0.2406, pruned_loss=0.02706, over 17212.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2616, pruned_loss=0.0485, over 2979068.83 frames. ], batch size: 46, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:16:23,534 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:23,709 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6972, 2.2615, 2.2689, 4.5941, 2.2723, 2.7083, 2.3525, 2.4547], device='cuda:5'), covar=tensor([0.0990, 0.3431, 0.2743, 0.0387, 0.4185, 0.2411, 0.3144, 0.3550], device='cuda:5'), in_proj_covar=tensor([0.0376, 0.0412, 0.0345, 0.0321, 0.0421, 0.0472, 0.0377, 0.0480], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:16:24,811 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:41,599 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:48,182 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9199, 2.6985, 2.6108, 1.9125, 2.5034, 2.7010, 2.5850, 1.7984], device='cuda:5'), covar=tensor([0.0395, 0.0083, 0.0062, 0.0345, 0.0114, 0.0108, 0.0097, 0.0399], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0074, 0.0074, 0.0130, 0.0086, 0.0097, 0.0085, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 02:17:03,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.083e+02 2.627e+02 3.131e+02 6.454e+02, threshold=5.253e+02, percent-clipped=1.0 2023-04-30 02:17:05,232 INFO [train.py:904] (5/8) Epoch 15, batch 500, loss[loss=0.1981, simple_loss=0.2685, pruned_loss=0.06383, over 16866.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2614, pruned_loss=0.04821, over 3060239.93 frames. ], batch size: 116, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:17:28,779 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142620.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:46,444 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:49,328 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:18:13,787 INFO [train.py:904] (5/8) Epoch 15, batch 550, loss[loss=0.1826, simple_loss=0.2726, pruned_loss=0.04634, over 17118.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2605, pruned_loss=0.04724, over 3113065.32 frames. ], batch size: 48, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:18:15,143 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8955, 3.8854, 4.3968, 2.1053, 4.6061, 4.7074, 3.2575, 3.5252], device='cuda:5'), covar=tensor([0.0748, 0.0212, 0.0221, 0.1173, 0.0065, 0.0124, 0.0405, 0.0387], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0102, 0.0087, 0.0138, 0.0071, 0.0112, 0.0124, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 02:19:22,100 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.274e+02 2.667e+02 3.434e+02 1.165e+03, threshold=5.335e+02, percent-clipped=9.0 2023-04-30 02:19:23,203 INFO [train.py:904] (5/8) Epoch 15, batch 600, loss[loss=0.1843, simple_loss=0.2573, pruned_loss=0.05564, over 12346.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.26, pruned_loss=0.04736, over 3156763.13 frames. ], batch size: 246, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:19:53,066 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:20:32,879 INFO [train.py:904] (5/8) Epoch 15, batch 650, loss[loss=0.1656, simple_loss=0.2374, pruned_loss=0.04689, over 16725.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.259, pruned_loss=0.04717, over 3199550.15 frames. ], batch size: 83, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:20:44,604 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 02:21:17,894 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:21:23,133 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2772, 3.5308, 3.9173, 2.0554, 3.1896, 2.4894, 3.7325, 3.6494], device='cuda:5'), covar=tensor([0.0287, 0.0907, 0.0457, 0.1987, 0.0772, 0.0932, 0.0640, 0.1154], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0150, 0.0161, 0.0147, 0.0139, 0.0127, 0.0139, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:21:27,225 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9621, 4.8204, 4.7775, 4.4572, 4.4619, 4.8408, 4.6835, 4.5469], device='cuda:5'), covar=tensor([0.0631, 0.0801, 0.0317, 0.0320, 0.0967, 0.0495, 0.0458, 0.0792], device='cuda:5'), in_proj_covar=tensor([0.0263, 0.0360, 0.0313, 0.0293, 0.0327, 0.0342, 0.0214, 0.0370], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:21:40,879 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.547e+02 2.460e+02 3.000e+02 3.809e+02 2.810e+03, threshold=6.001e+02, percent-clipped=15.0 2023-04-30 02:21:42,128 INFO [train.py:904] (5/8) Epoch 15, batch 700, loss[loss=0.1593, simple_loss=0.2401, pruned_loss=0.03924, over 16496.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2583, pruned_loss=0.04707, over 3220104.00 frames. ], batch size: 68, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:22:27,169 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 02:22:43,315 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 02:22:50,295 INFO [train.py:904] (5/8) Epoch 15, batch 750, loss[loss=0.1728, simple_loss=0.2526, pruned_loss=0.04649, over 15704.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.259, pruned_loss=0.04688, over 3238067.53 frames. ], batch size: 191, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:22:50,767 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9481, 4.7325, 4.9842, 5.1965, 5.3878, 4.7383, 5.3414, 5.3778], device='cuda:5'), covar=tensor([0.1841, 0.1413, 0.1682, 0.0756, 0.0538, 0.0911, 0.0558, 0.0572], device='cuda:5'), in_proj_covar=tensor([0.0589, 0.0726, 0.0863, 0.0742, 0.0557, 0.0577, 0.0589, 0.0689], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:23:28,721 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 02:23:57,717 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 2.171e+02 2.529e+02 2.994e+02 6.163e+02, threshold=5.058e+02, percent-clipped=1.0 2023-04-30 02:23:59,534 INFO [train.py:904] (5/8) Epoch 15, batch 800, loss[loss=0.1741, simple_loss=0.2635, pruned_loss=0.04233, over 17150.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2582, pruned_loss=0.0464, over 3254689.30 frames. ], batch size: 48, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:24:36,443 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:25:08,654 INFO [train.py:904] (5/8) Epoch 15, batch 850, loss[loss=0.1773, simple_loss=0.2473, pruned_loss=0.05361, over 12397.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2576, pruned_loss=0.04632, over 3254993.01 frames. ], batch size: 246, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:25:22,685 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-30 02:25:44,574 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 02:26:15,125 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 2.170e+02 2.499e+02 2.974e+02 6.190e+02, threshold=4.998e+02, percent-clipped=2.0 2023-04-30 02:26:16,300 INFO [train.py:904] (5/8) Epoch 15, batch 900, loss[loss=0.1669, simple_loss=0.2618, pruned_loss=0.03599, over 17086.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2574, pruned_loss=0.04585, over 3272439.71 frames. ], batch size: 49, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:27:07,787 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4807, 3.5921, 3.9354, 2.0767, 3.1721, 2.4021, 3.8636, 3.7893], device='cuda:5'), covar=tensor([0.0299, 0.0847, 0.0484, 0.1887, 0.0756, 0.1036, 0.0597, 0.1059], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0150, 0.0160, 0.0147, 0.0138, 0.0126, 0.0138, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:27:12,115 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5958, 3.6484, 1.9874, 3.7817, 2.7480, 3.7584, 1.9024, 2.7904], device='cuda:5'), covar=tensor([0.0225, 0.0324, 0.1677, 0.0318, 0.0681, 0.0670, 0.1731, 0.0669], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0171, 0.0192, 0.0146, 0.0170, 0.0212, 0.0200, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:27:24,797 INFO [train.py:904] (5/8) Epoch 15, batch 950, loss[loss=0.1508, simple_loss=0.2408, pruned_loss=0.03044, over 16849.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2574, pruned_loss=0.04576, over 3286354.34 frames. ], batch size: 42, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:28:02,501 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143080.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:28:30,231 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.127e+02 2.534e+02 2.926e+02 6.813e+02, threshold=5.068e+02, percent-clipped=2.0 2023-04-30 02:28:31,471 INFO [train.py:904] (5/8) Epoch 15, batch 1000, loss[loss=0.1785, simple_loss=0.2493, pruned_loss=0.05381, over 16848.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2561, pruned_loss=0.04497, over 3297559.96 frames. ], batch size: 102, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:28:47,401 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7143, 4.8526, 5.1657, 5.1806, 5.2312, 4.9535, 4.7211, 4.5995], device='cuda:5'), covar=tensor([0.0467, 0.0735, 0.0654, 0.0692, 0.0683, 0.0564, 0.1225, 0.0545], device='cuda:5'), in_proj_covar=tensor([0.0375, 0.0396, 0.0394, 0.0373, 0.0441, 0.0420, 0.0507, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 02:29:41,404 INFO [train.py:904] (5/8) Epoch 15, batch 1050, loss[loss=0.1611, simple_loss=0.2549, pruned_loss=0.03364, over 17265.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2547, pruned_loss=0.0443, over 3294360.71 frames. ], batch size: 52, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:30:20,379 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 02:30:30,301 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 02:30:46,965 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.203e+02 2.568e+02 3.063e+02 1.237e+03, threshold=5.135e+02, percent-clipped=3.0 2023-04-30 02:30:49,003 INFO [train.py:904] (5/8) Epoch 15, batch 1100, loss[loss=0.1639, simple_loss=0.2421, pruned_loss=0.04287, over 16885.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.254, pruned_loss=0.04475, over 3296765.46 frames. ], batch size: 90, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:31:26,636 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143229.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:31:40,534 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7939, 4.7728, 4.6359, 4.0854, 4.6846, 2.0733, 4.3502, 4.4203], device='cuda:5'), covar=tensor([0.0122, 0.0090, 0.0211, 0.0399, 0.0115, 0.2361, 0.0180, 0.0201], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0135, 0.0182, 0.0165, 0.0154, 0.0196, 0.0170, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:31:58,387 INFO [train.py:904] (5/8) Epoch 15, batch 1150, loss[loss=0.1659, simple_loss=0.2427, pruned_loss=0.04456, over 15879.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2531, pruned_loss=0.0443, over 3298431.64 frames. ], batch size: 35, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:31:58,949 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8568, 4.0073, 4.3587, 2.1670, 4.5360, 4.5686, 3.2892, 3.6331], device='cuda:5'), covar=tensor([0.0688, 0.0200, 0.0182, 0.1161, 0.0063, 0.0145, 0.0380, 0.0364], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0104, 0.0089, 0.0140, 0.0073, 0.0116, 0.0126, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 02:32:34,543 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:33:07,205 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.162e+02 2.475e+02 3.140e+02 4.983e+02, threshold=4.950e+02, percent-clipped=0.0 2023-04-30 02:33:08,294 INFO [train.py:904] (5/8) Epoch 15, batch 1200, loss[loss=0.1566, simple_loss=0.2386, pruned_loss=0.03727, over 16825.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2522, pruned_loss=0.04381, over 3306751.76 frames. ], batch size: 96, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:34:16,293 INFO [train.py:904] (5/8) Epoch 15, batch 1250, loss[loss=0.2019, simple_loss=0.2649, pruned_loss=0.06949, over 16451.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2528, pruned_loss=0.04437, over 3318402.98 frames. ], batch size: 75, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:34:27,645 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5631, 2.5769, 2.2481, 2.4942, 2.9171, 2.7184, 3.2709, 3.1685], device='cuda:5'), covar=tensor([0.0118, 0.0369, 0.0439, 0.0405, 0.0255, 0.0340, 0.0241, 0.0215], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0224, 0.0215, 0.0216, 0.0224, 0.0223, 0.0229, 0.0216], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:34:29,355 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7223, 3.0027, 2.6057, 5.0146, 4.0960, 4.5683, 1.6064, 3.2536], device='cuda:5'), covar=tensor([0.1405, 0.0724, 0.1245, 0.0188, 0.0284, 0.0345, 0.1559, 0.0721], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0166, 0.0187, 0.0167, 0.0198, 0.0214, 0.0190, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:34:43,212 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-30 02:34:56,726 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143380.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:35:25,422 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143400.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:35:26,168 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.199e+02 2.560e+02 2.965e+02 4.851e+02, threshold=5.120e+02, percent-clipped=0.0 2023-04-30 02:35:27,822 INFO [train.py:904] (5/8) Epoch 15, batch 1300, loss[loss=0.1802, simple_loss=0.2643, pruned_loss=0.04801, over 16554.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2529, pruned_loss=0.04442, over 3319977.05 frames. ], batch size: 68, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:35:54,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4785, 4.4718, 4.4541, 3.9717, 4.4481, 1.8290, 4.1844, 4.1509], device='cuda:5'), covar=tensor([0.0110, 0.0090, 0.0144, 0.0266, 0.0089, 0.2444, 0.0133, 0.0192], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0135, 0.0182, 0.0165, 0.0154, 0.0196, 0.0169, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:35:57,468 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0669, 5.5221, 5.6955, 5.4480, 5.5009, 6.1019, 5.5717, 5.3331], device='cuda:5'), covar=tensor([0.0847, 0.1941, 0.2338, 0.1976, 0.2669, 0.0988, 0.1493, 0.2330], device='cuda:5'), in_proj_covar=tensor([0.0386, 0.0550, 0.0604, 0.0466, 0.0626, 0.0632, 0.0479, 0.0618], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 02:36:03,859 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143428.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:36:34,852 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8831, 4.3132, 3.1398, 2.2539, 2.9658, 2.6490, 4.6939, 3.8828], device='cuda:5'), covar=tensor([0.2664, 0.0640, 0.1718, 0.2857, 0.2534, 0.1837, 0.0388, 0.1180], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0261, 0.0292, 0.0288, 0.0280, 0.0236, 0.0275, 0.0311], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 02:36:37,208 INFO [train.py:904] (5/8) Epoch 15, batch 1350, loss[loss=0.1723, simple_loss=0.2521, pruned_loss=0.04619, over 15502.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2532, pruned_loss=0.04397, over 3318955.49 frames. ], batch size: 190, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:38,291 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 02:36:50,767 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143461.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 02:37:06,611 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143473.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:37:45,753 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.266e+02 2.683e+02 3.217e+02 5.542e+02, threshold=5.366e+02, percent-clipped=1.0 2023-04-30 02:37:47,556 INFO [train.py:904] (5/8) Epoch 15, batch 1400, loss[loss=0.1822, simple_loss=0.2768, pruned_loss=0.04384, over 17037.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2532, pruned_loss=0.04407, over 3324263.59 frames. ], batch size: 55, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:38:17,894 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8362, 1.9695, 2.3747, 2.8621, 2.5906, 3.3715, 1.9738, 3.2310], device='cuda:5'), covar=tensor([0.0227, 0.0444, 0.0315, 0.0278, 0.0297, 0.0161, 0.0464, 0.0148], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0183, 0.0168, 0.0172, 0.0181, 0.0136, 0.0183, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:38:32,172 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143534.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 02:38:36,714 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0686, 3.9317, 4.0243, 4.2566, 4.3234, 3.8908, 4.0338, 4.3295], device='cuda:5'), covar=tensor([0.1432, 0.1205, 0.1585, 0.0784, 0.0692, 0.1546, 0.2403, 0.0792], device='cuda:5'), in_proj_covar=tensor([0.0605, 0.0746, 0.0893, 0.0762, 0.0570, 0.0591, 0.0600, 0.0709], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:38:39,903 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 02:38:56,003 INFO [train.py:904] (5/8) Epoch 15, batch 1450, loss[loss=0.17, simple_loss=0.2445, pruned_loss=0.04776, over 16766.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2524, pruned_loss=0.04362, over 3328275.75 frames. ], batch size: 124, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:38:57,276 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2244, 4.1151, 4.5666, 2.1555, 4.7501, 4.8116, 3.2449, 3.8125], device='cuda:5'), covar=tensor([0.0586, 0.0186, 0.0170, 0.1104, 0.0071, 0.0109, 0.0421, 0.0312], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0105, 0.0091, 0.0141, 0.0073, 0.0117, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 02:39:46,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-30 02:40:05,555 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.183e+02 2.517e+02 3.188e+02 7.484e+02, threshold=5.034e+02, percent-clipped=1.0 2023-04-30 02:40:06,130 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0255, 3.1852, 3.2223, 2.1412, 2.8274, 2.2859, 3.5197, 3.4712], device='cuda:5'), covar=tensor([0.0246, 0.0833, 0.0588, 0.1719, 0.0800, 0.0970, 0.0531, 0.0878], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0153, 0.0163, 0.0148, 0.0140, 0.0127, 0.0140, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:40:06,729 INFO [train.py:904] (5/8) Epoch 15, batch 1500, loss[loss=0.1791, simple_loss=0.2544, pruned_loss=0.05189, over 16883.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2535, pruned_loss=0.0441, over 3333756.30 frames. ], batch size: 116, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:41:14,539 INFO [train.py:904] (5/8) Epoch 15, batch 1550, loss[loss=0.195, simple_loss=0.2637, pruned_loss=0.06314, over 16291.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.256, pruned_loss=0.04638, over 3331194.23 frames. ], batch size: 145, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:42:22,879 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.352e+02 2.733e+02 3.255e+02 5.583e+02, threshold=5.465e+02, percent-clipped=1.0 2023-04-30 02:42:24,075 INFO [train.py:904] (5/8) Epoch 15, batch 1600, loss[loss=0.1571, simple_loss=0.2454, pruned_loss=0.03443, over 17113.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2575, pruned_loss=0.04676, over 3325602.77 frames. ], batch size: 49, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:02,521 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-30 02:43:35,436 INFO [train.py:904] (5/8) Epoch 15, batch 1650, loss[loss=0.1699, simple_loss=0.2592, pruned_loss=0.04027, over 17214.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2592, pruned_loss=0.04747, over 3326216.89 frames. ], batch size: 46, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:40,918 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143756.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 02:44:23,129 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7706, 3.9905, 4.1416, 2.9259, 3.5240, 4.1017, 3.7811, 2.2419], device='cuda:5'), covar=tensor([0.0418, 0.0109, 0.0050, 0.0330, 0.0116, 0.0099, 0.0081, 0.0452], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0075, 0.0074, 0.0130, 0.0087, 0.0098, 0.0085, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 02:44:46,127 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.448e+02 2.826e+02 3.842e+02 9.668e+02, threshold=5.653e+02, percent-clipped=8.0 2023-04-30 02:44:46,149 INFO [train.py:904] (5/8) Epoch 15, batch 1700, loss[loss=0.2034, simple_loss=0.2737, pruned_loss=0.06655, over 16911.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2614, pruned_loss=0.04822, over 3316777.06 frames. ], batch size: 109, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:44:54,493 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2023-04-30 02:45:22,454 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143829.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:45:22,971 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 02:45:31,720 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6257, 2.8849, 2.7489, 4.8981, 3.9334, 4.4317, 1.6099, 3.0449], device='cuda:5'), covar=tensor([0.1537, 0.0828, 0.1167, 0.0226, 0.0261, 0.0404, 0.1658, 0.0853], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0164, 0.0184, 0.0167, 0.0197, 0.0213, 0.0188, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:45:40,089 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4748, 2.3381, 2.3777, 4.3491, 2.2001, 2.7024, 2.4094, 2.4927], device='cuda:5'), covar=tensor([0.1095, 0.3462, 0.2632, 0.0438, 0.4037, 0.2388, 0.3272, 0.3437], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0417, 0.0349, 0.0328, 0.0425, 0.0482, 0.0383, 0.0489], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:45:51,801 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1345, 3.0048, 3.1861, 1.8107, 3.3012, 3.2162, 2.6679, 2.5452], device='cuda:5'), covar=tensor([0.0790, 0.0220, 0.0215, 0.1066, 0.0100, 0.0234, 0.0467, 0.0465], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0106, 0.0093, 0.0143, 0.0075, 0.0119, 0.0128, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:45:54,347 INFO [train.py:904] (5/8) Epoch 15, batch 1750, loss[loss=0.1745, simple_loss=0.2678, pruned_loss=0.0406, over 17109.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2619, pruned_loss=0.04733, over 3325596.95 frames. ], batch size: 49, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:46:40,200 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7458, 3.8193, 2.1930, 4.1009, 2.7379, 4.0514, 2.3833, 3.0200], device='cuda:5'), covar=tensor([0.0208, 0.0311, 0.1409, 0.0243, 0.0741, 0.0515, 0.1292, 0.0608], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0173, 0.0194, 0.0150, 0.0172, 0.0216, 0.0201, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:46:42,688 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 02:46:45,502 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0208, 3.3919, 3.0460, 5.1730, 4.3276, 4.5646, 2.1297, 3.2970], device='cuda:5'), covar=tensor([0.1228, 0.0641, 0.1006, 0.0228, 0.0286, 0.0436, 0.1353, 0.0755], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0163, 0.0183, 0.0166, 0.0196, 0.0211, 0.0187, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:46:52,122 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4607, 3.9081, 4.1697, 2.2895, 3.2831, 2.7910, 3.7589, 4.1878], device='cuda:5'), covar=tensor([0.0361, 0.0804, 0.0423, 0.1799, 0.0794, 0.0848, 0.0818, 0.0914], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0151, 0.0161, 0.0146, 0.0138, 0.0125, 0.0138, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:47:05,602 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.250e+02 2.638e+02 3.028e+02 6.204e+02, threshold=5.275e+02, percent-clipped=1.0 2023-04-30 02:47:05,618 INFO [train.py:904] (5/8) Epoch 15, batch 1800, loss[loss=0.1686, simple_loss=0.2658, pruned_loss=0.03571, over 17120.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2631, pruned_loss=0.04721, over 3315881.01 frames. ], batch size: 48, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:48:05,310 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-30 02:48:15,555 INFO [train.py:904] (5/8) Epoch 15, batch 1850, loss[loss=0.1363, simple_loss=0.2255, pruned_loss=0.02356, over 17227.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2638, pruned_loss=0.04696, over 3329282.44 frames. ], batch size: 45, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:48:55,191 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 02:48:59,009 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8910, 5.2834, 5.0431, 5.0500, 4.7388, 4.6173, 4.6374, 5.3826], device='cuda:5'), covar=tensor([0.1167, 0.0884, 0.0962, 0.0758, 0.0905, 0.1126, 0.1178, 0.0830], device='cuda:5'), in_proj_covar=tensor([0.0616, 0.0765, 0.0627, 0.0548, 0.0482, 0.0491, 0.0638, 0.0578], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:49:09,704 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2165, 5.1891, 4.9663, 4.4381, 5.0623, 1.8648, 4.8474, 4.9139], device='cuda:5'), covar=tensor([0.0080, 0.0070, 0.0181, 0.0392, 0.0090, 0.2735, 0.0129, 0.0205], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0137, 0.0186, 0.0170, 0.0157, 0.0198, 0.0173, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:49:30,763 INFO [train.py:904] (5/8) Epoch 15, batch 1900, loss[loss=0.1615, simple_loss=0.2456, pruned_loss=0.03871, over 16740.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2635, pruned_loss=0.04656, over 3328971.64 frames. ], batch size: 39, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:49:31,846 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.211e+02 2.636e+02 2.995e+02 6.158e+02, threshold=5.272e+02, percent-clipped=2.0 2023-04-30 02:49:34,503 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2358, 5.2380, 4.9879, 4.5010, 5.0911, 1.9135, 4.8540, 4.9947], device='cuda:5'), covar=tensor([0.0085, 0.0073, 0.0182, 0.0369, 0.0091, 0.2504, 0.0121, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0137, 0.0185, 0.0169, 0.0157, 0.0197, 0.0172, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:49:47,092 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6633, 6.0432, 5.7219, 5.8356, 5.4064, 5.3515, 5.4197, 6.1223], device='cuda:5'), covar=tensor([0.1155, 0.0851, 0.1097, 0.0705, 0.0909, 0.0699, 0.1092, 0.0901], device='cuda:5'), in_proj_covar=tensor([0.0613, 0.0762, 0.0626, 0.0546, 0.0481, 0.0489, 0.0637, 0.0577], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:50:39,896 INFO [train.py:904] (5/8) Epoch 15, batch 1950, loss[loss=0.1631, simple_loss=0.2567, pruned_loss=0.03478, over 17121.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2627, pruned_loss=0.04578, over 3337390.36 frames. ], batch size: 48, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:50:46,818 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144056.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:51:13,862 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2162, 4.1037, 4.2414, 4.4488, 4.5397, 4.0973, 4.2932, 4.5237], device='cuda:5'), covar=tensor([0.1549, 0.1103, 0.1481, 0.0667, 0.0632, 0.1287, 0.2280, 0.0658], device='cuda:5'), in_proj_covar=tensor([0.0604, 0.0744, 0.0894, 0.0766, 0.0571, 0.0594, 0.0604, 0.0711], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:51:49,541 INFO [train.py:904] (5/8) Epoch 15, batch 2000, loss[loss=0.1989, simple_loss=0.2699, pruned_loss=0.06395, over 16679.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2625, pruned_loss=0.04548, over 3344524.89 frames. ], batch size: 134, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:51:51,365 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.196e+02 2.584e+02 2.966e+02 4.475e+02, threshold=5.169e+02, percent-clipped=0.0 2023-04-30 02:51:52,652 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144104.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:52:10,111 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 02:52:27,284 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144129.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:52:28,618 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2189, 2.1138, 1.6708, 1.8859, 2.4059, 2.1491, 2.3535, 2.5476], device='cuda:5'), covar=tensor([0.0220, 0.0322, 0.0427, 0.0405, 0.0195, 0.0306, 0.0187, 0.0230], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0224, 0.0216, 0.0216, 0.0226, 0.0225, 0.0231, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:52:58,038 INFO [train.py:904] (5/8) Epoch 15, batch 2050, loss[loss=0.1818, simple_loss=0.2689, pruned_loss=0.04734, over 15877.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2625, pruned_loss=0.04594, over 3339701.28 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:53:32,909 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144177.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:54:07,935 INFO [train.py:904] (5/8) Epoch 15, batch 2100, loss[loss=0.1633, simple_loss=0.2454, pruned_loss=0.04063, over 15869.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2627, pruned_loss=0.04638, over 3333531.17 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:54:08,982 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.513e+02 2.931e+02 3.819e+02 1.829e+03, threshold=5.862e+02, percent-clipped=10.0 2023-04-30 02:55:17,948 INFO [train.py:904] (5/8) Epoch 15, batch 2150, loss[loss=0.1933, simple_loss=0.2916, pruned_loss=0.04751, over 16809.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2634, pruned_loss=0.04673, over 3323945.19 frames. ], batch size: 57, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:56:25,368 INFO [train.py:904] (5/8) Epoch 15, batch 2200, loss[loss=0.1684, simple_loss=0.2595, pruned_loss=0.03871, over 17197.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2641, pruned_loss=0.04703, over 3335372.75 frames. ], batch size: 46, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:56:27,077 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.322e+02 2.712e+02 3.377e+02 6.214e+02, threshold=5.423e+02, percent-clipped=1.0 2023-04-30 02:57:24,730 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 02:57:36,226 INFO [train.py:904] (5/8) Epoch 15, batch 2250, loss[loss=0.1719, simple_loss=0.2485, pruned_loss=0.04762, over 16848.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.265, pruned_loss=0.04758, over 3329990.52 frames. ], batch size: 96, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:57:47,097 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8295, 2.9933, 3.1874, 1.9650, 2.7527, 2.2710, 3.3884, 3.2694], device='cuda:5'), covar=tensor([0.0247, 0.0854, 0.0576, 0.1809, 0.0796, 0.0929, 0.0504, 0.0869], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0154, 0.0162, 0.0147, 0.0140, 0.0126, 0.0139, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 02:58:40,543 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3426, 2.2055, 2.3251, 4.1818, 2.2841, 2.6273, 2.2986, 2.4061], device='cuda:5'), covar=tensor([0.1271, 0.3640, 0.2700, 0.0525, 0.3626, 0.2405, 0.3747, 0.3051], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0416, 0.0349, 0.0326, 0.0423, 0.0480, 0.0382, 0.0487], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:58:46,694 INFO [train.py:904] (5/8) Epoch 15, batch 2300, loss[loss=0.1822, simple_loss=0.2854, pruned_loss=0.0395, over 17155.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2653, pruned_loss=0.04793, over 3329494.08 frames. ], batch size: 49, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:58:47,878 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.350e+02 2.907e+02 3.506e+02 6.150e+02, threshold=5.814e+02, percent-clipped=3.0 2023-04-30 02:58:59,098 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144411.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:59:21,092 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 02:59:36,571 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-30 02:59:51,633 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6628, 2.3066, 2.3265, 4.4792, 2.2137, 2.8275, 2.4339, 2.4990], device='cuda:5'), covar=tensor([0.1037, 0.3761, 0.2694, 0.0414, 0.3987, 0.2396, 0.3420, 0.3515], device='cuda:5'), in_proj_covar=tensor([0.0381, 0.0418, 0.0350, 0.0328, 0.0425, 0.0482, 0.0384, 0.0489], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 02:59:53,226 INFO [train.py:904] (5/8) Epoch 15, batch 2350, loss[loss=0.19, simple_loss=0.2823, pruned_loss=0.04889, over 17039.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2652, pruned_loss=0.04781, over 3329994.98 frames. ], batch size: 55, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 03:00:20,695 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:01:02,761 INFO [train.py:904] (5/8) Epoch 15, batch 2400, loss[loss=0.1815, simple_loss=0.2822, pruned_loss=0.04043, over 17052.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.266, pruned_loss=0.0478, over 3329717.27 frames. ], batch size: 53, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:01:04,729 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.391e+02 2.805e+02 3.317e+02 7.772e+02, threshold=5.609e+02, percent-clipped=1.0 2023-04-30 03:01:24,754 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8976, 1.9162, 2.4368, 2.8301, 2.6530, 3.3910, 2.2061, 3.2452], device='cuda:5'), covar=tensor([0.0202, 0.0410, 0.0310, 0.0275, 0.0301, 0.0151, 0.0410, 0.0147], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0183, 0.0168, 0.0172, 0.0181, 0.0138, 0.0183, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:01:47,046 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 03:02:10,317 INFO [train.py:904] (5/8) Epoch 15, batch 2450, loss[loss=0.2019, simple_loss=0.2715, pruned_loss=0.06617, over 16879.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2656, pruned_loss=0.04711, over 3336014.36 frames. ], batch size: 116, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:02:19,263 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1266, 3.9826, 4.3546, 2.1644, 4.5690, 4.5882, 3.3911, 3.5992], device='cuda:5'), covar=tensor([0.0641, 0.0221, 0.0191, 0.1131, 0.0060, 0.0152, 0.0376, 0.0344], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0105, 0.0091, 0.0139, 0.0073, 0.0117, 0.0125, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 03:02:25,728 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 03:02:26,872 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 03:02:55,761 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 03:03:13,097 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-04-30 03:03:17,681 INFO [train.py:904] (5/8) Epoch 15, batch 2500, loss[loss=0.1657, simple_loss=0.245, pruned_loss=0.04323, over 15891.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2654, pruned_loss=0.04666, over 3339728.01 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:03:18,672 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.258e+02 2.678e+02 3.424e+02 5.626e+02, threshold=5.355e+02, percent-clipped=1.0 2023-04-30 03:04:26,860 INFO [train.py:904] (5/8) Epoch 15, batch 2550, loss[loss=0.1928, simple_loss=0.2643, pruned_loss=0.06069, over 16889.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2662, pruned_loss=0.04718, over 3341160.16 frames. ], batch size: 109, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:34,895 INFO [train.py:904] (5/8) Epoch 15, batch 2600, loss[loss=0.1942, simple_loss=0.2864, pruned_loss=0.05096, over 16700.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2653, pruned_loss=0.04654, over 3339480.44 frames. ], batch size: 62, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:36,079 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.335e+02 2.561e+02 3.164e+02 7.288e+02, threshold=5.122e+02, percent-clipped=2.0 2023-04-30 03:06:25,336 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3609, 4.3837, 4.7820, 4.7848, 4.8250, 4.4586, 4.4883, 4.2342], device='cuda:5'), covar=tensor([0.0382, 0.0559, 0.0416, 0.0451, 0.0517, 0.0422, 0.0878, 0.0628], device='cuda:5'), in_proj_covar=tensor([0.0372, 0.0396, 0.0392, 0.0376, 0.0439, 0.0416, 0.0507, 0.0332], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 03:06:43,559 INFO [train.py:904] (5/8) Epoch 15, batch 2650, loss[loss=0.1861, simple_loss=0.2834, pruned_loss=0.04434, over 17116.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2654, pruned_loss=0.04575, over 3340486.07 frames. ], batch size: 49, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:07:05,914 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:07:52,873 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7890, 1.7917, 2.2532, 2.6465, 2.6880, 2.7073, 1.7558, 2.9120], device='cuda:5'), covar=tensor([0.0131, 0.0431, 0.0282, 0.0210, 0.0236, 0.0216, 0.0465, 0.0122], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0183, 0.0169, 0.0172, 0.0182, 0.0138, 0.0183, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:07:52,908 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8604, 2.8785, 2.5306, 4.2085, 3.5149, 4.1068, 1.5913, 2.9570], device='cuda:5'), covar=tensor([0.1300, 0.0618, 0.1070, 0.0159, 0.0202, 0.0388, 0.1413, 0.0734], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0166, 0.0186, 0.0170, 0.0202, 0.0215, 0.0189, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:07:53,572 INFO [train.py:904] (5/8) Epoch 15, batch 2700, loss[loss=0.1937, simple_loss=0.2742, pruned_loss=0.05661, over 16873.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.266, pruned_loss=0.04585, over 3339907.38 frames. ], batch size: 109, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:07:54,732 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.168e+02 2.529e+02 3.023e+02 4.642e+02, threshold=5.059e+02, percent-clipped=0.0 2023-04-30 03:08:45,759 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5991, 2.5449, 1.9985, 2.3146, 2.9351, 2.6842, 3.4622, 3.2234], device='cuda:5'), covar=tensor([0.0144, 0.0460, 0.0602, 0.0499, 0.0322, 0.0413, 0.0224, 0.0263], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0225, 0.0214, 0.0215, 0.0225, 0.0224, 0.0231, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:09:02,453 INFO [train.py:904] (5/8) Epoch 15, batch 2750, loss[loss=0.1656, simple_loss=0.2651, pruned_loss=0.03307, over 17052.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2656, pruned_loss=0.04548, over 3342481.97 frames. ], batch size: 50, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:10:11,029 INFO [train.py:904] (5/8) Epoch 15, batch 2800, loss[loss=0.1774, simple_loss=0.2516, pruned_loss=0.05161, over 16547.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2651, pruned_loss=0.04516, over 3341894.56 frames. ], batch size: 146, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:10:12,145 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.181e+02 2.488e+02 3.014e+02 5.995e+02, threshold=4.976e+02, percent-clipped=2.0 2023-04-30 03:11:21,055 INFO [train.py:904] (5/8) Epoch 15, batch 2850, loss[loss=0.193, simple_loss=0.2595, pruned_loss=0.06326, over 16408.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2644, pruned_loss=0.0451, over 3329267.33 frames. ], batch size: 146, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:08,801 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-04-30 03:12:22,226 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144996.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:12:31,828 INFO [train.py:904] (5/8) Epoch 15, batch 2900, loss[loss=0.1365, simple_loss=0.2292, pruned_loss=0.0219, over 17220.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2635, pruned_loss=0.04561, over 3327787.11 frames. ], batch size: 46, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:33,009 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.445e+02 2.844e+02 3.300e+02 6.709e+02, threshold=5.687e+02, percent-clipped=6.0 2023-04-30 03:12:37,412 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0626, 3.8389, 4.2211, 1.9981, 4.4047, 4.4796, 3.1450, 3.4409], device='cuda:5'), covar=tensor([0.0640, 0.0215, 0.0189, 0.1164, 0.0071, 0.0172, 0.0412, 0.0370], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0106, 0.0092, 0.0140, 0.0074, 0.0118, 0.0125, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 03:13:06,921 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:13:07,017 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8870, 2.8471, 2.5257, 4.6446, 3.6429, 4.2782, 1.7245, 3.1704], device='cuda:5'), covar=tensor([0.1372, 0.0798, 0.1312, 0.0225, 0.0250, 0.0440, 0.1611, 0.0763], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0164, 0.0185, 0.0169, 0.0199, 0.0212, 0.0188, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:13:28,252 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-04-30 03:13:40,930 INFO [train.py:904] (5/8) Epoch 15, batch 2950, loss[loss=0.2018, simple_loss=0.2688, pruned_loss=0.06738, over 16774.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2627, pruned_loss=0.04556, over 3336850.60 frames. ], batch size: 83, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:13:47,664 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145057.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:14:01,017 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145066.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:02,252 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145067.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:31,897 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:32,232 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 03:14:33,849 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 03:14:35,004 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5910, 3.9090, 4.2397, 2.4586, 3.2979, 2.6931, 4.2526, 4.1570], device='cuda:5'), covar=tensor([0.0280, 0.0810, 0.0413, 0.1700, 0.0752, 0.0920, 0.0531, 0.0911], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0154, 0.0162, 0.0147, 0.0139, 0.0126, 0.0140, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:14:49,686 INFO [train.py:904] (5/8) Epoch 15, batch 3000, loss[loss=0.1767, simple_loss=0.2628, pruned_loss=0.0453, over 17228.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2629, pruned_loss=0.046, over 3335959.54 frames. ], batch size: 45, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:14:49,687 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 03:14:58,806 INFO [train.py:938] (5/8) Epoch 15, validation: loss=0.138, simple_loss=0.2438, pruned_loss=0.01616, over 944034.00 frames. 2023-04-30 03:14:58,807 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 03:15:00,799 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.402e+02 2.841e+02 3.286e+02 6.614e+02, threshold=5.681e+02, percent-clipped=1.0 2023-04-30 03:15:17,419 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145115.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:15:34,582 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145127.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:16:07,543 INFO [train.py:904] (5/8) Epoch 15, batch 3050, loss[loss=0.2187, simple_loss=0.2952, pruned_loss=0.07108, over 12874.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2634, pruned_loss=0.04638, over 3327962.54 frames. ], batch size: 247, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:17:18,181 INFO [train.py:904] (5/8) Epoch 15, batch 3100, loss[loss=0.17, simple_loss=0.2391, pruned_loss=0.05046, over 16805.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2628, pruned_loss=0.04635, over 3324991.32 frames. ], batch size: 102, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:17:19,338 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.571e+02 2.445e+02 2.806e+02 3.389e+02 5.168e+02, threshold=5.611e+02, percent-clipped=0.0 2023-04-30 03:17:43,522 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9541, 3.1781, 2.7787, 5.0538, 4.1834, 4.5721, 1.6388, 3.2528], device='cuda:5'), covar=tensor([0.1302, 0.0687, 0.1124, 0.0182, 0.0254, 0.0362, 0.1556, 0.0702], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0171, 0.0201, 0.0213, 0.0188, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:18:28,450 INFO [train.py:904] (5/8) Epoch 15, batch 3150, loss[loss=0.1645, simple_loss=0.2581, pruned_loss=0.03546, over 17210.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2617, pruned_loss=0.04629, over 3329639.37 frames. ], batch size: 45, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:19:37,247 INFO [train.py:904] (5/8) Epoch 15, batch 3200, loss[loss=0.1938, simple_loss=0.2661, pruned_loss=0.06073, over 16772.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.261, pruned_loss=0.04575, over 3333855.46 frames. ], batch size: 124, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:19:38,468 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.241e+02 2.735e+02 3.234e+02 5.514e+02, threshold=5.469e+02, percent-clipped=0.0 2023-04-30 03:19:41,612 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2369, 5.2183, 5.0141, 4.4038, 5.0305, 1.7602, 4.8102, 4.9889], device='cuda:5'), covar=tensor([0.0082, 0.0077, 0.0185, 0.0423, 0.0100, 0.2770, 0.0142, 0.0185], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0140, 0.0189, 0.0175, 0.0160, 0.0200, 0.0176, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:20:29,878 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5034, 3.6040, 2.1621, 3.7442, 2.7615, 3.7214, 2.2308, 2.8488], device='cuda:5'), covar=tensor([0.0219, 0.0300, 0.1431, 0.0277, 0.0703, 0.0675, 0.1291, 0.0598], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0172, 0.0191, 0.0152, 0.0169, 0.0216, 0.0199, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:20:29,924 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7223, 2.7266, 2.4069, 4.0980, 3.4524, 4.1063, 1.4803, 2.8349], device='cuda:5'), covar=tensor([0.1325, 0.0593, 0.1093, 0.0142, 0.0131, 0.0319, 0.1426, 0.0780], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0164, 0.0184, 0.0170, 0.0200, 0.0212, 0.0187, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:20:46,527 INFO [train.py:904] (5/8) Epoch 15, batch 3250, loss[loss=0.186, simple_loss=0.2746, pruned_loss=0.04871, over 16901.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2604, pruned_loss=0.04569, over 3333530.75 frames. ], batch size: 96, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:20:46,789 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145352.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:20:57,664 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5804, 5.9815, 5.6749, 5.8043, 5.3282, 5.3467, 5.4024, 6.0570], device='cuda:5'), covar=tensor([0.1370, 0.0891, 0.1095, 0.0702, 0.0983, 0.0692, 0.1056, 0.0891], device='cuda:5'), in_proj_covar=tensor([0.0630, 0.0787, 0.0643, 0.0562, 0.0497, 0.0501, 0.0654, 0.0596], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:21:13,401 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:30,556 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145383.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:57,382 INFO [train.py:904] (5/8) Epoch 15, batch 3300, loss[loss=0.1741, simple_loss=0.26, pruned_loss=0.04409, over 16653.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2616, pruned_loss=0.04597, over 3335882.05 frames. ], batch size: 76, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:21:58,628 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.241e+02 2.723e+02 3.266e+02 5.157e+02, threshold=5.447e+02, percent-clipped=0.0 2023-04-30 03:22:03,197 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2927, 3.6718, 3.8284, 2.0809, 3.1272, 2.4046, 3.7114, 3.7846], device='cuda:5'), covar=tensor([0.0288, 0.0786, 0.0476, 0.1842, 0.0757, 0.0936, 0.0595, 0.0912], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0155, 0.0162, 0.0147, 0.0139, 0.0126, 0.0140, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:22:25,245 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:22:38,882 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:23:06,160 INFO [train.py:904] (5/8) Epoch 15, batch 3350, loss[loss=0.1667, simple_loss=0.2464, pruned_loss=0.04355, over 16871.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2625, pruned_loss=0.0462, over 3326165.16 frames. ], batch size: 96, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:23:37,944 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7726, 4.8509, 5.0399, 4.8392, 4.8234, 5.4885, 5.0561, 4.7333], device='cuda:5'), covar=tensor([0.1217, 0.1810, 0.2092, 0.2163, 0.2925, 0.1004, 0.1427, 0.2547], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0557, 0.0611, 0.0475, 0.0638, 0.0637, 0.0483, 0.0631], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 03:24:17,515 INFO [train.py:904] (5/8) Epoch 15, batch 3400, loss[loss=0.1517, simple_loss=0.2452, pruned_loss=0.02914, over 17234.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2624, pruned_loss=0.04613, over 3319525.90 frames. ], batch size: 45, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:18,610 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.153e+02 2.635e+02 3.209e+02 5.771e+02, threshold=5.270e+02, percent-clipped=2.0 2023-04-30 03:24:37,707 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 03:25:17,186 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 03:25:28,525 INFO [train.py:904] (5/8) Epoch 15, batch 3450, loss[loss=0.169, simple_loss=0.2447, pruned_loss=0.04668, over 16854.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2613, pruned_loss=0.04591, over 3315940.97 frames. ], batch size: 90, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:26:27,273 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145594.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:26:38,102 INFO [train.py:904] (5/8) Epoch 15, batch 3500, loss[loss=0.2144, simple_loss=0.2913, pruned_loss=0.06869, over 16325.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2605, pruned_loss=0.04581, over 3317431.13 frames. ], batch size: 165, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:26:39,234 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 2.268e+02 2.638e+02 3.199e+02 5.613e+02, threshold=5.276e+02, percent-clipped=1.0 2023-04-30 03:27:35,953 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:27:47,277 INFO [train.py:904] (5/8) Epoch 15, batch 3550, loss[loss=0.1911, simple_loss=0.2695, pruned_loss=0.0563, over 16796.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2594, pruned_loss=0.0456, over 3320087.07 frames. ], batch size: 102, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:27:48,312 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145652.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:27:53,429 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145655.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:27:57,219 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8781, 4.0565, 2.6529, 4.6537, 3.0946, 4.6231, 2.5919, 3.3548], device='cuda:5'), covar=tensor([0.0266, 0.0332, 0.1307, 0.0204, 0.0714, 0.0473, 0.1341, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0152, 0.0170, 0.0215, 0.0200, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:28:13,762 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3647, 4.2315, 4.3917, 4.5411, 4.6813, 4.2475, 4.4687, 4.6428], device='cuda:5'), covar=tensor([0.1593, 0.0974, 0.1367, 0.0705, 0.0578, 0.1049, 0.1855, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0620, 0.0767, 0.0921, 0.0787, 0.0586, 0.0612, 0.0617, 0.0724], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:28:27,671 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5255, 2.2886, 1.7181, 2.0276, 2.6866, 2.4030, 2.6627, 2.7736], device='cuda:5'), covar=tensor([0.0151, 0.0333, 0.0474, 0.0390, 0.0189, 0.0278, 0.0174, 0.0209], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0222, 0.0213, 0.0215, 0.0224, 0.0224, 0.0232, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:28:31,882 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145683.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:31,956 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4227, 2.2617, 1.7641, 2.0668, 2.5913, 2.3522, 2.5750, 2.7157], device='cuda:5'), covar=tensor([0.0184, 0.0312, 0.0461, 0.0364, 0.0200, 0.0282, 0.0202, 0.0217], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0222, 0.0213, 0.0216, 0.0224, 0.0224, 0.0232, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:28:42,605 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3312, 3.1585, 3.4820, 1.8696, 3.5993, 3.6057, 2.9896, 2.6688], device='cuda:5'), covar=tensor([0.0721, 0.0224, 0.0169, 0.1067, 0.0085, 0.0180, 0.0357, 0.0438], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0106, 0.0091, 0.0139, 0.0073, 0.0118, 0.0123, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 03:28:50,430 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1705, 4.6442, 3.1705, 2.4571, 3.1658, 2.6335, 4.8143, 3.9810], device='cuda:5'), covar=tensor([0.2439, 0.0578, 0.1782, 0.2537, 0.2710, 0.2031, 0.0392, 0.1230], device='cuda:5'), in_proj_covar=tensor([0.0313, 0.0267, 0.0296, 0.0296, 0.0291, 0.0241, 0.0281, 0.0321], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 03:28:54,823 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145700.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:57,600 INFO [train.py:904] (5/8) Epoch 15, batch 3600, loss[loss=0.1966, simple_loss=0.2846, pruned_loss=0.05431, over 16705.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2585, pruned_loss=0.0451, over 3326197.33 frames. ], batch size: 57, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:28:58,726 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.177e+02 2.263e+02 2.507e+02 3.007e+02 5.256e+02, threshold=5.015e+02, percent-clipped=0.0 2023-04-30 03:29:00,366 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:27,023 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145722.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:34,304 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145727.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:39,861 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145731.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:47,465 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1482, 3.4075, 3.5498, 2.4810, 3.3061, 3.6998, 3.4732, 2.0238], device='cuda:5'), covar=tensor([0.0473, 0.0110, 0.0051, 0.0339, 0.0085, 0.0079, 0.0079, 0.0399], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0075, 0.0075, 0.0130, 0.0088, 0.0100, 0.0086, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 03:30:10,705 INFO [train.py:904] (5/8) Epoch 15, batch 3650, loss[loss=0.1896, simple_loss=0.2715, pruned_loss=0.0539, over 15591.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2581, pruned_loss=0.04606, over 3316071.04 frames. ], batch size: 191, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:30:37,278 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6176, 3.6771, 2.1333, 3.8719, 2.7912, 3.8137, 2.2514, 2.8996], device='cuda:5'), covar=tensor([0.0222, 0.0392, 0.1580, 0.0278, 0.0698, 0.0760, 0.1366, 0.0621], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0172, 0.0191, 0.0152, 0.0169, 0.0216, 0.0200, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:30:38,224 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:30:58,535 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 03:31:24,501 INFO [train.py:904] (5/8) Epoch 15, batch 3700, loss[loss=0.1782, simple_loss=0.2562, pruned_loss=0.05007, over 16732.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.257, pruned_loss=0.04761, over 3291184.81 frames. ], batch size: 124, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:31:26,282 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.174e+02 2.701e+02 3.170e+02 5.249e+02, threshold=5.403e+02, percent-clipped=2.0 2023-04-30 03:31:53,406 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145821.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:32:38,918 INFO [train.py:904] (5/8) Epoch 15, batch 3750, loss[loss=0.1866, simple_loss=0.272, pruned_loss=0.05062, over 16620.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2578, pruned_loss=0.04898, over 3296004.28 frames. ], batch size: 62, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:23,527 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145882.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:33:45,455 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6144, 2.4857, 2.3128, 3.6385, 2.9249, 3.8169, 1.4612, 2.6267], device='cuda:5'), covar=tensor([0.1402, 0.0698, 0.1172, 0.0213, 0.0185, 0.0368, 0.1559, 0.0859], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0165, 0.0185, 0.0170, 0.0201, 0.0212, 0.0188, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:33:51,794 INFO [train.py:904] (5/8) Epoch 15, batch 3800, loss[loss=0.1696, simple_loss=0.2439, pruned_loss=0.04769, over 16892.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2582, pruned_loss=0.05022, over 3290814.78 frames. ], batch size: 109, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:52,974 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9189, 3.9731, 4.2754, 4.2832, 4.3002, 4.0221, 4.0892, 3.9919], device='cuda:5'), covar=tensor([0.0336, 0.0659, 0.0413, 0.0383, 0.0465, 0.0443, 0.0739, 0.0589], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0408, 0.0403, 0.0382, 0.0450, 0.0424, 0.0520, 0.0339], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 03:33:53,682 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.221e+02 2.643e+02 3.115e+02 5.182e+02, threshold=5.286e+02, percent-clipped=0.0 2023-04-30 03:34:30,785 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8742, 2.9572, 2.5337, 5.0184, 3.9923, 4.3788, 1.6826, 2.8786], device='cuda:5'), covar=tensor([0.1219, 0.0719, 0.1306, 0.0119, 0.0355, 0.0381, 0.1439, 0.0992], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0166, 0.0186, 0.0170, 0.0201, 0.0212, 0.0188, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 03:35:01,662 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145950.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:35:04,375 INFO [train.py:904] (5/8) Epoch 15, batch 3850, loss[loss=0.1662, simple_loss=0.2431, pruned_loss=0.04471, over 16425.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2583, pruned_loss=0.05095, over 3292211.46 frames. ], batch size: 75, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:13,426 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145999.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:36:21,016 INFO [train.py:904] (5/8) Epoch 15, batch 3900, loss[loss=0.1844, simple_loss=0.2582, pruned_loss=0.05533, over 16848.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2581, pruned_loss=0.05127, over 3300067.99 frames. ], batch size: 116, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:22,209 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.255e+02 2.647e+02 3.185e+02 6.041e+02, threshold=5.295e+02, percent-clipped=2.0 2023-04-30 03:36:53,532 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1363, 5.6532, 5.8463, 5.5505, 5.4648, 6.1684, 5.6967, 5.4036], device='cuda:5'), covar=tensor([0.0737, 0.1488, 0.1543, 0.1660, 0.2362, 0.0827, 0.1245, 0.2049], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0552, 0.0604, 0.0471, 0.0627, 0.0627, 0.0480, 0.0623], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 03:36:57,955 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:37:19,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4493, 3.4343, 2.6730, 2.1169, 2.2457, 2.2519, 3.4644, 3.1094], device='cuda:5'), covar=tensor([0.2738, 0.0681, 0.1680, 0.2567, 0.2581, 0.2003, 0.0546, 0.1210], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0265, 0.0294, 0.0293, 0.0290, 0.0238, 0.0279, 0.0318], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 03:37:32,707 INFO [train.py:904] (5/8) Epoch 15, batch 3950, loss[loss=0.1652, simple_loss=0.2369, pruned_loss=0.04676, over 16802.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2582, pruned_loss=0.05166, over 3291941.34 frames. ], batch size: 90, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:06,972 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:38:46,145 INFO [train.py:904] (5/8) Epoch 15, batch 4000, loss[loss=0.2006, simple_loss=0.2819, pruned_loss=0.05965, over 16192.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2586, pruned_loss=0.05226, over 3285361.96 frames. ], batch size: 165, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:47,411 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.290e+02 2.701e+02 3.084e+02 7.730e+02, threshold=5.403e+02, percent-clipped=2.0 2023-04-30 03:39:56,276 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:39:59,985 INFO [train.py:904] (5/8) Epoch 15, batch 4050, loss[loss=0.1628, simple_loss=0.2434, pruned_loss=0.04107, over 16614.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2593, pruned_loss=0.05158, over 3267213.77 frames. ], batch size: 57, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:40:36,989 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146177.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:41:13,957 INFO [train.py:904] (5/8) Epoch 15, batch 4100, loss[loss=0.2233, simple_loss=0.3, pruned_loss=0.07331, over 12368.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2603, pruned_loss=0.05048, over 3253384.52 frames. ], batch size: 248, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:41:15,752 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 1.975e+02 2.401e+02 2.875e+02 5.931e+02, threshold=4.803e+02, percent-clipped=1.0 2023-04-30 03:41:26,675 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146210.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:41:26,778 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4305, 3.2475, 2.6143, 2.1191, 2.2427, 2.2385, 3.3764, 3.1318], device='cuda:5'), covar=tensor([0.2777, 0.0678, 0.1692, 0.2378, 0.2350, 0.2034, 0.0496, 0.1083], device='cuda:5'), in_proj_covar=tensor([0.0312, 0.0266, 0.0295, 0.0295, 0.0291, 0.0239, 0.0280, 0.0319], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 03:41:52,060 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1647, 3.4665, 3.5766, 3.5662, 3.5467, 3.4063, 3.4304, 3.4287], device='cuda:5'), covar=tensor([0.0386, 0.0544, 0.0419, 0.0404, 0.0510, 0.0466, 0.0748, 0.0505], device='cuda:5'), in_proj_covar=tensor([0.0373, 0.0403, 0.0395, 0.0375, 0.0444, 0.0416, 0.0511, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 03:42:30,247 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146250.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:42:32,945 INFO [train.py:904] (5/8) Epoch 15, batch 4150, loss[loss=0.1955, simple_loss=0.2915, pruned_loss=0.04974, over 16538.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2675, pruned_loss=0.05298, over 3220213.12 frames. ], batch size: 75, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:42:59,753 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 03:43:45,580 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:46,734 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:50,404 INFO [train.py:904] (5/8) Epoch 15, batch 4200, loss[loss=0.2296, simple_loss=0.3029, pruned_loss=0.07818, over 11377.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.274, pruned_loss=0.0546, over 3182731.68 frames. ], batch size: 246, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:43:53,467 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.342e+02 2.800e+02 3.448e+02 4.997e+02, threshold=5.600e+02, percent-clipped=3.0 2023-04-30 03:44:30,150 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8539, 5.1649, 4.9540, 4.9629, 4.7080, 4.6682, 4.5612, 5.2600], device='cuda:5'), covar=tensor([0.1094, 0.0801, 0.0897, 0.0736, 0.0774, 0.0833, 0.1064, 0.0740], device='cuda:5'), in_proj_covar=tensor([0.0610, 0.0761, 0.0627, 0.0549, 0.0479, 0.0487, 0.0635, 0.0578], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:44:59,536 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:45:06,399 INFO [train.py:904] (5/8) Epoch 15, batch 4250, loss[loss=0.1753, simple_loss=0.266, pruned_loss=0.04228, over 15436.00 frames. ], tot_loss[loss=0.194, simple_loss=0.278, pruned_loss=0.05493, over 3156717.20 frames. ], batch size: 191, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:19,359 INFO [train.py:904] (5/8) Epoch 15, batch 4300, loss[loss=0.1964, simple_loss=0.2749, pruned_loss=0.059, over 11784.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2798, pruned_loss=0.05409, over 3170492.98 frames. ], batch size: 246, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:23,349 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.412e+02 2.971e+02 3.359e+02 7.082e+02, threshold=5.941e+02, percent-clipped=4.0 2023-04-30 03:47:31,147 INFO [train.py:904] (5/8) Epoch 15, batch 4350, loss[loss=0.2028, simple_loss=0.2805, pruned_loss=0.06254, over 11574.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2828, pruned_loss=0.05501, over 3161427.99 frames. ], batch size: 248, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:48:08,951 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146477.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:48:40,832 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7098, 4.7259, 4.5087, 3.9019, 4.6563, 1.5744, 4.4170, 4.2003], device='cuda:5'), covar=tensor([0.0058, 0.0042, 0.0136, 0.0277, 0.0056, 0.2832, 0.0086, 0.0197], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0136, 0.0186, 0.0172, 0.0157, 0.0196, 0.0174, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:48:45,759 INFO [train.py:904] (5/8) Epoch 15, batch 4400, loss[loss=0.2073, simple_loss=0.2959, pruned_loss=0.0594, over 17126.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2851, pruned_loss=0.05596, over 3161257.46 frames. ], batch size: 47, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:48:50,401 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.574e+02 2.972e+02 3.574e+02 6.742e+02, threshold=5.944e+02, percent-clipped=2.0 2023-04-30 03:48:51,429 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146505.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:49:21,074 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146525.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:49:58,636 INFO [train.py:904] (5/8) Epoch 15, batch 4450, loss[loss=0.2083, simple_loss=0.2916, pruned_loss=0.06251, over 16337.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2888, pruned_loss=0.05721, over 3171119.50 frames. ], batch size: 35, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:50:21,450 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7295, 3.8446, 4.0973, 4.0756, 4.0834, 3.8517, 3.8645, 3.8085], device='cuda:5'), covar=tensor([0.0324, 0.0526, 0.0358, 0.0381, 0.0436, 0.0423, 0.0821, 0.0507], device='cuda:5'), in_proj_covar=tensor([0.0360, 0.0387, 0.0382, 0.0362, 0.0429, 0.0403, 0.0495, 0.0323], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 03:50:30,413 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0153, 4.0998, 4.4233, 4.3957, 4.3988, 4.1179, 4.1386, 4.0108], device='cuda:5'), covar=tensor([0.0294, 0.0447, 0.0305, 0.0349, 0.0419, 0.0392, 0.0790, 0.0539], device='cuda:5'), in_proj_covar=tensor([0.0360, 0.0387, 0.0382, 0.0362, 0.0428, 0.0403, 0.0494, 0.0323], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 03:50:57,171 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146591.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:51:12,440 INFO [train.py:904] (5/8) Epoch 15, batch 4500, loss[loss=0.2075, simple_loss=0.2825, pruned_loss=0.06625, over 16873.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2891, pruned_loss=0.05787, over 3172205.15 frames. ], batch size: 42, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:51:16,067 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 1.915e+02 2.182e+02 2.470e+02 4.715e+02, threshold=4.363e+02, percent-clipped=0.0 2023-04-30 03:51:25,054 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-30 03:51:47,932 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:51:48,220 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 03:52:25,489 INFO [train.py:904] (5/8) Epoch 15, batch 4550, loss[loss=0.2045, simple_loss=0.2903, pruned_loss=0.05934, over 16401.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2892, pruned_loss=0.05841, over 3182326.05 frames. ], batch size: 68, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:52:25,966 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146652.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:52:37,163 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 03:53:16,753 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146687.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 03:53:37,400 INFO [train.py:904] (5/8) Epoch 15, batch 4600, loss[loss=0.1879, simple_loss=0.2721, pruned_loss=0.0518, over 16623.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2898, pruned_loss=0.05811, over 3193645.51 frames. ], batch size: 57, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:53:41,728 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 1.940e+02 2.192e+02 2.697e+02 4.440e+02, threshold=4.384e+02, percent-clipped=1.0 2023-04-30 03:53:50,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5358, 5.5225, 5.2497, 4.6739, 5.5091, 1.8958, 5.1956, 5.0490], device='cuda:5'), covar=tensor([0.0039, 0.0034, 0.0100, 0.0253, 0.0035, 0.2483, 0.0070, 0.0119], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0133, 0.0184, 0.0169, 0.0154, 0.0193, 0.0171, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:54:49,173 INFO [train.py:904] (5/8) Epoch 15, batch 4650, loss[loss=0.208, simple_loss=0.2926, pruned_loss=0.06171, over 16480.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2892, pruned_loss=0.05867, over 3187786.66 frames. ], batch size: 146, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:55:36,047 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0861, 1.9096, 2.5765, 2.9496, 2.7594, 3.4104, 2.0399, 3.3910], device='cuda:5'), covar=tensor([0.0143, 0.0439, 0.0245, 0.0223, 0.0256, 0.0131, 0.0440, 0.0104], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0185, 0.0169, 0.0175, 0.0184, 0.0140, 0.0186, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:55:48,361 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3195, 2.4426, 2.0252, 2.3073, 2.8752, 2.4316, 2.9990, 3.0819], device='cuda:5'), covar=tensor([0.0081, 0.0322, 0.0444, 0.0364, 0.0182, 0.0311, 0.0149, 0.0196], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0217, 0.0211, 0.0210, 0.0217, 0.0218, 0.0223, 0.0214], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:56:03,174 INFO [train.py:904] (5/8) Epoch 15, batch 4700, loss[loss=0.1784, simple_loss=0.268, pruned_loss=0.04436, over 16492.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2864, pruned_loss=0.05758, over 3194074.91 frames. ], batch size: 75, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:56:07,890 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 1.906e+02 2.324e+02 2.777e+02 7.777e+02, threshold=4.648e+02, percent-clipped=3.0 2023-04-30 03:56:08,857 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146805.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:07,667 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:18,598 INFO [train.py:904] (5/8) Epoch 15, batch 4750, loss[loss=0.1678, simple_loss=0.2557, pruned_loss=0.03997, over 16651.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2821, pruned_loss=0.05546, over 3196955.67 frames. ], batch size: 68, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:57:20,575 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:58:27,332 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 03:58:31,091 INFO [train.py:904] (5/8) Epoch 15, batch 4800, loss[loss=0.1913, simple_loss=0.2861, pruned_loss=0.04826, over 16887.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2782, pruned_loss=0.05309, over 3204817.33 frames. ], batch size: 96, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:58:36,183 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.051e+02 1.856e+02 2.167e+02 2.629e+02 4.962e+02, threshold=4.334e+02, percent-clipped=1.0 2023-04-30 03:58:36,830 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146905.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:59:08,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9443, 1.9915, 2.1711, 3.6224, 2.0037, 2.3720, 2.0959, 2.1828], device='cuda:5'), covar=tensor([0.1324, 0.3669, 0.2576, 0.0499, 0.3865, 0.2334, 0.3601, 0.3111], device='cuda:5'), in_proj_covar=tensor([0.0377, 0.0416, 0.0344, 0.0320, 0.0421, 0.0478, 0.0381, 0.0486], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 03:59:40,651 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:59:47,626 INFO [train.py:904] (5/8) Epoch 15, batch 4850, loss[loss=0.2084, simple_loss=0.2931, pruned_loss=0.06189, over 16701.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2788, pruned_loss=0.05211, over 3200422.64 frames. ], batch size: 62, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:00:34,906 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146982.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 04:01:03,675 INFO [train.py:904] (5/8) Epoch 15, batch 4900, loss[loss=0.1716, simple_loss=0.2634, pruned_loss=0.03993, over 16792.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2781, pruned_loss=0.05101, over 3194498.64 frames. ], batch size: 83, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:01:08,002 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 1.989e+02 2.229e+02 2.704e+02 6.823e+02, threshold=4.458e+02, percent-clipped=4.0 2023-04-30 04:02:16,326 INFO [train.py:904] (5/8) Epoch 15, batch 4950, loss[loss=0.1644, simple_loss=0.2636, pruned_loss=0.03259, over 16854.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2777, pruned_loss=0.05069, over 3183551.12 frames. ], batch size: 102, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:02:32,356 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8762, 5.2524, 5.4177, 5.2266, 5.2284, 5.7947, 5.2685, 4.9839], device='cuda:5'), covar=tensor([0.0813, 0.1495, 0.1584, 0.1569, 0.2001, 0.0718, 0.1352, 0.2106], device='cuda:5'), in_proj_covar=tensor([0.0372, 0.0523, 0.0569, 0.0447, 0.0592, 0.0600, 0.0454, 0.0595], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 04:03:26,066 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-30 04:03:28,738 INFO [train.py:904] (5/8) Epoch 15, batch 5000, loss[loss=0.2281, simple_loss=0.3097, pruned_loss=0.07321, over 12007.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2791, pruned_loss=0.05057, over 3198166.64 frames. ], batch size: 246, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:03:32,271 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.177e+02 2.680e+02 3.002e+02 5.827e+02, threshold=5.360e+02, percent-clipped=3.0 2023-04-30 04:04:39,317 INFO [train.py:904] (5/8) Epoch 15, batch 5050, loss[loss=0.1983, simple_loss=0.295, pruned_loss=0.05077, over 16448.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2799, pruned_loss=0.05024, over 3220903.06 frames. ], batch size: 146, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:04:57,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0883, 2.0887, 2.2088, 3.6502, 1.9802, 2.3924, 2.1722, 2.2435], device='cuda:5'), covar=tensor([0.1223, 0.3414, 0.2575, 0.0504, 0.4017, 0.2375, 0.3550, 0.3085], device='cuda:5'), in_proj_covar=tensor([0.0377, 0.0417, 0.0344, 0.0319, 0.0420, 0.0478, 0.0381, 0.0486], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:05:46,846 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:05:48,960 INFO [train.py:904] (5/8) Epoch 15, batch 5100, loss[loss=0.1696, simple_loss=0.2575, pruned_loss=0.04091, over 16692.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2785, pruned_loss=0.04953, over 3226190.58 frames. ], batch size: 57, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:50,738 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5828, 2.4981, 2.5334, 4.2768, 3.2917, 4.0329, 1.4583, 2.9138], device='cuda:5'), covar=tensor([0.1365, 0.0829, 0.1224, 0.0140, 0.0241, 0.0346, 0.1656, 0.0810], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0166, 0.0188, 0.0168, 0.0202, 0.0211, 0.0191, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 04:05:52,966 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.025e+02 2.374e+02 2.786e+02 3.985e+02, threshold=4.748e+02, percent-clipped=0.0 2023-04-30 04:06:45,539 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:53,118 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147247.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:59,929 INFO [train.py:904] (5/8) Epoch 15, batch 5150, loss[loss=0.1829, simple_loss=0.2736, pruned_loss=0.04608, over 16544.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2786, pruned_loss=0.0491, over 3209207.66 frames. ], batch size: 68, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:07:45,089 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:03,064 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:14,135 INFO [train.py:904] (5/8) Epoch 15, batch 5200, loss[loss=0.1879, simple_loss=0.2679, pruned_loss=0.05391, over 16900.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2772, pruned_loss=0.04876, over 3202855.99 frames. ], batch size: 116, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:08:14,612 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147302.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:17,788 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.189e+02 2.515e+02 3.060e+02 5.067e+02, threshold=5.031e+02, percent-clipped=2.0 2023-04-30 04:08:25,189 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147310.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:55,676 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147330.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:09:27,027 INFO [train.py:904] (5/8) Epoch 15, batch 5250, loss[loss=0.1859, simple_loss=0.2536, pruned_loss=0.05907, over 16479.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2748, pruned_loss=0.04852, over 3205813.35 frames. ], batch size: 35, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:09:55,033 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:10:06,907 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7389, 3.9072, 2.8615, 2.3626, 2.6130, 2.4695, 4.1029, 3.5213], device='cuda:5'), covar=tensor([0.2485, 0.0610, 0.1707, 0.2374, 0.2358, 0.1738, 0.0427, 0.1014], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0263, 0.0293, 0.0292, 0.0285, 0.0234, 0.0277, 0.0312], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 04:10:37,515 INFO [train.py:904] (5/8) Epoch 15, batch 5300, loss[loss=0.1399, simple_loss=0.2342, pruned_loss=0.02276, over 16876.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2706, pruned_loss=0.04701, over 3213511.20 frames. ], batch size: 96, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:10:40,970 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 1.977e+02 2.284e+02 2.754e+02 4.909e+02, threshold=4.569e+02, percent-clipped=0.0 2023-04-30 04:10:45,216 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147407.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:10:58,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6109, 4.5869, 4.3194, 2.9713, 3.8081, 4.4035, 3.8605, 2.4266], device='cuda:5'), covar=tensor([0.0466, 0.0018, 0.0030, 0.0311, 0.0070, 0.0060, 0.0083, 0.0418], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0073, 0.0073, 0.0130, 0.0088, 0.0097, 0.0085, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 04:11:23,772 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147434.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:34,907 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6361, 4.4175, 4.2313, 2.9096, 3.6683, 4.2879, 3.7676, 2.3846], device='cuda:5'), covar=tensor([0.0408, 0.0021, 0.0029, 0.0297, 0.0084, 0.0062, 0.0070, 0.0385], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0073, 0.0073, 0.0130, 0.0089, 0.0097, 0.0085, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 04:11:49,982 INFO [train.py:904] (5/8) Epoch 15, batch 5350, loss[loss=0.1645, simple_loss=0.2579, pruned_loss=0.03557, over 17128.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2692, pruned_loss=0.04653, over 3212329.98 frames. ], batch size: 47, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:12:14,846 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:12:34,644 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2936, 4.3374, 4.1354, 3.8562, 3.8487, 4.2562, 3.9711, 3.9883], device='cuda:5'), covar=tensor([0.0588, 0.0520, 0.0292, 0.0287, 0.0911, 0.0457, 0.0683, 0.0643], device='cuda:5'), in_proj_covar=tensor([0.0266, 0.0365, 0.0315, 0.0298, 0.0329, 0.0345, 0.0212, 0.0371], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:12:53,075 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:01,235 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147500.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:03,290 INFO [train.py:904] (5/8) Epoch 15, batch 5400, loss[loss=0.1706, simple_loss=0.2676, pruned_loss=0.03681, over 16519.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2721, pruned_loss=0.04752, over 3189872.12 frames. ], batch size: 75, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:13:07,674 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 2.104e+02 2.571e+02 3.291e+02 5.861e+02, threshold=5.143e+02, percent-clipped=4.0 2023-04-30 04:13:28,638 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8140, 3.8299, 2.2428, 4.6861, 2.9018, 4.4420, 2.3625, 2.9594], device='cuda:5'), covar=tensor([0.0251, 0.0358, 0.1757, 0.0105, 0.0859, 0.0443, 0.1596, 0.0845], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0190, 0.0142, 0.0169, 0.0210, 0.0198, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 04:14:05,015 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8358, 3.7447, 3.9180, 3.7367, 3.8306, 4.2551, 3.8885, 3.5599], device='cuda:5'), covar=tensor([0.1915, 0.2285, 0.1988, 0.2556, 0.2743, 0.1748, 0.1537, 0.2773], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0529, 0.0573, 0.0450, 0.0602, 0.0608, 0.0456, 0.0601], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 04:14:13,937 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:14:20,734 INFO [train.py:904] (5/8) Epoch 15, batch 5450, loss[loss=0.2126, simple_loss=0.2994, pruned_loss=0.06291, over 16419.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2745, pruned_loss=0.04868, over 3186444.22 frames. ], batch size: 146, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:15:32,478 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147597.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:15:40,047 INFO [train.py:904] (5/8) Epoch 15, batch 5500, loss[loss=0.2653, simple_loss=0.3326, pruned_loss=0.09897, over 11862.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2821, pruned_loss=0.05329, over 3144922.97 frames. ], batch size: 248, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:15:43,662 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-30 04:15:45,690 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.456e+02 3.016e+02 4.198e+02 6.055e+02, threshold=6.032e+02, percent-clipped=7.0 2023-04-30 04:16:40,290 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2159, 3.7612, 3.7057, 2.1994, 3.3973, 3.7769, 3.5599, 1.9517], device='cuda:5'), covar=tensor([0.0491, 0.0040, 0.0047, 0.0428, 0.0096, 0.0102, 0.0069, 0.0429], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0073, 0.0073, 0.0130, 0.0089, 0.0097, 0.0085, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 04:16:58,376 INFO [train.py:904] (5/8) Epoch 15, batch 5550, loss[loss=0.2988, simple_loss=0.3465, pruned_loss=0.1255, over 11244.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2901, pruned_loss=0.05914, over 3122760.32 frames. ], batch size: 247, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:17:10,670 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:22,927 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:38,986 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 04:17:46,896 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5006, 3.4809, 2.7711, 2.1190, 2.3854, 2.2355, 3.6623, 3.3460], device='cuda:5'), covar=tensor([0.2728, 0.0727, 0.1593, 0.2431, 0.2230, 0.1878, 0.0460, 0.1035], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0261, 0.0292, 0.0293, 0.0284, 0.0234, 0.0276, 0.0311], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:18:21,668 INFO [train.py:904] (5/8) Epoch 15, batch 5600, loss[loss=0.3068, simple_loss=0.3526, pruned_loss=0.1305, over 11070.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2953, pruned_loss=0.06388, over 3088303.37 frames. ], batch size: 248, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:18:28,274 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.312e+02 3.387e+02 4.165e+02 5.212e+02 1.017e+03, threshold=8.330e+02, percent-clipped=15.0 2023-04-30 04:18:35,817 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4879, 3.5276, 3.2994, 2.9682, 3.1739, 3.4505, 3.2702, 3.2773], device='cuda:5'), covar=tensor([0.0551, 0.0569, 0.0236, 0.0267, 0.0492, 0.0431, 0.1232, 0.0453], device='cuda:5'), in_proj_covar=tensor([0.0262, 0.0363, 0.0311, 0.0294, 0.0325, 0.0341, 0.0208, 0.0368], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:18:53,078 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:19:46,184 INFO [train.py:904] (5/8) Epoch 15, batch 5650, loss[loss=0.2915, simple_loss=0.3525, pruned_loss=0.1153, over 11329.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3013, pruned_loss=0.06876, over 3046124.88 frames. ], batch size: 250, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:20:04,646 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147763.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:20:13,268 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4675, 3.4488, 3.4122, 2.6386, 3.3146, 2.0912, 3.1394, 2.8258], device='cuda:5'), covar=tensor([0.0142, 0.0108, 0.0175, 0.0214, 0.0091, 0.2099, 0.0127, 0.0212], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0132, 0.0180, 0.0168, 0.0152, 0.0190, 0.0168, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:20:47,701 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:21:05,470 INFO [train.py:904] (5/8) Epoch 15, batch 5700, loss[loss=0.2009, simple_loss=0.3001, pruned_loss=0.05086, over 16754.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3034, pruned_loss=0.07061, over 3031370.37 frames. ], batch size: 83, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:21:11,570 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.489e+02 3.643e+02 4.297e+02 4.973e+02 1.168e+03, threshold=8.593e+02, percent-clipped=2.0 2023-04-30 04:21:26,335 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:22:00,146 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3545, 2.9297, 2.6942, 2.2423, 2.3264, 2.2627, 2.8841, 2.8881], device='cuda:5'), covar=tensor([0.2172, 0.0712, 0.1356, 0.2140, 0.2009, 0.1875, 0.0452, 0.1001], device='cuda:5'), in_proj_covar=tensor([0.0309, 0.0260, 0.0290, 0.0291, 0.0282, 0.0232, 0.0275, 0.0309], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:22:23,096 INFO [train.py:904] (5/8) Epoch 15, batch 5750, loss[loss=0.2019, simple_loss=0.2933, pruned_loss=0.05526, over 16785.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3064, pruned_loss=0.07249, over 3001225.10 frames. ], batch size: 83, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:22:32,849 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:03,218 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:35,071 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 04:23:36,819 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147897.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:44,674 INFO [train.py:904] (5/8) Epoch 15, batch 5800, loss[loss=0.2601, simple_loss=0.3199, pruned_loss=0.1002, over 12132.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3058, pruned_loss=0.07133, over 2987937.36 frames. ], batch size: 246, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:23:51,420 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.104e+02 2.959e+02 3.306e+02 4.369e+02 1.266e+03, threshold=6.612e+02, percent-clipped=1.0 2023-04-30 04:24:00,827 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:12,299 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147918.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:55,226 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:05,483 INFO [train.py:904] (5/8) Epoch 15, batch 5850, loss[loss=0.1855, simple_loss=0.2749, pruned_loss=0.04802, over 17132.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3027, pruned_loss=0.06922, over 2999914.41 frames. ], batch size: 47, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:25:29,110 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147966.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:38,373 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:54,829 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0116, 1.9364, 2.4870, 2.9916, 2.8870, 3.4380, 2.1238, 3.4461], device='cuda:5'), covar=tensor([0.0175, 0.0444, 0.0296, 0.0256, 0.0230, 0.0114, 0.0442, 0.0095], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0180, 0.0165, 0.0170, 0.0178, 0.0136, 0.0183, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:26:30,242 INFO [train.py:904] (5/8) Epoch 15, batch 5900, loss[loss=0.1887, simple_loss=0.2711, pruned_loss=0.05313, over 16614.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3023, pruned_loss=0.06851, over 3026451.56 frames. ], batch size: 57, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:26:39,378 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.736e+02 3.208e+02 4.005e+02 8.372e+02, threshold=6.416e+02, percent-clipped=2.0 2023-04-30 04:26:52,639 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148014.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:26:54,081 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:27:00,680 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 04:27:50,084 INFO [train.py:904] (5/8) Epoch 15, batch 5950, loss[loss=0.2001, simple_loss=0.2854, pruned_loss=0.0574, over 16601.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.3025, pruned_loss=0.06712, over 3049743.31 frames. ], batch size: 62, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:28:09,485 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:28:49,050 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:29:08,856 INFO [train.py:904] (5/8) Epoch 15, batch 6000, loss[loss=0.2103, simple_loss=0.2911, pruned_loss=0.06474, over 17045.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.3016, pruned_loss=0.06682, over 3050580.55 frames. ], batch size: 55, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:29:08,857 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 04:29:19,436 INFO [train.py:938] (5/8) Epoch 15, validation: loss=0.1559, simple_loss=0.2691, pruned_loss=0.0214, over 944034.00 frames. 2023-04-30 04:29:19,437 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 04:29:26,131 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.753e+02 3.314e+02 4.296e+02 7.936e+02, threshold=6.628e+02, percent-clipped=2.0 2023-04-30 04:29:33,717 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:30:15,356 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148138.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:30:36,864 INFO [train.py:904] (5/8) Epoch 15, batch 6050, loss[loss=0.1961, simple_loss=0.2972, pruned_loss=0.04753, over 16707.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.3003, pruned_loss=0.0661, over 3059580.01 frames. ], batch size: 83, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:30:39,623 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6154, 3.6024, 3.9341, 1.9162, 4.1287, 4.1503, 3.1762, 3.0728], device='cuda:5'), covar=tensor([0.0728, 0.0196, 0.0149, 0.1156, 0.0052, 0.0117, 0.0323, 0.0416], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0104, 0.0090, 0.0136, 0.0072, 0.0114, 0.0122, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 04:31:08,634 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:31:59,321 INFO [train.py:904] (5/8) Epoch 15, batch 6100, loss[loss=0.2058, simple_loss=0.2808, pruned_loss=0.0654, over 16990.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2992, pruned_loss=0.06431, over 3089501.94 frames. ], batch size: 55, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:32:08,601 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.651e+02 3.170e+02 3.946e+02 8.387e+02, threshold=6.339e+02, percent-clipped=2.0 2023-04-30 04:32:18,627 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:33:18,253 INFO [train.py:904] (5/8) Epoch 15, batch 6150, loss[loss=0.2083, simple_loss=0.279, pruned_loss=0.06883, over 11643.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2972, pruned_loss=0.0636, over 3103399.43 frames. ], batch size: 246, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:33:29,748 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3823, 4.6580, 4.4340, 4.4487, 4.1688, 4.1517, 4.1880, 4.6968], device='cuda:5'), covar=tensor([0.1098, 0.0851, 0.0991, 0.0793, 0.0781, 0.1456, 0.0980, 0.0897], device='cuda:5'), in_proj_covar=tensor([0.0598, 0.0745, 0.0615, 0.0535, 0.0467, 0.0479, 0.0615, 0.0563], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:33:43,180 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:15,526 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:37,597 INFO [train.py:904] (5/8) Epoch 15, batch 6200, loss[loss=0.1952, simple_loss=0.2817, pruned_loss=0.05431, over 16778.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2951, pruned_loss=0.06293, over 3121673.81 frames. ], batch size: 83, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:34:46,175 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.083e+02 3.031e+02 3.685e+02 4.615e+02 1.155e+03, threshold=7.370e+02, percent-clipped=8.0 2023-04-30 04:34:58,575 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148315.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:33,787 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3972, 3.3459, 3.4145, 3.5205, 3.5426, 3.2891, 3.5257, 3.6029], device='cuda:5'), covar=tensor([0.1140, 0.0951, 0.1038, 0.0602, 0.0673, 0.2339, 0.0941, 0.0693], device='cuda:5'), in_proj_covar=tensor([0.0570, 0.0705, 0.0848, 0.0714, 0.0539, 0.0565, 0.0572, 0.0668], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:35:48,366 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148348.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:54,859 INFO [train.py:904] (5/8) Epoch 15, batch 6250, loss[loss=0.246, simple_loss=0.3129, pruned_loss=0.08952, over 11671.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2949, pruned_loss=0.06287, over 3123696.31 frames. ], batch size: 248, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:36:11,840 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148363.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:36:41,616 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 04:36:55,229 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 04:37:11,941 INFO [train.py:904] (5/8) Epoch 15, batch 6300, loss[loss=0.2348, simple_loss=0.3141, pruned_loss=0.07779, over 16476.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.295, pruned_loss=0.06258, over 3111658.99 frames. ], batch size: 146, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:37:21,869 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.809e+02 3.245e+02 4.323e+02 1.479e+03, threshold=6.491e+02, percent-clipped=2.0 2023-04-30 04:37:22,295 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8549, 4.8881, 5.3005, 5.2253, 5.2783, 4.9159, 4.8621, 4.5766], device='cuda:5'), covar=tensor([0.0274, 0.0491, 0.0377, 0.0443, 0.0424, 0.0364, 0.0957, 0.0488], device='cuda:5'), in_proj_covar=tensor([0.0365, 0.0391, 0.0385, 0.0368, 0.0437, 0.0408, 0.0504, 0.0329], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 04:37:57,330 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148429.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:38:33,534 INFO [train.py:904] (5/8) Epoch 15, batch 6350, loss[loss=0.1971, simple_loss=0.2829, pruned_loss=0.05568, over 16766.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2958, pruned_loss=0.06346, over 3120338.12 frames. ], batch size: 102, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:39:03,921 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:34,181 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:45,777 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:51,472 INFO [train.py:904] (5/8) Epoch 15, batch 6400, loss[loss=0.1913, simple_loss=0.2692, pruned_loss=0.05666, over 16331.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.295, pruned_loss=0.06368, over 3140459.69 frames. ], batch size: 35, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:39:59,986 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.110e+02 3.263e+02 3.782e+02 4.494e+02 8.205e+02, threshold=7.565e+02, percent-clipped=3.0 2023-04-30 04:40:09,519 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:40:18,144 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:40:51,693 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0874, 3.5117, 3.4301, 1.8253, 2.8505, 2.2878, 3.4831, 3.7958], device='cuda:5'), covar=tensor([0.0282, 0.0715, 0.0587, 0.2068, 0.0848, 0.0977, 0.0577, 0.0720], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0154, 0.0163, 0.0147, 0.0139, 0.0127, 0.0140, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 04:41:02,646 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6252, 2.0316, 1.5911, 1.8604, 2.4560, 2.0089, 2.3795, 2.6341], device='cuda:5'), covar=tensor([0.0174, 0.0398, 0.0579, 0.0466, 0.0247, 0.0355, 0.0288, 0.0222], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0216, 0.0210, 0.0211, 0.0217, 0.0216, 0.0218, 0.0211], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:41:07,998 INFO [train.py:904] (5/8) Epoch 15, batch 6450, loss[loss=0.1825, simple_loss=0.2766, pruned_loss=0.04421, over 16533.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2953, pruned_loss=0.06331, over 3133223.52 frames. ], batch size: 68, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:41:19,854 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:22,167 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:26,581 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-30 04:41:32,443 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:42:10,691 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0615, 2.3170, 2.2975, 2.6763, 1.7833, 3.1443, 1.8005, 2.5866], device='cuda:5'), covar=tensor([0.1091, 0.0633, 0.1094, 0.0151, 0.0130, 0.0405, 0.1393, 0.0753], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0167, 0.0201, 0.0210, 0.0190, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 04:42:27,015 INFO [train.py:904] (5/8) Epoch 15, batch 6500, loss[loss=0.2464, simple_loss=0.3105, pruned_loss=0.09111, over 11178.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2935, pruned_loss=0.06276, over 3132016.39 frames. ], batch size: 248, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:42:37,000 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.869e+02 3.331e+02 4.011e+02 8.248e+02, threshold=6.661e+02, percent-clipped=1.0 2023-04-30 04:42:41,918 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7202, 2.5259, 2.3071, 3.4963, 2.3434, 3.7328, 1.3765, 2.7047], device='cuda:5'), covar=tensor([0.1301, 0.0709, 0.1223, 0.0162, 0.0190, 0.0372, 0.1625, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0165, 0.0186, 0.0167, 0.0201, 0.0210, 0.0190, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 04:42:47,764 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:33,201 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:47,222 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148651.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:47,993 INFO [train.py:904] (5/8) Epoch 15, batch 6550, loss[loss=0.2152, simple_loss=0.3092, pruned_loss=0.06064, over 16703.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2963, pruned_loss=0.0637, over 3128160.80 frames. ], batch size: 134, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:45:05,108 INFO [train.py:904] (5/8) Epoch 15, batch 6600, loss[loss=0.2327, simple_loss=0.3142, pruned_loss=0.07556, over 16521.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2989, pruned_loss=0.06413, over 3132500.12 frames. ], batch size: 68, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:45:13,947 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.578e+02 3.041e+02 3.810e+02 4.934e+02 1.256e+03, threshold=7.620e+02, percent-clipped=11.0 2023-04-30 04:45:19,931 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:45:26,334 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:45:33,671 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 04:46:14,063 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 04:46:22,125 INFO [train.py:904] (5/8) Epoch 15, batch 6650, loss[loss=0.2138, simple_loss=0.2953, pruned_loss=0.06618, over 16359.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2985, pruned_loss=0.06452, over 3122038.76 frames. ], batch size: 146, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:47:00,949 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:47:13,532 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:47:39,126 INFO [train.py:904] (5/8) Epoch 15, batch 6700, loss[loss=0.2345, simple_loss=0.3109, pruned_loss=0.07903, over 16398.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2976, pruned_loss=0.06512, over 3103184.58 frames. ], batch size: 146, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:47:47,146 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.803e+02 3.506e+02 4.354e+02 9.571e+02, threshold=7.012e+02, percent-clipped=2.0 2023-04-30 04:48:54,941 INFO [train.py:904] (5/8) Epoch 15, batch 6750, loss[loss=0.1914, simple_loss=0.2766, pruned_loss=0.05311, over 16644.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2961, pruned_loss=0.06478, over 3107611.59 frames. ], batch size: 134, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:48:58,423 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:48:58,534 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8269, 4.5811, 4.8154, 5.0270, 5.2036, 4.6617, 5.1384, 5.1155], device='cuda:5'), covar=tensor([0.1536, 0.1221, 0.1584, 0.0669, 0.0524, 0.0825, 0.0598, 0.0654], device='cuda:5'), in_proj_covar=tensor([0.0564, 0.0699, 0.0838, 0.0708, 0.0533, 0.0560, 0.0568, 0.0661], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:49:11,557 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9712, 3.2572, 3.1138, 2.1183, 2.9381, 3.1879, 3.0417, 1.8614], device='cuda:5'), covar=tensor([0.0492, 0.0043, 0.0061, 0.0402, 0.0102, 0.0121, 0.0082, 0.0432], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0075, 0.0074, 0.0133, 0.0089, 0.0099, 0.0086, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 04:50:09,787 INFO [train.py:904] (5/8) Epoch 15, batch 6800, loss[loss=0.2046, simple_loss=0.2991, pruned_loss=0.05499, over 16888.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2968, pruned_loss=0.0651, over 3114345.52 frames. ], batch size: 96, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:50:10,490 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8775, 2.0962, 2.3654, 3.1727, 2.1509, 2.2961, 2.2932, 2.1639], device='cuda:5'), covar=tensor([0.1087, 0.3011, 0.2175, 0.0572, 0.3652, 0.2154, 0.2727, 0.3077], device='cuda:5'), in_proj_covar=tensor([0.0374, 0.0412, 0.0343, 0.0317, 0.0421, 0.0474, 0.0379, 0.0482], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:50:21,236 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.837e+02 3.542e+02 4.429e+02 7.153e+02, threshold=7.083e+02, percent-clipped=1.0 2023-04-30 04:50:39,669 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9669, 2.8897, 3.1560, 1.7245, 3.2558, 3.3181, 2.5904, 2.5151], device='cuda:5'), covar=tensor([0.0923, 0.0248, 0.0187, 0.1193, 0.0088, 0.0171, 0.0463, 0.0490], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0105, 0.0091, 0.0138, 0.0073, 0.0115, 0.0123, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 04:50:41,093 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6452, 4.8883, 5.0508, 4.7869, 4.8740, 5.4146, 4.8971, 4.7402], device='cuda:5'), covar=tensor([0.1194, 0.1729, 0.2061, 0.1970, 0.2476, 0.0955, 0.1700, 0.2298], device='cuda:5'), in_proj_covar=tensor([0.0381, 0.0532, 0.0582, 0.0454, 0.0601, 0.0611, 0.0464, 0.0606], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 04:51:15,911 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148943.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:51:28,466 INFO [train.py:904] (5/8) Epoch 15, batch 6850, loss[loss=0.2159, simple_loss=0.313, pruned_loss=0.05939, over 16782.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2972, pruned_loss=0.06478, over 3125208.30 frames. ], batch size: 124, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:52:00,656 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3638, 5.4705, 5.1968, 4.8067, 4.6060, 5.3272, 5.2495, 4.8873], device='cuda:5'), covar=tensor([0.1149, 0.1196, 0.0448, 0.0456, 0.1475, 0.0937, 0.0559, 0.1228], device='cuda:5'), in_proj_covar=tensor([0.0260, 0.0360, 0.0308, 0.0291, 0.0323, 0.0338, 0.0209, 0.0366], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:52:26,495 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148991.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:43,276 INFO [train.py:904] (5/8) Epoch 15, batch 6900, loss[loss=0.2997, simple_loss=0.3675, pruned_loss=0.116, over 11716.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2996, pruned_loss=0.06488, over 3117073.15 frames. ], batch size: 247, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:52:52,145 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149007.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:54,701 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.687e+02 3.079e+02 3.976e+02 7.116e+02, threshold=6.157e+02, percent-clipped=1.0 2023-04-30 04:53:58,323 INFO [train.py:904] (5/8) Epoch 15, batch 6950, loss[loss=0.2848, simple_loss=0.3514, pruned_loss=0.1091, over 11217.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.3014, pruned_loss=0.06646, over 3102792.51 frames. ], batch size: 248, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:54:02,659 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4090, 2.8302, 2.6583, 2.2848, 2.2881, 2.2994, 2.9177, 2.8617], device='cuda:5'), covar=tensor([0.2278, 0.1003, 0.1401, 0.2100, 0.1917, 0.1811, 0.0528, 0.1152], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0262, 0.0291, 0.0292, 0.0284, 0.0235, 0.0276, 0.0311], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:54:27,430 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149072.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:54:45,980 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:54:54,893 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.52 vs. limit=5.0 2023-04-30 04:55:10,884 INFO [train.py:904] (5/8) Epoch 15, batch 7000, loss[loss=0.2053, simple_loss=0.303, pruned_loss=0.0538, over 17012.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.3013, pruned_loss=0.06591, over 3090401.55 frames. ], batch size: 55, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:55:23,350 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.893e+02 3.767e+02 4.998e+02 1.151e+03, threshold=7.533e+02, percent-clipped=10.0 2023-04-30 04:55:58,209 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149133.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:56:25,570 INFO [train.py:904] (5/8) Epoch 15, batch 7050, loss[loss=0.2558, simple_loss=0.3206, pruned_loss=0.09556, over 11157.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.3015, pruned_loss=0.06552, over 3099950.99 frames. ], batch size: 246, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:56:29,212 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:40,346 INFO [train.py:904] (5/8) Epoch 15, batch 7100, loss[loss=0.2249, simple_loss=0.3052, pruned_loss=0.07228, over 15314.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.3005, pruned_loss=0.06524, over 3094844.79 frames. ], batch size: 191, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:57:40,786 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:53,982 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.990e+02 3.583e+02 4.304e+02 1.223e+03, threshold=7.166e+02, percent-clipped=1.0 2023-04-30 04:58:55,149 INFO [train.py:904] (5/8) Epoch 15, batch 7150, loss[loss=0.2767, simple_loss=0.3308, pruned_loss=0.1113, over 10971.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2991, pruned_loss=0.06535, over 3095167.55 frames. ], batch size: 246, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:59:34,128 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-04-30 04:59:51,777 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1398, 5.4489, 5.1979, 5.2229, 4.8814, 4.8908, 4.8978, 5.5452], device='cuda:5'), covar=tensor([0.1055, 0.0734, 0.0970, 0.0762, 0.0817, 0.0781, 0.1009, 0.0755], device='cuda:5'), in_proj_covar=tensor([0.0592, 0.0729, 0.0608, 0.0528, 0.0458, 0.0474, 0.0608, 0.0552], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 04:59:52,341 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 05:00:05,795 INFO [train.py:904] (5/8) Epoch 15, batch 7200, loss[loss=0.1849, simple_loss=0.2793, pruned_loss=0.04527, over 16957.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2963, pruned_loss=0.0633, over 3092100.98 frames. ], batch size: 109, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:00:13,315 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:00:17,917 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 2.693e+02 3.090e+02 3.791e+02 7.265e+02, threshold=6.181e+02, percent-clipped=1.0 2023-04-30 05:00:51,701 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-30 05:00:58,515 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-30 05:01:26,145 INFO [train.py:904] (5/8) Epoch 15, batch 7250, loss[loss=0.1956, simple_loss=0.2829, pruned_loss=0.05414, over 16742.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2934, pruned_loss=0.06179, over 3089374.13 frames. ], batch size: 89, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:01:31,289 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149355.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:01:57,348 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:02:42,302 INFO [train.py:904] (5/8) Epoch 15, batch 7300, loss[loss=0.2093, simple_loss=0.2969, pruned_loss=0.0608, over 16548.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2927, pruned_loss=0.06158, over 3100052.36 frames. ], batch size: 62, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:02:55,637 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 2.888e+02 3.399e+02 4.220e+02 6.338e+02, threshold=6.799e+02, percent-clipped=2.0 2023-04-30 05:03:10,461 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:03:42,259 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-30 05:03:58,898 INFO [train.py:904] (5/8) Epoch 15, batch 7350, loss[loss=0.2014, simple_loss=0.2887, pruned_loss=0.05699, over 16712.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2933, pruned_loss=0.06252, over 3071785.73 frames. ], batch size: 89, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:04:08,815 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 05:04:25,902 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 05:05:17,916 INFO [train.py:904] (5/8) Epoch 15, batch 7400, loss[loss=0.2133, simple_loss=0.3024, pruned_loss=0.06209, over 16503.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2947, pruned_loss=0.06303, over 3080379.18 frames. ], batch size: 68, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:05:32,177 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.749e+02 3.166e+02 4.051e+02 9.173e+02, threshold=6.332e+02, percent-clipped=4.0 2023-04-30 05:05:37,308 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:06:37,036 INFO [train.py:904] (5/8) Epoch 15, batch 7450, loss[loss=0.1936, simple_loss=0.2941, pruned_loss=0.04652, over 16806.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2967, pruned_loss=0.06511, over 3054445.29 frames. ], batch size: 83, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:07:00,471 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149565.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:07:16,479 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:08:00,167 INFO [train.py:904] (5/8) Epoch 15, batch 7500, loss[loss=0.2311, simple_loss=0.3018, pruned_loss=0.08022, over 11520.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2973, pruned_loss=0.06464, over 3040983.47 frames. ], batch size: 248, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:08:16,083 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.987e+02 3.634e+02 4.459e+02 1.121e+03, threshold=7.267e+02, percent-clipped=5.0 2023-04-30 05:08:40,017 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149626.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:09:18,815 INFO [train.py:904] (5/8) Epoch 15, batch 7550, loss[loss=0.2049, simple_loss=0.281, pruned_loss=0.06441, over 15515.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2966, pruned_loss=0.06524, over 3016933.76 frames. ], batch size: 190, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:09:53,195 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-04-30 05:10:35,818 INFO [train.py:904] (5/8) Epoch 15, batch 7600, loss[loss=0.197, simple_loss=0.2992, pruned_loss=0.04743, over 16896.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2953, pruned_loss=0.06442, over 3053476.10 frames. ], batch size: 96, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:10:50,669 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.884e+02 3.508e+02 4.101e+02 6.446e+02, threshold=7.017e+02, percent-clipped=0.0 2023-04-30 05:11:00,764 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8330, 2.6529, 2.2549, 3.5013, 2.4214, 3.6537, 1.5475, 2.6457], device='cuda:5'), covar=tensor([0.1204, 0.0597, 0.1238, 0.0168, 0.0153, 0.0473, 0.1513, 0.0876], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0165, 0.0186, 0.0166, 0.0201, 0.0210, 0.0191, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 05:11:45,931 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149746.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:11:55,070 INFO [train.py:904] (5/8) Epoch 15, batch 7650, loss[loss=0.2236, simple_loss=0.3083, pruned_loss=0.0695, over 16689.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2961, pruned_loss=0.06556, over 3031528.25 frames. ], batch size: 134, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:11:59,617 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-30 05:12:46,475 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5876, 3.1314, 2.8666, 1.8028, 2.5975, 2.0083, 3.0357, 3.3435], device='cuda:5'), covar=tensor([0.0258, 0.0654, 0.0698, 0.2183, 0.0982, 0.1136, 0.0662, 0.0740], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0153, 0.0162, 0.0147, 0.0139, 0.0126, 0.0140, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 05:13:13,701 INFO [train.py:904] (5/8) Epoch 15, batch 7700, loss[loss=0.1999, simple_loss=0.2914, pruned_loss=0.0542, over 16216.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2961, pruned_loss=0.06605, over 3034884.60 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:13:22,702 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149807.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:13:29,264 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 3.235e+02 4.085e+02 5.006e+02 1.161e+03, threshold=8.169e+02, percent-clipped=3.0 2023-04-30 05:14:26,384 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4680, 3.5103, 3.2508, 3.0037, 3.1158, 3.3833, 3.3097, 3.2176], device='cuda:5'), covar=tensor([0.0628, 0.0571, 0.0250, 0.0262, 0.0515, 0.0463, 0.1070, 0.0495], device='cuda:5'), in_proj_covar=tensor([0.0257, 0.0355, 0.0303, 0.0288, 0.0320, 0.0334, 0.0208, 0.0362], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 05:14:33,198 INFO [train.py:904] (5/8) Epoch 15, batch 7750, loss[loss=0.276, simple_loss=0.3385, pruned_loss=0.1067, over 11514.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2963, pruned_loss=0.06614, over 3032894.60 frames. ], batch size: 248, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:15:02,375 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:15:07,438 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6883, 4.6679, 4.5349, 3.7924, 4.5808, 1.5976, 4.3418, 4.2747], device='cuda:5'), covar=tensor([0.0104, 0.0089, 0.0176, 0.0410, 0.0095, 0.2808, 0.0141, 0.0248], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0131, 0.0178, 0.0166, 0.0150, 0.0190, 0.0166, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 05:15:51,928 INFO [train.py:904] (5/8) Epoch 15, batch 7800, loss[loss=0.2603, simple_loss=0.3196, pruned_loss=0.1005, over 11520.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2975, pruned_loss=0.0667, over 3045976.06 frames. ], batch size: 248, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:16:07,342 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 3.237e+02 3.890e+02 4.639e+02 7.802e+02, threshold=7.781e+02, percent-clipped=0.0 2023-04-30 05:16:21,593 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149921.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:17:06,643 INFO [train.py:904] (5/8) Epoch 15, batch 7850, loss[loss=0.1968, simple_loss=0.2916, pruned_loss=0.05098, over 16765.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2978, pruned_loss=0.066, over 3048505.27 frames. ], batch size: 89, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:18:25,142 INFO [train.py:904] (5/8) Epoch 15, batch 7900, loss[loss=0.1996, simple_loss=0.2959, pruned_loss=0.05169, over 16848.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2969, pruned_loss=0.06509, over 3063393.30 frames. ], batch size: 102, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:18:40,306 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.219e+02 2.824e+02 3.475e+02 4.207e+02 7.504e+02, threshold=6.949e+02, percent-clipped=0.0 2023-04-30 05:19:43,698 INFO [train.py:904] (5/8) Epoch 15, batch 7950, loss[loss=0.2404, simple_loss=0.306, pruned_loss=0.08743, over 11551.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2973, pruned_loss=0.06569, over 3065350.98 frames. ], batch size: 246, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:19:59,504 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5235, 3.4910, 3.4701, 2.8093, 3.3944, 2.1475, 3.1037, 2.8630], device='cuda:5'), covar=tensor([0.0138, 0.0110, 0.0158, 0.0200, 0.0090, 0.1995, 0.0129, 0.0185], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0131, 0.0179, 0.0165, 0.0151, 0.0191, 0.0166, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 05:20:07,392 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4567, 4.0858, 4.0902, 2.8687, 3.7033, 4.1499, 3.7439, 2.3502], device='cuda:5'), covar=tensor([0.0453, 0.0048, 0.0037, 0.0309, 0.0078, 0.0077, 0.0066, 0.0377], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0074, 0.0073, 0.0132, 0.0088, 0.0098, 0.0085, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 05:20:25,874 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:20:44,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4000, 3.3117, 2.6755, 2.0854, 2.2145, 2.1039, 3.5084, 3.1127], device='cuda:5'), covar=tensor([0.2950, 0.0754, 0.1731, 0.2703, 0.2659, 0.2135, 0.0492, 0.1229], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0262, 0.0294, 0.0294, 0.0286, 0.0236, 0.0278, 0.0312], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 05:20:58,690 INFO [train.py:904] (5/8) Epoch 15, batch 8000, loss[loss=0.2122, simple_loss=0.3005, pruned_loss=0.06191, over 16837.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2983, pruned_loss=0.06656, over 3049810.41 frames. ], batch size: 116, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:20:59,783 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150102.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:21:14,670 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.932e+02 3.545e+02 4.036e+02 7.060e+02, threshold=7.089e+02, percent-clipped=1.0 2023-04-30 05:21:56,014 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150140.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:22:14,253 INFO [train.py:904] (5/8) Epoch 15, batch 8050, loss[loss=0.2152, simple_loss=0.3014, pruned_loss=0.06447, over 16221.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2979, pruned_loss=0.06598, over 3061088.93 frames. ], batch size: 165, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:22:41,359 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:22:42,881 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 05:23:30,584 INFO [train.py:904] (5/8) Epoch 15, batch 8100, loss[loss=0.1954, simple_loss=0.2822, pruned_loss=0.0543, over 16993.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2971, pruned_loss=0.06524, over 3061478.15 frames. ], batch size: 55, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:23:31,127 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0129, 1.9365, 2.4583, 2.9321, 2.7445, 3.3605, 2.0595, 3.3809], device='cuda:5'), covar=tensor([0.0159, 0.0446, 0.0296, 0.0234, 0.0285, 0.0137, 0.0444, 0.0103], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0181, 0.0165, 0.0170, 0.0179, 0.0137, 0.0183, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 05:23:39,271 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-04-30 05:23:45,514 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.677e+02 3.415e+02 4.132e+02 1.188e+03, threshold=6.830e+02, percent-clipped=3.0 2023-04-30 05:23:56,038 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:23:59,916 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:24:45,704 INFO [train.py:904] (5/8) Epoch 15, batch 8150, loss[loss=0.248, simple_loss=0.3101, pruned_loss=0.09297, over 11283.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2944, pruned_loss=0.06374, over 3066268.07 frames. ], batch size: 247, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:25:11,110 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150269.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:19,671 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:57,617 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-04-30 05:26:00,385 INFO [train.py:904] (5/8) Epoch 15, batch 8200, loss[loss=0.1934, simple_loss=0.2771, pruned_loss=0.05486, over 16605.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2926, pruned_loss=0.06348, over 3065789.58 frames. ], batch size: 62, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:26:16,346 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 2.882e+02 3.369e+02 3.879e+02 7.748e+02, threshold=6.737e+02, percent-clipped=3.0 2023-04-30 05:26:54,615 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:27:22,320 INFO [train.py:904] (5/8) Epoch 15, batch 8250, loss[loss=0.1818, simple_loss=0.265, pruned_loss=0.04932, over 12059.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2915, pruned_loss=0.06124, over 3055732.10 frames. ], batch size: 248, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:27:39,320 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-30 05:28:22,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4400, 4.4995, 4.6458, 4.5251, 4.4523, 5.0370, 4.5291, 4.2331], device='cuda:5'), covar=tensor([0.1341, 0.1752, 0.1795, 0.1887, 0.2547, 0.1004, 0.1559, 0.2491], device='cuda:5'), in_proj_covar=tensor([0.0382, 0.0536, 0.0589, 0.0454, 0.0605, 0.0618, 0.0467, 0.0607], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 05:28:45,009 INFO [train.py:904] (5/8) Epoch 15, batch 8300, loss[loss=0.1981, simple_loss=0.2903, pruned_loss=0.05296, over 16383.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2887, pruned_loss=0.05829, over 3052204.81 frames. ], batch size: 165, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:28:45,642 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150402.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:29:01,272 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.394e+02 2.913e+02 3.586e+02 6.520e+02, threshold=5.826e+02, percent-clipped=0.0 2023-04-30 05:29:39,980 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150435.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:30:05,307 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150450.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:30:07,759 INFO [train.py:904] (5/8) Epoch 15, batch 8350, loss[loss=0.202, simple_loss=0.3027, pruned_loss=0.05064, over 15286.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2878, pruned_loss=0.05666, over 3034073.34 frames. ], batch size: 191, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:30:47,646 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3095, 1.7049, 2.0205, 2.3157, 2.4560, 2.5169, 1.8237, 2.5694], device='cuda:5'), covar=tensor([0.0189, 0.0436, 0.0295, 0.0283, 0.0240, 0.0200, 0.0426, 0.0125], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0179, 0.0163, 0.0168, 0.0177, 0.0135, 0.0179, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 05:31:29,421 INFO [train.py:904] (5/8) Epoch 15, batch 8400, loss[loss=0.1794, simple_loss=0.2774, pruned_loss=0.04067, over 16408.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2853, pruned_loss=0.0544, over 3030656.42 frames. ], batch size: 146, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:31:37,446 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 05:31:46,293 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.409e+02 2.912e+02 3.516e+02 5.302e+02, threshold=5.823e+02, percent-clipped=0.0 2023-04-30 05:32:24,188 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6483, 4.4679, 4.7115, 4.8284, 5.0103, 4.5092, 4.9770, 5.0238], device='cuda:5'), covar=tensor([0.1669, 0.1190, 0.1442, 0.0737, 0.0496, 0.0828, 0.0486, 0.0517], device='cuda:5'), in_proj_covar=tensor([0.0546, 0.0680, 0.0806, 0.0687, 0.0523, 0.0543, 0.0553, 0.0644], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 05:32:27,414 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150537.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:32:49,416 INFO [train.py:904] (5/8) Epoch 15, batch 8450, loss[loss=0.1666, simple_loss=0.2733, pruned_loss=0.02999, over 16698.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2834, pruned_loss=0.05258, over 3048569.65 frames. ], batch size: 83, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:33:43,503 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 05:34:02,268 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:34:08,634 INFO [train.py:904] (5/8) Epoch 15, batch 8500, loss[loss=0.2031, simple_loss=0.2876, pruned_loss=0.05928, over 16238.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2799, pruned_loss=0.05051, over 3033600.53 frames. ], batch size: 165, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:34:25,008 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 2.200e+02 2.703e+02 3.335e+02 6.908e+02, threshold=5.407e+02, percent-clipped=1.0 2023-04-30 05:34:27,558 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6220, 3.0178, 3.1885, 1.9640, 2.7673, 2.1571, 3.1721, 3.2088], device='cuda:5'), covar=tensor([0.0324, 0.0858, 0.0536, 0.2073, 0.0823, 0.1069, 0.0669, 0.1004], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0149, 0.0157, 0.0144, 0.0136, 0.0124, 0.0137, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 05:34:54,287 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150630.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:35:30,248 INFO [train.py:904] (5/8) Epoch 15, batch 8550, loss[loss=0.1802, simple_loss=0.2773, pruned_loss=0.04155, over 16628.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2774, pruned_loss=0.04926, over 3014538.45 frames. ], batch size: 69, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:37:12,300 INFO [train.py:904] (5/8) Epoch 15, batch 8600, loss[loss=0.1801, simple_loss=0.2818, pruned_loss=0.03917, over 16182.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2776, pruned_loss=0.04794, over 3032020.71 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:37:32,423 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.521e+02 3.034e+02 3.543e+02 4.971e+02, threshold=6.069e+02, percent-clipped=0.0 2023-04-30 05:37:53,583 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-30 05:38:14,991 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:20,244 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:51,970 INFO [train.py:904] (5/8) Epoch 15, batch 8650, loss[loss=0.1853, simple_loss=0.2779, pruned_loss=0.04636, over 16812.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2756, pruned_loss=0.04668, over 3028849.95 frames. ], batch size: 124, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:39:27,482 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6523, 3.1580, 3.4399, 1.9950, 2.9517, 2.1707, 3.2881, 3.2572], device='cuda:5'), covar=tensor([0.0201, 0.0673, 0.0470, 0.1988, 0.0699, 0.1015, 0.0546, 0.0790], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0150, 0.0158, 0.0145, 0.0137, 0.0125, 0.0138, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 05:40:03,724 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:40:04,150 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-04-30 05:40:23,655 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150793.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 05:40:39,990 INFO [train.py:904] (5/8) Epoch 15, batch 8700, loss[loss=0.168, simple_loss=0.2547, pruned_loss=0.04058, over 12389.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2723, pruned_loss=0.0451, over 3040752.80 frames. ], batch size: 248, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:41:01,904 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.169e+02 2.751e+02 3.261e+02 6.645e+02, threshold=5.502e+02, percent-clipped=1.0 2023-04-30 05:42:16,619 INFO [train.py:904] (5/8) Epoch 15, batch 8750, loss[loss=0.1941, simple_loss=0.2732, pruned_loss=0.05743, over 12007.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.272, pruned_loss=0.04463, over 3040805.14 frames. ], batch size: 247, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:43:06,228 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6330, 2.6904, 1.8692, 2.7824, 2.1560, 2.8106, 2.1011, 2.3702], device='cuda:5'), covar=tensor([0.0289, 0.0331, 0.1233, 0.0237, 0.0695, 0.0424, 0.1152, 0.0598], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0163, 0.0182, 0.0136, 0.0163, 0.0199, 0.0192, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 05:43:52,666 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150893.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:44:06,372 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150900.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:44:08,767 INFO [train.py:904] (5/8) Epoch 15, batch 8800, loss[loss=0.188, simple_loss=0.28, pruned_loss=0.04797, over 16212.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2708, pruned_loss=0.04346, over 3054560.98 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:44:28,835 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.483e+02 3.172e+02 3.684e+02 7.481e+02, threshold=6.344e+02, percent-clipped=2.0 2023-04-30 05:45:06,399 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150930.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:45:50,518 INFO [train.py:904] (5/8) Epoch 15, batch 8850, loss[loss=0.1803, simple_loss=0.2852, pruned_loss=0.03777, over 16786.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2733, pruned_loss=0.04295, over 3047631.68 frames. ], batch size: 124, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:46:10,221 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:46:46,307 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:47:36,867 INFO [train.py:904] (5/8) Epoch 15, batch 8900, loss[loss=0.2031, simple_loss=0.2997, pruned_loss=0.05329, over 16929.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2735, pruned_loss=0.0422, over 3048392.47 frames. ], batch size: 109, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:47:56,130 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7724, 2.3879, 2.3252, 3.5742, 2.1023, 3.7291, 1.5013, 2.8161], device='cuda:5'), covar=tensor([0.1349, 0.0762, 0.1165, 0.0173, 0.0103, 0.0318, 0.1561, 0.0729], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0163, 0.0185, 0.0161, 0.0194, 0.0205, 0.0188, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-30 05:47:59,490 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.270e+02 2.672e+02 3.305e+02 7.174e+02, threshold=5.344e+02, percent-clipped=2.0 2023-04-30 05:49:35,108 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9682, 3.1303, 2.8543, 5.0121, 3.8238, 4.4782, 1.5786, 3.3667], device='cuda:5'), covar=tensor([0.1184, 0.0615, 0.0991, 0.0109, 0.0174, 0.0264, 0.1456, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0162, 0.0184, 0.0160, 0.0192, 0.0203, 0.0187, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-30 05:49:42,136 INFO [train.py:904] (5/8) Epoch 15, batch 8950, loss[loss=0.1565, simple_loss=0.2516, pruned_loss=0.03064, over 16751.00 frames. ], tot_loss[loss=0.179, simple_loss=0.273, pruned_loss=0.04248, over 3055428.90 frames. ], batch size: 83, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:50:13,713 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 05:51:02,413 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151088.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:51:30,998 INFO [train.py:904] (5/8) Epoch 15, batch 9000, loss[loss=0.1744, simple_loss=0.2638, pruned_loss=0.04247, over 16697.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2699, pruned_loss=0.04106, over 3067375.76 frames. ], batch size: 134, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:51:30,999 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 05:51:40,828 INFO [train.py:938] (5/8) Epoch 15, validation: loss=0.15, simple_loss=0.2539, pruned_loss=0.02307, over 944034.00 frames. 2023-04-30 05:51:40,828 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 05:52:03,721 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 2.164e+02 2.577e+02 3.241e+02 6.734e+02, threshold=5.154e+02, percent-clipped=2.0 2023-04-30 05:53:23,292 INFO [train.py:904] (5/8) Epoch 15, batch 9050, loss[loss=0.168, simple_loss=0.2463, pruned_loss=0.04487, over 16449.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2712, pruned_loss=0.0417, over 3092703.56 frames. ], batch size: 68, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:53:33,222 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6122, 3.6534, 3.4790, 3.1341, 3.2785, 3.5721, 3.3557, 3.3949], device='cuda:5'), covar=tensor([0.0540, 0.0556, 0.0265, 0.0230, 0.0490, 0.0404, 0.1078, 0.0451], device='cuda:5'), in_proj_covar=tensor([0.0254, 0.0350, 0.0299, 0.0283, 0.0310, 0.0328, 0.0205, 0.0353], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 05:54:06,353 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 05:54:28,395 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:54:51,804 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:55:09,974 INFO [train.py:904] (5/8) Epoch 15, batch 9100, loss[loss=0.1805, simple_loss=0.2767, pruned_loss=0.04208, over 16393.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2706, pruned_loss=0.04203, over 3076854.90 frames. ], batch size: 146, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:55:31,076 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.276e+02 2.684e+02 3.281e+02 5.650e+02, threshold=5.369e+02, percent-clipped=3.0 2023-04-30 05:56:26,225 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 05:56:43,008 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-30 05:56:44,655 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:56:54,474 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151245.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:57:08,978 INFO [train.py:904] (5/8) Epoch 15, batch 9150, loss[loss=0.1692, simple_loss=0.2603, pruned_loss=0.03908, over 16427.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2708, pruned_loss=0.04139, over 3092516.62 frames. ], batch size: 68, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:57:18,586 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151256.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:12,747 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:46,466 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:51,681 INFO [train.py:904] (5/8) Epoch 15, batch 9200, loss[loss=0.1695, simple_loss=0.2585, pruned_loss=0.04021, over 16325.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.266, pruned_loss=0.04042, over 3076782.38 frames. ], batch size: 146, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:59:12,240 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 2.499e+02 3.414e+02 4.138e+02 7.678e+02, threshold=6.827e+02, percent-clipped=7.0 2023-04-30 06:00:09,390 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151343.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:00:22,808 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0424, 3.3504, 2.9370, 5.1157, 4.0278, 4.4467, 1.6738, 3.2889], device='cuda:5'), covar=tensor([0.1190, 0.0580, 0.0976, 0.0136, 0.0164, 0.0288, 0.1438, 0.0637], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0160, 0.0182, 0.0159, 0.0190, 0.0202, 0.0185, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:5') 2023-04-30 06:00:24,881 INFO [train.py:904] (5/8) Epoch 15, batch 9250, loss[loss=0.1895, simple_loss=0.2778, pruned_loss=0.05057, over 16773.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2658, pruned_loss=0.04073, over 3059070.59 frames. ], batch size: 124, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:00:41,006 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:00:58,282 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5789, 3.5298, 3.5315, 2.8535, 3.4540, 2.0214, 3.2327, 2.9016], device='cuda:5'), covar=tensor([0.0109, 0.0099, 0.0135, 0.0152, 0.0081, 0.2165, 0.0119, 0.0192], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0129, 0.0175, 0.0158, 0.0149, 0.0189, 0.0163, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:01:43,156 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:02:14,995 INFO [train.py:904] (5/8) Epoch 15, batch 9300, loss[loss=0.1748, simple_loss=0.2649, pruned_loss=0.04241, over 16375.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2641, pruned_loss=0.04034, over 3033427.90 frames. ], batch size: 146, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:02:26,177 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 06:02:37,916 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.225e+02 2.604e+02 3.286e+02 5.862e+02, threshold=5.207e+02, percent-clipped=0.0 2023-04-30 06:03:29,844 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:03:59,125 INFO [train.py:904] (5/8) Epoch 15, batch 9350, loss[loss=0.2022, simple_loss=0.2893, pruned_loss=0.05755, over 16224.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2644, pruned_loss=0.04047, over 3046976.05 frames. ], batch size: 165, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:04:08,111 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 06:04:33,095 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8029, 3.7615, 3.9378, 3.7761, 3.8935, 4.2888, 3.9494, 3.6778], device='cuda:5'), covar=tensor([0.2079, 0.2470, 0.2005, 0.2467, 0.2786, 0.1631, 0.1330, 0.2354], device='cuda:5'), in_proj_covar=tensor([0.0357, 0.0506, 0.0551, 0.0425, 0.0564, 0.0583, 0.0437, 0.0564], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 06:05:36,962 INFO [train.py:904] (5/8) Epoch 15, batch 9400, loss[loss=0.1699, simple_loss=0.2491, pruned_loss=0.04533, over 12452.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2643, pruned_loss=0.04023, over 3041551.34 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:05:59,183 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.126e+02 2.557e+02 3.057e+02 4.455e+02, threshold=5.114e+02, percent-clipped=0.0 2023-04-30 06:06:55,074 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151540.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:07:17,466 INFO [train.py:904] (5/8) Epoch 15, batch 9450, loss[loss=0.1773, simple_loss=0.2671, pruned_loss=0.04378, over 16926.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2659, pruned_loss=0.04018, over 3047064.05 frames. ], batch size: 116, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:07:24,710 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151556.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:07:29,543 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6295, 4.0459, 4.0721, 2.8339, 3.5798, 4.0053, 3.8315, 2.6177], device='cuda:5'), covar=tensor([0.0378, 0.0031, 0.0031, 0.0326, 0.0078, 0.0072, 0.0059, 0.0341], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0071, 0.0071, 0.0129, 0.0086, 0.0094, 0.0083, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 06:08:06,472 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2835, 2.4089, 1.9377, 2.1901, 2.8219, 2.4846, 2.9166, 3.0123], device='cuda:5'), covar=tensor([0.0118, 0.0406, 0.0514, 0.0465, 0.0242, 0.0363, 0.0155, 0.0222], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0214, 0.0208, 0.0207, 0.0213, 0.0213, 0.0210, 0.0203], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:08:53,453 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7356, 5.0439, 4.8631, 4.8364, 4.5924, 4.6254, 4.4937, 5.1080], device='cuda:5'), covar=tensor([0.1108, 0.0880, 0.0847, 0.0661, 0.0757, 0.0844, 0.1018, 0.0863], device='cuda:5'), in_proj_covar=tensor([0.0571, 0.0711, 0.0574, 0.0510, 0.0444, 0.0460, 0.0586, 0.0538], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:08:55,359 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3775, 1.7024, 1.9777, 2.3175, 2.3354, 2.5277, 1.7841, 2.5586], device='cuda:5'), covar=tensor([0.0176, 0.0438, 0.0294, 0.0290, 0.0284, 0.0201, 0.0449, 0.0139], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0176, 0.0161, 0.0165, 0.0175, 0.0133, 0.0177, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:5') 2023-04-30 06:08:58,272 INFO [train.py:904] (5/8) Epoch 15, batch 9500, loss[loss=0.1626, simple_loss=0.2469, pruned_loss=0.03919, over 12859.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2651, pruned_loss=0.03974, over 3061476.00 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:09:04,139 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:09:22,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 2.144e+02 2.700e+02 3.388e+02 6.768e+02, threshold=5.400e+02, percent-clipped=4.0 2023-04-30 06:10:11,909 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151638.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:10:44,461 INFO [train.py:904] (5/8) Epoch 15, batch 9550, loss[loss=0.1714, simple_loss=0.2671, pruned_loss=0.03782, over 15373.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2648, pruned_loss=0.03971, over 3048509.43 frames. ], batch size: 191, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:10:49,463 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151654.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:11:59,322 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151688.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:12:24,738 INFO [train.py:904] (5/8) Epoch 15, batch 9600, loss[loss=0.2052, simple_loss=0.301, pruned_loss=0.0547, over 15233.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2669, pruned_loss=0.04107, over 3025923.79 frames. ], batch size: 190, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:12:44,302 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.311e+02 2.682e+02 3.396e+02 6.074e+02, threshold=5.365e+02, percent-clipped=2.0 2023-04-30 06:12:48,560 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2263, 5.5437, 5.3448, 5.3789, 5.0618, 5.0347, 4.9625, 5.6608], device='cuda:5'), covar=tensor([0.1080, 0.0865, 0.0871, 0.0716, 0.0737, 0.0663, 0.1095, 0.0784], device='cuda:5'), in_proj_covar=tensor([0.0570, 0.0706, 0.0571, 0.0508, 0.0443, 0.0458, 0.0584, 0.0536], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:13:10,833 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-30 06:14:04,278 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:14:11,688 INFO [train.py:904] (5/8) Epoch 15, batch 9650, loss[loss=0.1972, simple_loss=0.2942, pruned_loss=0.05005, over 17021.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2684, pruned_loss=0.04101, over 3031519.35 frames. ], batch size: 55, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:14:59,330 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8164, 3.6890, 3.9176, 3.7511, 3.9048, 4.2802, 3.9631, 3.5972], device='cuda:5'), covar=tensor([0.2084, 0.2605, 0.2186, 0.2480, 0.2878, 0.1829, 0.1492, 0.2715], device='cuda:5'), in_proj_covar=tensor([0.0359, 0.0510, 0.0555, 0.0427, 0.0569, 0.0588, 0.0439, 0.0566], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 06:15:18,609 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0581, 3.8929, 4.1239, 4.2246, 4.3775, 3.9239, 4.3492, 4.3856], device='cuda:5'), covar=tensor([0.1686, 0.1139, 0.1267, 0.0708, 0.0526, 0.1324, 0.0551, 0.0681], device='cuda:5'), in_proj_covar=tensor([0.0537, 0.0665, 0.0783, 0.0680, 0.0512, 0.0533, 0.0544, 0.0631], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:15:58,125 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:15:58,771 INFO [train.py:904] (5/8) Epoch 15, batch 9700, loss[loss=0.17, simple_loss=0.2758, pruned_loss=0.03212, over 16725.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2676, pruned_loss=0.0407, over 3046451.94 frames. ], batch size: 83, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:16:19,906 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.225e+02 2.684e+02 3.461e+02 6.863e+02, threshold=5.368e+02, percent-clipped=3.0 2023-04-30 06:16:24,604 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2486, 4.3087, 4.4756, 4.3232, 4.3592, 4.8258, 4.4070, 4.1268], device='cuda:5'), covar=tensor([0.1602, 0.2034, 0.2261, 0.2035, 0.2518, 0.1091, 0.1462, 0.2318], device='cuda:5'), in_proj_covar=tensor([0.0359, 0.0509, 0.0554, 0.0426, 0.0568, 0.0588, 0.0438, 0.0565], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 06:17:18,862 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151840.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:17:41,657 INFO [train.py:904] (5/8) Epoch 15, batch 9750, loss[loss=0.1593, simple_loss=0.2556, pruned_loss=0.03146, over 16795.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2665, pruned_loss=0.04083, over 3047340.64 frames. ], batch size: 102, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:18:01,721 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:18:11,091 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 06:18:56,065 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:19:18,831 INFO [train.py:904] (5/8) Epoch 15, batch 9800, loss[loss=0.181, simple_loss=0.2779, pruned_loss=0.04199, over 16952.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2667, pruned_loss=0.03991, over 3050413.59 frames. ], batch size: 125, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:19:40,662 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.231e+02 2.655e+02 3.314e+02 7.232e+02, threshold=5.310e+02, percent-clipped=1.0 2023-04-30 06:20:30,679 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:21:03,087 INFO [train.py:904] (5/8) Epoch 15, batch 9850, loss[loss=0.1858, simple_loss=0.2768, pruned_loss=0.04733, over 15463.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2679, pruned_loss=0.0398, over 3050681.29 frames. ], batch size: 191, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:21:08,574 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:22:15,329 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 06:22:17,651 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151986.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:22:40,870 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4764, 3.4772, 3.9137, 1.8971, 4.0348, 4.1488, 3.0802, 3.0292], device='cuda:5'), covar=tensor([0.0727, 0.0239, 0.0155, 0.1107, 0.0060, 0.0106, 0.0321, 0.0405], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0100, 0.0085, 0.0133, 0.0068, 0.0108, 0.0118, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 06:22:49,660 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-30 06:22:57,945 INFO [train.py:904] (5/8) Epoch 15, batch 9900, loss[loss=0.1798, simple_loss=0.2806, pruned_loss=0.03956, over 16242.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2683, pruned_loss=0.03951, over 3054486.71 frames. ], batch size: 165, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:22:58,860 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152002.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:22:59,290 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-30 06:23:24,826 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.244e+02 2.678e+02 3.249e+02 7.284e+02, threshold=5.355e+02, percent-clipped=2.0 2023-04-30 06:24:37,933 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:24:55,587 INFO [train.py:904] (5/8) Epoch 15, batch 9950, loss[loss=0.1914, simple_loss=0.2814, pruned_loss=0.05069, over 12536.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2696, pruned_loss=0.03978, over 3053889.12 frames. ], batch size: 248, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:26:56,978 INFO [train.py:904] (5/8) Epoch 15, batch 10000, loss[loss=0.162, simple_loss=0.2619, pruned_loss=0.03109, over 16512.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2682, pruned_loss=0.03956, over 3048368.90 frames. ], batch size: 75, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:27:18,767 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.235e+02 2.836e+02 3.494e+02 9.282e+02, threshold=5.672e+02, percent-clipped=5.0 2023-04-30 06:28:35,919 INFO [train.py:904] (5/8) Epoch 15, batch 10050, loss[loss=0.1765, simple_loss=0.2781, pruned_loss=0.03749, over 16200.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2682, pruned_loss=0.03943, over 3061793.60 frames. ], batch size: 165, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:28:45,789 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152157.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:28:46,030 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9492, 3.3627, 3.5115, 1.9410, 2.9162, 2.2772, 3.3655, 3.4149], device='cuda:5'), covar=tensor([0.0270, 0.0738, 0.0549, 0.2009, 0.0798, 0.0956, 0.0718, 0.0987], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0144, 0.0157, 0.0145, 0.0136, 0.0123, 0.0136, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 06:29:04,150 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6464, 3.7229, 2.4136, 4.1670, 2.7132, 4.0674, 2.3730, 2.9819], device='cuda:5'), covar=tensor([0.0230, 0.0333, 0.1418, 0.0178, 0.0836, 0.0513, 0.1408, 0.0671], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0161, 0.0183, 0.0135, 0.0166, 0.0198, 0.0194, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 06:30:08,504 INFO [train.py:904] (5/8) Epoch 15, batch 10100, loss[loss=0.1704, simple_loss=0.2599, pruned_loss=0.04048, over 16361.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2677, pruned_loss=0.03913, over 3074017.93 frames. ], batch size: 146, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:30:28,187 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.142e+02 2.576e+02 3.219e+02 1.174e+03, threshold=5.152e+02, percent-clipped=3.0 2023-04-30 06:30:55,491 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:31:05,892 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9503, 2.0882, 2.3749, 3.2354, 2.1733, 2.2851, 2.2242, 2.1583], device='cuda:5'), covar=tensor([0.1132, 0.3517, 0.2387, 0.0591, 0.4119, 0.2531, 0.3133, 0.3443], device='cuda:5'), in_proj_covar=tensor([0.0364, 0.0399, 0.0338, 0.0306, 0.0412, 0.0455, 0.0369, 0.0465], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:31:08,374 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7694, 3.6983, 2.9865, 2.2759, 2.6187, 2.3729, 4.1675, 3.3826], device='cuda:5'), covar=tensor([0.2797, 0.0973, 0.1699, 0.2657, 0.2556, 0.2008, 0.0511, 0.1355], device='cuda:5'), in_proj_covar=tensor([0.0301, 0.0251, 0.0283, 0.0281, 0.0265, 0.0228, 0.0266, 0.0297], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:31:53,397 INFO [train.py:904] (5/8) Epoch 16, batch 0, loss[loss=0.2455, simple_loss=0.3127, pruned_loss=0.0892, over 16726.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3127, pruned_loss=0.0892, over 16726.00 frames. ], batch size: 124, lr: 4.32e-03, grad_scale: 8.0 2023-04-30 06:31:53,397 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 06:32:00,894 INFO [train.py:938] (5/8) Epoch 16, validation: loss=0.1492, simple_loss=0.2525, pruned_loss=0.02288, over 944034.00 frames. 2023-04-30 06:32:00,895 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 06:32:12,334 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1362, 5.1107, 4.9148, 4.6121, 4.9100, 2.1172, 4.7746, 4.7706], device='cuda:5'), covar=tensor([0.0070, 0.0068, 0.0175, 0.0253, 0.0091, 0.2195, 0.0124, 0.0201], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0127, 0.0171, 0.0153, 0.0147, 0.0187, 0.0161, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:32:29,800 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:32:30,291 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 06:32:48,492 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:33:09,976 INFO [train.py:904] (5/8) Epoch 16, batch 50, loss[loss=0.1802, simple_loss=0.2697, pruned_loss=0.04531, over 17073.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2804, pruned_loss=0.05581, over 740941.73 frames. ], batch size: 53, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:33:12,135 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4516, 4.1786, 4.1961, 4.5823, 4.6853, 4.3332, 4.6549, 4.6721], device='cuda:5'), covar=tensor([0.2078, 0.2170, 0.2837, 0.1511, 0.1242, 0.1332, 0.1626, 0.2605], device='cuda:5'), in_proj_covar=tensor([0.0539, 0.0670, 0.0789, 0.0685, 0.0515, 0.0534, 0.0550, 0.0633], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:33:29,876 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 2.475e+02 3.034e+02 3.841e+02 6.460e+02, threshold=6.067e+02, percent-clipped=5.0 2023-04-30 06:33:53,538 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:34:08,302 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:34:18,731 INFO [train.py:904] (5/8) Epoch 16, batch 100, loss[loss=0.1766, simple_loss=0.2546, pruned_loss=0.04933, over 16806.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2745, pruned_loss=0.05368, over 1312481.85 frames. ], batch size: 102, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:34:25,695 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 06:35:14,871 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:35:26,677 INFO [train.py:904] (5/8) Epoch 16, batch 150, loss[loss=0.1685, simple_loss=0.2583, pruned_loss=0.03942, over 17102.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2714, pruned_loss=0.05201, over 1764991.39 frames. ], batch size: 47, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:35:48,056 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.712e+02 2.454e+02 2.817e+02 3.397e+02 1.160e+03, threshold=5.634e+02, percent-clipped=3.0 2023-04-30 06:36:24,993 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:36:35,180 INFO [train.py:904] (5/8) Epoch 16, batch 200, loss[loss=0.1798, simple_loss=0.2753, pruned_loss=0.04212, over 17132.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2722, pruned_loss=0.05211, over 2107489.73 frames. ], batch size: 49, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:36:42,864 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:36:58,572 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8170, 4.7540, 4.6924, 4.2790, 4.7464, 1.9727, 4.5012, 4.4659], device='cuda:5'), covar=tensor([0.0120, 0.0093, 0.0150, 0.0262, 0.0104, 0.2321, 0.0136, 0.0168], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0131, 0.0176, 0.0159, 0.0151, 0.0192, 0.0165, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:37:09,275 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7063, 4.9330, 5.0903, 4.9130, 4.9039, 5.5255, 5.0534, 4.8066], device='cuda:5'), covar=tensor([0.1688, 0.2226, 0.3054, 0.2274, 0.2977, 0.1345, 0.1822, 0.2527], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0536, 0.0583, 0.0450, 0.0596, 0.0622, 0.0458, 0.0597], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 06:37:44,242 INFO [train.py:904] (5/8) Epoch 16, batch 250, loss[loss=0.1939, simple_loss=0.2692, pruned_loss=0.05928, over 16450.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2703, pruned_loss=0.0523, over 2377506.44 frames. ], batch size: 146, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:37:48,058 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152505.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:50,106 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:51,343 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7680, 2.6219, 2.3799, 4.1044, 3.2522, 4.0932, 1.5917, 2.7872], device='cuda:5'), covar=tensor([0.1407, 0.0702, 0.1207, 0.0167, 0.0164, 0.0350, 0.1540, 0.0842], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0163, 0.0185, 0.0162, 0.0191, 0.0207, 0.0188, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-30 06:38:05,713 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.341e+02 2.836e+02 3.695e+02 5.924e+02, threshold=5.672e+02, percent-clipped=1.0 2023-04-30 06:38:53,039 INFO [train.py:904] (5/8) Epoch 16, batch 300, loss[loss=0.1557, simple_loss=0.2408, pruned_loss=0.03529, over 16827.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2671, pruned_loss=0.05003, over 2582123.08 frames. ], batch size: 42, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:39:33,808 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152581.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:35,023 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:43,005 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:40:01,551 INFO [train.py:904] (5/8) Epoch 16, batch 350, loss[loss=0.1983, simple_loss=0.286, pruned_loss=0.05527, over 17090.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2646, pruned_loss=0.04912, over 2751513.88 frames. ], batch size: 53, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:40:20,731 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.239e+02 2.608e+02 3.067e+02 7.666e+02, threshold=5.216e+02, percent-clipped=3.0 2023-04-30 06:40:38,469 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:40:58,729 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:07,029 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152650.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:08,839 INFO [train.py:904] (5/8) Epoch 16, batch 400, loss[loss=0.1865, simple_loss=0.274, pruned_loss=0.04951, over 17093.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2622, pruned_loss=0.04748, over 2881912.21 frames. ], batch size: 53, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:41:12,736 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-30 06:42:10,585 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2585, 4.2602, 4.2603, 3.7313, 4.2362, 1.8534, 4.0402, 3.8780], device='cuda:5'), covar=tensor([0.0122, 0.0101, 0.0147, 0.0289, 0.0105, 0.2442, 0.0133, 0.0219], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0135, 0.0181, 0.0164, 0.0155, 0.0196, 0.0170, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:42:16,094 INFO [train.py:904] (5/8) Epoch 16, batch 450, loss[loss=0.1815, simple_loss=0.2524, pruned_loss=0.05536, over 16755.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2606, pruned_loss=0.04644, over 2981721.21 frames. ], batch size: 134, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:42:31,259 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6172, 4.5942, 4.5605, 4.0297, 4.5311, 1.9466, 4.3390, 4.2256], device='cuda:5'), covar=tensor([0.0107, 0.0091, 0.0162, 0.0323, 0.0114, 0.2381, 0.0144, 0.0202], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0134, 0.0180, 0.0164, 0.0155, 0.0196, 0.0169, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:42:36,236 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.224e+02 2.572e+02 2.956e+02 7.682e+02, threshold=5.144e+02, percent-clipped=1.0 2023-04-30 06:42:50,509 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:43:25,842 INFO [train.py:904] (5/8) Epoch 16, batch 500, loss[loss=0.1759, simple_loss=0.2504, pruned_loss=0.05074, over 12177.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.259, pruned_loss=0.04579, over 3058020.01 frames. ], batch size: 246, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:44:14,813 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152787.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:33,484 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:34,225 INFO [train.py:904] (5/8) Epoch 16, batch 550, loss[loss=0.1728, simple_loss=0.2712, pruned_loss=0.03722, over 17037.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2583, pruned_loss=0.04512, over 3107300.27 frames. ], batch size: 50, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:44:40,355 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6302, 6.0406, 5.7435, 5.7650, 5.3605, 5.3874, 5.4376, 6.1407], device='cuda:5'), covar=tensor([0.1250, 0.0821, 0.1100, 0.0798, 0.0979, 0.0646, 0.1069, 0.0786], device='cuda:5'), in_proj_covar=tensor([0.0614, 0.0759, 0.0616, 0.0546, 0.0476, 0.0487, 0.0631, 0.0577], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:44:55,530 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 2.218e+02 2.743e+02 3.343e+02 5.376e+02, threshold=5.487e+02, percent-clipped=2.0 2023-04-30 06:45:17,906 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 06:45:46,508 INFO [train.py:904] (5/8) Epoch 16, batch 600, loss[loss=0.1527, simple_loss=0.2414, pruned_loss=0.03201, over 17180.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.258, pruned_loss=0.04571, over 3156068.91 frames. ], batch size: 46, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:46:15,371 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 06:46:25,177 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:46:41,317 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 06:46:53,590 INFO [train.py:904] (5/8) Epoch 16, batch 650, loss[loss=0.1737, simple_loss=0.2669, pruned_loss=0.04028, over 16731.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2572, pruned_loss=0.04515, over 3190479.42 frames. ], batch size: 57, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:47:14,338 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.133e+02 2.528e+02 3.105e+02 5.746e+02, threshold=5.056e+02, percent-clipped=1.0 2023-04-30 06:47:30,635 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:30,802 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:43,504 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:52,173 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:48:00,904 INFO [train.py:904] (5/8) Epoch 16, batch 700, loss[loss=0.173, simple_loss=0.2579, pruned_loss=0.04408, over 16820.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2575, pruned_loss=0.04528, over 3224059.64 frames. ], batch size: 102, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:48:37,714 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:49:10,745 INFO [train.py:904] (5/8) Epoch 16, batch 750, loss[loss=0.1944, simple_loss=0.2721, pruned_loss=0.05837, over 12285.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2578, pruned_loss=0.04469, over 3246111.61 frames. ], batch size: 246, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:49:31,105 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.303e+02 2.598e+02 3.090e+02 5.870e+02, threshold=5.196e+02, percent-clipped=1.0 2023-04-30 06:49:47,384 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 06:49:57,740 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7967, 2.4872, 2.0069, 2.2621, 2.9059, 2.6993, 2.9349, 2.9859], device='cuda:5'), covar=tensor([0.0199, 0.0368, 0.0475, 0.0457, 0.0219, 0.0324, 0.0245, 0.0254], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0225, 0.0217, 0.0219, 0.0225, 0.0225, 0.0226, 0.0217], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:50:08,607 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 06:50:17,794 INFO [train.py:904] (5/8) Epoch 16, batch 800, loss[loss=0.1825, simple_loss=0.2712, pruned_loss=0.04686, over 16699.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2568, pruned_loss=0.04445, over 3266633.98 frames. ], batch size: 57, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:50:21,710 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5139, 2.2169, 2.3014, 4.3484, 2.1980, 2.6955, 2.3103, 2.3590], device='cuda:5'), covar=tensor([0.1141, 0.3673, 0.2792, 0.0444, 0.4096, 0.2479, 0.3473, 0.3708], device='cuda:5'), in_proj_covar=tensor([0.0380, 0.0418, 0.0352, 0.0322, 0.0427, 0.0478, 0.0386, 0.0489], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:51:00,664 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:25,490 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153101.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:26,416 INFO [train.py:904] (5/8) Epoch 16, batch 850, loss[loss=0.1708, simple_loss=0.2423, pruned_loss=0.0496, over 16887.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2562, pruned_loss=0.04459, over 3265861.58 frames. ], batch size: 109, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:51:31,563 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 06:51:46,530 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 2.045e+02 2.529e+02 2.957e+02 4.458e+02, threshold=5.058e+02, percent-clipped=0.0 2023-04-30 06:51:46,994 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153117.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:52:32,207 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:52:35,401 INFO [train.py:904] (5/8) Epoch 16, batch 900, loss[loss=0.1826, simple_loss=0.2582, pruned_loss=0.05354, over 16829.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2553, pruned_loss=0.04404, over 3280385.95 frames. ], batch size: 102, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:53:11,020 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:53:14,266 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1450, 4.0952, 4.4656, 4.4657, 4.4878, 4.2060, 4.2171, 4.1163], device='cuda:5'), covar=tensor([0.0344, 0.0635, 0.0438, 0.0441, 0.0522, 0.0438, 0.0831, 0.0577], device='cuda:5'), in_proj_covar=tensor([0.0374, 0.0406, 0.0395, 0.0378, 0.0445, 0.0419, 0.0510, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 06:53:43,187 INFO [train.py:904] (5/8) Epoch 16, batch 950, loss[loss=0.1628, simple_loss=0.2477, pruned_loss=0.03895, over 17216.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2553, pruned_loss=0.0445, over 3274052.02 frames. ], batch size: 44, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:53:53,519 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 06:54:04,604 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.182e+02 2.578e+02 3.357e+02 5.954e+02, threshold=5.156e+02, percent-clipped=2.0 2023-04-30 06:54:33,130 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:42,480 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153245.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:53,049 INFO [train.py:904] (5/8) Epoch 16, batch 1000, loss[loss=0.1547, simple_loss=0.2386, pruned_loss=0.03535, over 16961.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2543, pruned_loss=0.04418, over 3282412.94 frames. ], batch size: 41, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:55:29,855 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153278.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:55:40,270 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:55:51,058 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:56:02,854 INFO [train.py:904] (5/8) Epoch 16, batch 1050, loss[loss=0.1876, simple_loss=0.27, pruned_loss=0.05258, over 16828.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2543, pruned_loss=0.04444, over 3300024.65 frames. ], batch size: 109, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:56:20,183 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6058, 4.6272, 4.9329, 4.9407, 4.9632, 4.6718, 4.6494, 4.4709], device='cuda:5'), covar=tensor([0.0308, 0.0638, 0.0401, 0.0401, 0.0466, 0.0377, 0.0779, 0.0498], device='cuda:5'), in_proj_covar=tensor([0.0375, 0.0406, 0.0396, 0.0379, 0.0446, 0.0419, 0.0511, 0.0332], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 06:56:24,685 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.258e+02 2.836e+02 3.297e+02 6.244e+02, threshold=5.672e+02, percent-clipped=2.0 2023-04-30 06:56:55,477 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153339.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:56:55,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4806, 2.8956, 2.6580, 2.2544, 2.2325, 2.2874, 2.8587, 2.7509], device='cuda:5'), covar=tensor([0.2406, 0.0819, 0.1525, 0.2284, 0.2281, 0.1902, 0.0511, 0.1212], device='cuda:5'), in_proj_covar=tensor([0.0313, 0.0263, 0.0294, 0.0291, 0.0282, 0.0237, 0.0278, 0.0315], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 06:57:12,827 INFO [train.py:904] (5/8) Epoch 16, batch 1100, loss[loss=0.157, simple_loss=0.2523, pruned_loss=0.03084, over 17129.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2536, pruned_loss=0.04396, over 3303369.88 frames. ], batch size: 47, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:57:39,637 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 06:57:54,341 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:58:21,564 INFO [train.py:904] (5/8) Epoch 16, batch 1150, loss[loss=0.157, simple_loss=0.2355, pruned_loss=0.03926, over 16885.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2527, pruned_loss=0.04318, over 3315368.64 frames. ], batch size: 90, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:58:24,627 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6178, 4.9381, 4.7561, 4.7339, 4.4846, 4.4302, 4.4140, 4.9967], device='cuda:5'), covar=tensor([0.1198, 0.0853, 0.0953, 0.0874, 0.0803, 0.1244, 0.1058, 0.0889], device='cuda:5'), in_proj_covar=tensor([0.0629, 0.0775, 0.0631, 0.0558, 0.0487, 0.0498, 0.0648, 0.0593], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 06:58:30,340 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6798, 4.7902, 4.9823, 4.8031, 4.7876, 5.4318, 4.9334, 4.6406], device='cuda:5'), covar=tensor([0.1439, 0.2168, 0.2343, 0.2363, 0.2877, 0.1147, 0.1814, 0.2628], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0558, 0.0605, 0.0469, 0.0624, 0.0637, 0.0474, 0.0616], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 06:58:42,016 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.113e+02 2.608e+02 3.086e+02 6.684e+02, threshold=5.217e+02, percent-clipped=1.0 2023-04-30 06:58:46,834 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:58:58,553 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153430.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:59:21,311 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 06:59:27,884 INFO [train.py:904] (5/8) Epoch 16, batch 1200, loss[loss=0.1735, simple_loss=0.2472, pruned_loss=0.04987, over 16805.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2526, pruned_loss=0.04278, over 3315183.92 frames. ], batch size: 102, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 06:59:57,488 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:10,781 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:37,164 INFO [train.py:904] (5/8) Epoch 16, batch 1250, loss[loss=0.1865, simple_loss=0.2575, pruned_loss=0.0577, over 16558.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2529, pruned_loss=0.04331, over 3313399.93 frames. ], batch size: 146, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:00:57,400 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.233e+02 2.659e+02 3.231e+02 7.143e+02, threshold=5.319e+02, percent-clipped=2.0 2023-04-30 07:01:24,211 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5405, 4.4939, 4.5053, 4.0237, 4.4634, 1.8353, 4.2616, 4.1909], device='cuda:5'), covar=tensor([0.0115, 0.0095, 0.0156, 0.0296, 0.0094, 0.2473, 0.0131, 0.0181], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0139, 0.0186, 0.0170, 0.0159, 0.0199, 0.0174, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:01:43,336 INFO [train.py:904] (5/8) Epoch 16, batch 1300, loss[loss=0.1326, simple_loss=0.2174, pruned_loss=0.02384, over 15944.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2525, pruned_loss=0.04314, over 3302489.90 frames. ], batch size: 35, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:02:04,145 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-30 07:02:52,654 INFO [train.py:904] (5/8) Epoch 16, batch 1350, loss[loss=0.187, simple_loss=0.2654, pruned_loss=0.05433, over 16605.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.253, pruned_loss=0.04312, over 3306072.85 frames. ], batch size: 68, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:03:12,848 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.319e+02 2.697e+02 3.051e+02 7.500e+02, threshold=5.394e+02, percent-clipped=1.0 2023-04-30 07:03:37,069 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:04:02,091 INFO [train.py:904] (5/8) Epoch 16, batch 1400, loss[loss=0.1566, simple_loss=0.2331, pruned_loss=0.04001, over 15576.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2527, pruned_loss=0.04293, over 3308056.19 frames. ], batch size: 191, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:04:56,871 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3394, 4.2659, 4.2499, 3.7973, 4.2828, 1.8304, 4.0680, 3.9162], device='cuda:5'), covar=tensor([0.0103, 0.0098, 0.0152, 0.0268, 0.0083, 0.2419, 0.0123, 0.0180], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0138, 0.0186, 0.0170, 0.0159, 0.0198, 0.0174, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:05:12,118 INFO [train.py:904] (5/8) Epoch 16, batch 1450, loss[loss=0.1634, simple_loss=0.2382, pruned_loss=0.04435, over 16746.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2524, pruned_loss=0.04328, over 3308160.69 frames. ], batch size: 89, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:05:33,604 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-04-30 07:05:34,106 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.274e+02 2.601e+02 3.378e+02 7.170e+02, threshold=5.202e+02, percent-clipped=2.0 2023-04-30 07:06:22,462 INFO [train.py:904] (5/8) Epoch 16, batch 1500, loss[loss=0.158, simple_loss=0.2399, pruned_loss=0.03806, over 16301.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2519, pruned_loss=0.04304, over 3316873.42 frames. ], batch size: 36, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:06:27,648 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3696, 2.3127, 2.3641, 4.2689, 2.2614, 2.7073, 2.3207, 2.4765], device='cuda:5'), covar=tensor([0.1253, 0.3617, 0.2819, 0.0511, 0.3891, 0.2514, 0.3642, 0.3156], device='cuda:5'), in_proj_covar=tensor([0.0385, 0.0421, 0.0354, 0.0325, 0.0428, 0.0485, 0.0390, 0.0495], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:06:50,976 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:06:56,060 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:07:09,864 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7680, 3.9058, 2.6489, 4.5251, 3.0570, 4.4433, 2.6040, 3.1834], device='cuda:5'), covar=tensor([0.0265, 0.0342, 0.1275, 0.0177, 0.0750, 0.0442, 0.1280, 0.0655], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0172, 0.0192, 0.0152, 0.0173, 0.0214, 0.0201, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 07:07:30,685 INFO [train.py:904] (5/8) Epoch 16, batch 1550, loss[loss=0.1461, simple_loss=0.2286, pruned_loss=0.03176, over 16765.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2529, pruned_loss=0.04428, over 3320261.84 frames. ], batch size: 39, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:07:53,743 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.189e+02 2.739e+02 3.070e+02 6.481e+02, threshold=5.478e+02, percent-clipped=4.0 2023-04-30 07:07:58,991 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153821.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:08:22,337 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4923, 3.8249, 3.9910, 2.1601, 3.1109, 2.5931, 3.9307, 4.0239], device='cuda:5'), covar=tensor([0.0289, 0.0807, 0.0483, 0.1865, 0.0814, 0.0951, 0.0565, 0.0925], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0154, 0.0160, 0.0148, 0.0139, 0.0125, 0.0140, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 07:08:40,378 INFO [train.py:904] (5/8) Epoch 16, batch 1600, loss[loss=0.1883, simple_loss=0.2669, pruned_loss=0.05484, over 16446.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2553, pruned_loss=0.04514, over 3321275.76 frames. ], batch size: 68, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:09:15,309 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0403, 2.1807, 2.6276, 2.9704, 2.7311, 3.4175, 2.5077, 3.5464], device='cuda:5'), covar=tensor([0.0194, 0.0428, 0.0251, 0.0242, 0.0291, 0.0161, 0.0368, 0.0127], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0185, 0.0171, 0.0174, 0.0184, 0.0140, 0.0185, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:09:47,767 INFO [train.py:904] (5/8) Epoch 16, batch 1650, loss[loss=0.1824, simple_loss=0.2792, pruned_loss=0.0428, over 17034.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.258, pruned_loss=0.04634, over 3312626.77 frames. ], batch size: 50, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:10:09,062 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.549e+02 2.974e+02 3.703e+02 7.236e+02, threshold=5.949e+02, percent-clipped=5.0 2023-04-30 07:10:11,690 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8978, 5.4310, 5.5580, 5.3108, 5.2593, 5.9600, 5.4326, 5.1034], device='cuda:5'), covar=tensor([0.1116, 0.2017, 0.2203, 0.2020, 0.3078, 0.1007, 0.1370, 0.2488], device='cuda:5'), in_proj_covar=tensor([0.0395, 0.0568, 0.0617, 0.0475, 0.0638, 0.0649, 0.0483, 0.0630], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 07:10:13,599 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2893, 4.2113, 4.2654, 3.8012, 4.2640, 1.9164, 4.0620, 3.8982], device='cuda:5'), covar=tensor([0.0115, 0.0099, 0.0138, 0.0284, 0.0086, 0.2350, 0.0125, 0.0185], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0138, 0.0186, 0.0170, 0.0159, 0.0198, 0.0174, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:10:15,978 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:32,350 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153934.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:56,045 INFO [train.py:904] (5/8) Epoch 16, batch 1700, loss[loss=0.1848, simple_loss=0.2619, pruned_loss=0.05382, over 16821.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2596, pruned_loss=0.04657, over 3313456.02 frames. ], batch size: 116, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:11:21,840 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8945, 4.3401, 3.2761, 2.4155, 2.8318, 2.5399, 4.6881, 3.7452], device='cuda:5'), covar=tensor([0.2528, 0.0585, 0.1516, 0.2537, 0.2703, 0.1870, 0.0336, 0.1159], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0263, 0.0292, 0.0291, 0.0282, 0.0237, 0.0278, 0.0314], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 07:11:38,411 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153982.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:11:40,916 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:12:09,300 INFO [train.py:904] (5/8) Epoch 16, batch 1750, loss[loss=0.1641, simple_loss=0.2446, pruned_loss=0.04178, over 16625.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2603, pruned_loss=0.04628, over 3319236.35 frames. ], batch size: 89, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:12:33,143 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.261e+02 2.728e+02 3.333e+02 9.430e+02, threshold=5.456e+02, percent-clipped=2.0 2023-04-30 07:13:15,084 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154049.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:13:18,933 INFO [train.py:904] (5/8) Epoch 16, batch 1800, loss[loss=0.1946, simple_loss=0.2755, pruned_loss=0.05681, over 16865.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2613, pruned_loss=0.04643, over 3313215.06 frames. ], batch size: 96, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:13:53,206 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:13:54,295 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:03,099 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:06,428 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3689, 5.3146, 5.0555, 4.5322, 5.1646, 2.1027, 4.9108, 5.0681], device='cuda:5'), covar=tensor([0.0073, 0.0064, 0.0181, 0.0413, 0.0090, 0.2339, 0.0115, 0.0176], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0139, 0.0187, 0.0172, 0.0160, 0.0199, 0.0176, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:14:28,063 INFO [train.py:904] (5/8) Epoch 16, batch 1850, loss[loss=0.2111, simple_loss=0.287, pruned_loss=0.06762, over 16730.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2623, pruned_loss=0.04637, over 3316469.91 frames. ], batch size: 134, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:14:28,793 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-04-30 07:14:39,356 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154110.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 07:14:50,224 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.254e+02 2.561e+02 3.083e+02 7.849e+02, threshold=5.121e+02, percent-clipped=3.0 2023-04-30 07:14:59,482 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:15:13,946 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6199, 2.2541, 1.8215, 2.1494, 2.6985, 2.4870, 2.7556, 2.8249], device='cuda:5'), covar=tensor([0.0182, 0.0369, 0.0471, 0.0388, 0.0204, 0.0290, 0.0197, 0.0245], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0227, 0.0220, 0.0219, 0.0228, 0.0229, 0.0232, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:15:15,709 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154137.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:15:25,879 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:15:33,374 INFO [train.py:904] (5/8) Epoch 16, batch 1900, loss[loss=0.1748, simple_loss=0.2553, pruned_loss=0.04713, over 16793.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2614, pruned_loss=0.04634, over 3316272.32 frames. ], batch size: 124, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:16:28,077 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7788, 3.8832, 4.1455, 4.1202, 4.1300, 3.8653, 3.9289, 3.8352], device='cuda:5'), covar=tensor([0.0344, 0.0473, 0.0350, 0.0381, 0.0430, 0.0436, 0.0668, 0.0520], device='cuda:5'), in_proj_covar=tensor([0.0383, 0.0412, 0.0400, 0.0382, 0.0452, 0.0430, 0.0524, 0.0341], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 07:16:44,644 INFO [train.py:904] (5/8) Epoch 16, batch 1950, loss[loss=0.1523, simple_loss=0.234, pruned_loss=0.03526, over 15781.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2609, pruned_loss=0.04579, over 3310119.43 frames. ], batch size: 35, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:16:54,271 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8268, 3.9335, 2.6015, 4.5648, 3.1670, 4.4734, 2.6090, 3.1466], device='cuda:5'), covar=tensor([0.0288, 0.0376, 0.1503, 0.0239, 0.0782, 0.0490, 0.1476, 0.0764], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0175, 0.0195, 0.0154, 0.0174, 0.0217, 0.0203, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 07:17:04,525 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.235e+02 2.551e+02 3.026e+02 6.742e+02, threshold=5.103e+02, percent-clipped=2.0 2023-04-30 07:17:10,519 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:17:51,474 INFO [train.py:904] (5/8) Epoch 16, batch 2000, loss[loss=0.1896, simple_loss=0.2764, pruned_loss=0.05137, over 15551.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2601, pruned_loss=0.04528, over 3315797.94 frames. ], batch size: 191, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:18:27,780 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154279.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:33,449 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154282.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:18:38,749 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:40,029 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0318, 2.2683, 2.6933, 3.0363, 2.7978, 3.4277, 2.3222, 3.4935], device='cuda:5'), covar=tensor([0.0195, 0.0372, 0.0246, 0.0238, 0.0256, 0.0155, 0.0388, 0.0139], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0185, 0.0171, 0.0175, 0.0185, 0.0141, 0.0186, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:18:43,457 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7224, 2.7799, 2.5304, 2.6209, 3.0081, 2.8580, 3.4616, 3.2966], device='cuda:5'), covar=tensor([0.0113, 0.0330, 0.0386, 0.0389, 0.0233, 0.0297, 0.0202, 0.0234], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0225, 0.0218, 0.0219, 0.0227, 0.0229, 0.0232, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:18:44,574 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:59,384 INFO [train.py:904] (5/8) Epoch 16, batch 2050, loss[loss=0.1935, simple_loss=0.264, pruned_loss=0.06154, over 16853.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.26, pruned_loss=0.04517, over 3314039.98 frames. ], batch size: 109, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:19:19,612 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.241e+02 2.674e+02 3.116e+02 4.900e+02, threshold=5.347e+02, percent-clipped=0.0 2023-04-30 07:19:46,327 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154337.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:20:01,094 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:06,451 INFO [train.py:904] (5/8) Epoch 16, batch 2100, loss[loss=0.1529, simple_loss=0.2338, pruned_loss=0.03597, over 17004.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2612, pruned_loss=0.0453, over 3325811.13 frames. ], batch size: 41, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:20:06,977 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154352.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:46,456 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 07:21:09,683 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154398.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:21:14,695 INFO [train.py:904] (5/8) Epoch 16, batch 2150, loss[loss=0.1484, simple_loss=0.2316, pruned_loss=0.03256, over 16748.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2616, pruned_loss=0.04515, over 3332145.11 frames. ], batch size: 39, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:21:19,327 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154405.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:21:36,521 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.339e+02 2.706e+02 3.339e+02 5.041e+02, threshold=5.411e+02, percent-clipped=0.0 2023-04-30 07:21:57,916 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154432.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:21:58,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6532, 2.3074, 2.3481, 4.4470, 2.1859, 2.7768, 2.3714, 2.5580], device='cuda:5'), covar=tensor([0.1112, 0.3718, 0.2786, 0.0423, 0.4131, 0.2490, 0.3538, 0.3475], device='cuda:5'), in_proj_covar=tensor([0.0384, 0.0421, 0.0352, 0.0324, 0.0425, 0.0484, 0.0390, 0.0493], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:22:07,251 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154440.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:22:25,102 INFO [train.py:904] (5/8) Epoch 16, batch 2200, loss[loss=0.1702, simple_loss=0.2622, pruned_loss=0.03916, over 17121.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2625, pruned_loss=0.04562, over 3327198.92 frames. ], batch size: 47, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:22:55,744 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:23:32,249 INFO [train.py:904] (5/8) Epoch 16, batch 2250, loss[loss=0.1628, simple_loss=0.2581, pruned_loss=0.03371, over 17042.00 frames. ], tot_loss[loss=0.178, simple_loss=0.263, pruned_loss=0.04645, over 3328864.04 frames. ], batch size: 50, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:23:55,134 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.236e+02 2.597e+02 3.012e+02 8.725e+02, threshold=5.194e+02, percent-clipped=2.0 2023-04-30 07:24:18,878 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154535.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:24:25,912 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 07:24:31,596 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7518, 4.5534, 4.7966, 4.9597, 5.1239, 4.4749, 5.1084, 5.1461], device='cuda:5'), covar=tensor([0.1663, 0.1343, 0.1575, 0.0727, 0.0540, 0.1095, 0.0649, 0.0538], device='cuda:5'), in_proj_covar=tensor([0.0613, 0.0762, 0.0904, 0.0769, 0.0577, 0.0603, 0.0614, 0.0714], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:24:40,030 INFO [train.py:904] (5/8) Epoch 16, batch 2300, loss[loss=0.1794, simple_loss=0.2701, pruned_loss=0.04434, over 16668.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2623, pruned_loss=0.04575, over 3336285.09 frames. ], batch size: 62, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:25:16,759 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154577.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:25:20,043 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154579.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:25:51,262 INFO [train.py:904] (5/8) Epoch 16, batch 2350, loss[loss=0.1787, simple_loss=0.2697, pruned_loss=0.04389, over 16975.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2628, pruned_loss=0.04637, over 3329433.81 frames. ], batch size: 41, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:26:10,021 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4615, 4.7870, 5.1357, 5.0827, 5.0904, 4.7514, 4.4299, 4.4790], device='cuda:5'), covar=tensor([0.0630, 0.0619, 0.0558, 0.0707, 0.0764, 0.0626, 0.1613, 0.0609], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0408, 0.0396, 0.0378, 0.0449, 0.0423, 0.0518, 0.0335], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 07:26:11,866 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.266e+02 2.563e+02 3.089e+02 4.962e+02, threshold=5.126e+02, percent-clipped=0.0 2023-04-30 07:26:21,977 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 07:26:25,330 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154627.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:26,798 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:45,841 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154642.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:52,322 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154647.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:57,946 INFO [train.py:904] (5/8) Epoch 16, batch 2400, loss[loss=0.2056, simple_loss=0.2796, pruned_loss=0.0658, over 16866.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2648, pruned_loss=0.04721, over 3315285.84 frames. ], batch size: 116, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:27:50,996 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154689.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:27:55,619 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154693.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 07:28:06,464 INFO [train.py:904] (5/8) Epoch 16, batch 2450, loss[loss=0.1967, simple_loss=0.2711, pruned_loss=0.06117, over 16732.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2659, pruned_loss=0.04703, over 3313038.19 frames. ], batch size: 124, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:28:12,160 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154705.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:28:28,922 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.204e+02 2.505e+02 3.030e+02 6.585e+02, threshold=5.010e+02, percent-clipped=3.0 2023-04-30 07:28:31,871 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:50,300 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:59,905 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:01,183 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2268, 5.2591, 5.0400, 4.4882, 5.0534, 1.9657, 4.8815, 5.0803], device='cuda:5'), covar=tensor([0.0078, 0.0073, 0.0190, 0.0413, 0.0107, 0.2522, 0.0151, 0.0175], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0142, 0.0191, 0.0175, 0.0163, 0.0202, 0.0179, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:29:16,069 INFO [train.py:904] (5/8) Epoch 16, batch 2500, loss[loss=0.1636, simple_loss=0.2616, pruned_loss=0.03284, over 17070.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2655, pruned_loss=0.04627, over 3318777.76 frames. ], batch size: 50, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:29:18,086 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154753.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:29:53,791 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:55,112 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154781.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:30:05,087 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:30:24,150 INFO [train.py:904] (5/8) Epoch 16, batch 2550, loss[loss=0.2003, simple_loss=0.2719, pruned_loss=0.06432, over 16710.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2654, pruned_loss=0.04608, over 3321842.55 frames. ], batch size: 124, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:30:47,017 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.496e+02 2.136e+02 2.491e+02 3.165e+02 6.684e+02, threshold=4.981e+02, percent-clipped=5.0 2023-04-30 07:31:02,632 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:31:32,693 INFO [train.py:904] (5/8) Epoch 16, batch 2600, loss[loss=0.1595, simple_loss=0.2517, pruned_loss=0.03366, over 17209.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2655, pruned_loss=0.04647, over 3320131.06 frames. ], batch size: 45, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:32:07,865 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154877.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:32:28,371 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7440, 2.2383, 1.6169, 1.9363, 2.6602, 2.4600, 2.8632, 2.7796], device='cuda:5'), covar=tensor([0.0219, 0.0499, 0.0675, 0.0590, 0.0261, 0.0406, 0.0266, 0.0350], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0225, 0.0218, 0.0218, 0.0227, 0.0228, 0.0233, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:32:31,423 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-30 07:32:40,430 INFO [train.py:904] (5/8) Epoch 16, batch 2650, loss[loss=0.1884, simple_loss=0.2711, pruned_loss=0.05287, over 16836.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.266, pruned_loss=0.04659, over 3316502.75 frames. ], batch size: 109, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:32:48,567 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3907, 4.2236, 4.4257, 4.5751, 4.7018, 4.2543, 4.5435, 4.6858], device='cuda:5'), covar=tensor([0.1598, 0.1238, 0.1401, 0.0778, 0.0730, 0.1191, 0.1770, 0.1005], device='cuda:5'), in_proj_covar=tensor([0.0618, 0.0769, 0.0913, 0.0777, 0.0584, 0.0615, 0.0619, 0.0723], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:33:01,396 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.264e+02 2.671e+02 3.159e+02 6.185e+02, threshold=5.341e+02, percent-clipped=3.0 2023-04-30 07:33:12,799 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:32,307 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 07:33:35,823 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:37,563 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 07:33:43,669 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:49,084 INFO [train.py:904] (5/8) Epoch 16, batch 2700, loss[loss=0.1821, simple_loss=0.2609, pruned_loss=0.0516, over 16398.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2659, pruned_loss=0.04628, over 3322393.12 frames. ], batch size: 146, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:34:22,473 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154976.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:32,704 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:41,398 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:45,846 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154993.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:34:47,987 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154995.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:56,800 INFO [train.py:904] (5/8) Epoch 16, batch 2750, loss[loss=0.1775, simple_loss=0.2651, pruned_loss=0.04495, over 17208.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2653, pruned_loss=0.04559, over 3323007.00 frames. ], batch size: 46, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:35:18,258 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.250e+02 2.149e+02 2.425e+02 2.902e+02 6.172e+02, threshold=4.850e+02, percent-clipped=1.0 2023-04-30 07:35:18,506 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1202, 5.6184, 5.7131, 5.4726, 5.4796, 6.0940, 5.6366, 5.3207], device='cuda:5'), covar=tensor([0.0824, 0.1815, 0.1904, 0.1957, 0.2593, 0.0932, 0.1242, 0.2132], device='cuda:5'), in_proj_covar=tensor([0.0401, 0.0569, 0.0618, 0.0481, 0.0639, 0.0651, 0.0485, 0.0635], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 07:35:46,153 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:35:47,545 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-04-30 07:35:50,516 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155041.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:36:02,189 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8862, 1.9734, 2.3338, 2.9176, 2.6331, 3.2991, 2.1931, 3.2876], device='cuda:5'), covar=tensor([0.0216, 0.0424, 0.0328, 0.0281, 0.0309, 0.0159, 0.0424, 0.0125], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0186, 0.0172, 0.0176, 0.0186, 0.0142, 0.0185, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:36:04,028 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9899, 4.5411, 4.5590, 3.3052, 3.7526, 4.4937, 3.9712, 2.8239], device='cuda:5'), covar=tensor([0.0432, 0.0048, 0.0032, 0.0309, 0.0113, 0.0073, 0.0085, 0.0366], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0078, 0.0077, 0.0133, 0.0090, 0.0100, 0.0089, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 07:36:04,757 INFO [train.py:904] (5/8) Epoch 16, batch 2800, loss[loss=0.1555, simple_loss=0.2459, pruned_loss=0.03257, over 16882.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2652, pruned_loss=0.0457, over 3318308.92 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:36:28,459 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:38,310 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:59,259 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 07:37:15,732 INFO [train.py:904] (5/8) Epoch 16, batch 2850, loss[loss=0.188, simple_loss=0.2866, pruned_loss=0.04466, over 17305.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2647, pruned_loss=0.04538, over 3315192.22 frames. ], batch size: 52, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:37:36,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 2.238e+02 2.628e+02 3.275e+02 9.061e+02, threshold=5.256e+02, percent-clipped=3.0 2023-04-30 07:37:43,787 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 07:37:54,451 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:37:54,547 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:38:23,957 INFO [train.py:904] (5/8) Epoch 16, batch 2900, loss[loss=0.1581, simple_loss=0.2491, pruned_loss=0.03355, over 17147.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2636, pruned_loss=0.04544, over 3321417.91 frames. ], batch size: 47, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:01,148 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:39:22,327 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0049, 5.4117, 5.6148, 5.3819, 5.3742, 5.9945, 5.4658, 5.1237], device='cuda:5'), covar=tensor([0.1035, 0.1847, 0.2033, 0.1944, 0.2608, 0.0957, 0.1441, 0.2517], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0572, 0.0619, 0.0484, 0.0640, 0.0649, 0.0487, 0.0638], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 07:39:32,504 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9684, 3.8524, 4.2273, 2.1611, 4.4470, 4.4731, 3.2543, 3.4749], device='cuda:5'), covar=tensor([0.0682, 0.0238, 0.0279, 0.1148, 0.0080, 0.0204, 0.0436, 0.0375], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0138, 0.0074, 0.0119, 0.0126, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 07:39:33,220 INFO [train.py:904] (5/8) Epoch 16, batch 2950, loss[loss=0.1905, simple_loss=0.2799, pruned_loss=0.05053, over 17016.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2633, pruned_loss=0.04615, over 3322717.58 frames. ], batch size: 55, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:53,710 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-30 07:39:54,115 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.372e+02 2.734e+02 3.300e+02 7.430e+02, threshold=5.468e+02, percent-clipped=3.0 2023-04-30 07:40:40,763 INFO [train.py:904] (5/8) Epoch 16, batch 3000, loss[loss=0.1994, simple_loss=0.2774, pruned_loss=0.06066, over 16872.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2637, pruned_loss=0.0468, over 3320859.36 frames. ], batch size: 116, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:40:40,764 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 07:40:47,323 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8663, 4.8707, 5.2339, 5.2023, 5.1860, 4.9044, 4.8364, 4.7348], device='cuda:5'), covar=tensor([0.0289, 0.0482, 0.0370, 0.0379, 0.0395, 0.0345, 0.0852, 0.0344], device='cuda:5'), in_proj_covar=tensor([0.0389, 0.0421, 0.0409, 0.0387, 0.0460, 0.0436, 0.0532, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 07:40:49,854 INFO [train.py:938] (5/8) Epoch 16, validation: loss=0.1368, simple_loss=0.2429, pruned_loss=0.01541, over 944034.00 frames. 2023-04-30 07:40:49,855 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 07:41:27,924 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1679, 2.1936, 2.6664, 3.0916, 2.8692, 3.4124, 2.4935, 3.5631], device='cuda:5'), covar=tensor([0.0176, 0.0413, 0.0266, 0.0263, 0.0267, 0.0169, 0.0365, 0.0118], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0186, 0.0172, 0.0175, 0.0186, 0.0142, 0.0184, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:41:34,712 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:41:49,302 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0017, 4.9485, 4.7932, 4.2316, 4.8805, 1.8886, 4.6017, 4.6692], device='cuda:5'), covar=tensor([0.0139, 0.0121, 0.0245, 0.0408, 0.0130, 0.2704, 0.0172, 0.0209], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0142, 0.0190, 0.0175, 0.0162, 0.0201, 0.0178, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:41:59,659 INFO [train.py:904] (5/8) Epoch 16, batch 3050, loss[loss=0.1831, simple_loss=0.2733, pruned_loss=0.04644, over 17132.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2627, pruned_loss=0.0463, over 3328720.48 frames. ], batch size: 48, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:42:21,042 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.286e+02 2.843e+02 3.408e+02 8.038e+02, threshold=5.686e+02, percent-clipped=2.0 2023-04-30 07:42:27,362 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 07:42:42,354 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:42:42,373 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:42:44,731 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8736, 5.1431, 5.2899, 5.1358, 5.1428, 5.7513, 5.2864, 4.9549], device='cuda:5'), covar=tensor([0.1192, 0.1813, 0.2150, 0.2049, 0.2514, 0.1034, 0.1532, 0.2469], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0568, 0.0613, 0.0481, 0.0637, 0.0643, 0.0482, 0.0633], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 07:42:52,669 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1874, 4.8250, 5.1302, 5.3405, 5.5567, 4.8463, 5.5241, 5.5080], device='cuda:5'), covar=tensor([0.1411, 0.1293, 0.1611, 0.0695, 0.0453, 0.0827, 0.0444, 0.0507], device='cuda:5'), in_proj_covar=tensor([0.0619, 0.0769, 0.0905, 0.0781, 0.0581, 0.0611, 0.0614, 0.0724], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:43:10,106 INFO [train.py:904] (5/8) Epoch 16, batch 3100, loss[loss=0.1744, simple_loss=0.2451, pruned_loss=0.05186, over 16769.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2625, pruned_loss=0.04627, over 3319347.17 frames. ], batch size: 124, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:43:42,806 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155376.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:44:10,968 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4972, 3.7048, 4.0910, 2.2796, 3.2640, 2.5661, 3.9753, 3.8781], device='cuda:5'), covar=tensor([0.0241, 0.0741, 0.0429, 0.1803, 0.0754, 0.0867, 0.0548, 0.0981], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0157, 0.0161, 0.0148, 0.0140, 0.0126, 0.0141, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 07:44:17,694 INFO [train.py:904] (5/8) Epoch 16, batch 3150, loss[loss=0.1643, simple_loss=0.2461, pruned_loss=0.04128, over 15952.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2608, pruned_loss=0.04558, over 3323236.46 frames. ], batch size: 35, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:44:39,890 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.373e+02 2.704e+02 3.278e+02 4.749e+02, threshold=5.408e+02, percent-clipped=0.0 2023-04-30 07:44:47,893 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:44:49,096 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:45:00,597 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 07:45:27,257 INFO [train.py:904] (5/8) Epoch 16, batch 3200, loss[loss=0.1548, simple_loss=0.2386, pruned_loss=0.03552, over 15914.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2612, pruned_loss=0.04599, over 3314939.25 frames. ], batch size: 35, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:45:34,187 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:46:32,204 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5640, 4.9897, 4.4888, 4.8146, 4.5410, 4.4857, 4.5268, 5.0420], device='cuda:5'), covar=tensor([0.2412, 0.1596, 0.2470, 0.1432, 0.1686, 0.2102, 0.2246, 0.1951], device='cuda:5'), in_proj_covar=tensor([0.0649, 0.0803, 0.0650, 0.0579, 0.0504, 0.0512, 0.0668, 0.0616], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:46:36,142 INFO [train.py:904] (5/8) Epoch 16, batch 3250, loss[loss=0.1601, simple_loss=0.2583, pruned_loss=0.03098, over 17125.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2614, pruned_loss=0.0461, over 3322597.06 frames. ], batch size: 49, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:46:58,465 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.206e+02 2.546e+02 3.042e+02 5.764e+02, threshold=5.092e+02, percent-clipped=1.0 2023-04-30 07:46:58,932 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:47:45,926 INFO [train.py:904] (5/8) Epoch 16, batch 3300, loss[loss=0.1569, simple_loss=0.2557, pruned_loss=0.02907, over 17126.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2621, pruned_loss=0.04607, over 3332550.50 frames. ], batch size: 47, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:48:21,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7197, 2.8837, 2.9113, 4.9935, 4.1188, 4.5524, 1.6092, 3.1624], device='cuda:5'), covar=tensor([0.1399, 0.0776, 0.1021, 0.0176, 0.0227, 0.0349, 0.1628, 0.0771], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0167, 0.0188, 0.0175, 0.0202, 0.0214, 0.0189, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 07:48:56,955 INFO [train.py:904] (5/8) Epoch 16, batch 3350, loss[loss=0.1474, simple_loss=0.2313, pruned_loss=0.03176, over 16734.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2624, pruned_loss=0.046, over 3332367.53 frames. ], batch size: 39, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:49:01,203 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155605.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:49:19,484 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.838e+02 2.398e+02 2.739e+02 3.162e+02 6.710e+02, threshold=5.477e+02, percent-clipped=3.0 2023-04-30 07:49:39,770 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:49:58,962 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8250, 4.0223, 2.5479, 4.5551, 3.0923, 4.5534, 2.7261, 3.2173], device='cuda:5'), covar=tensor([0.0272, 0.0362, 0.1443, 0.0210, 0.0782, 0.0388, 0.1257, 0.0627], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0175, 0.0194, 0.0156, 0.0174, 0.0218, 0.0202, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 07:50:08,454 INFO [train.py:904] (5/8) Epoch 16, batch 3400, loss[loss=0.1459, simple_loss=0.2327, pruned_loss=0.02953, over 15898.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2616, pruned_loss=0.04542, over 3332539.08 frames. ], batch size: 35, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:50:16,700 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155657.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:27,060 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2023-04-30 07:50:28,090 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:45,713 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 07:50:49,342 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:50,803 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 07:51:10,365 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0138, 5.4075, 5.1451, 5.2198, 4.8990, 4.8612, 4.8909, 5.5227], device='cuda:5'), covar=tensor([0.1301, 0.0851, 0.1018, 0.0772, 0.0905, 0.0865, 0.1003, 0.0916], device='cuda:5'), in_proj_covar=tensor([0.0643, 0.0798, 0.0644, 0.0575, 0.0499, 0.0508, 0.0661, 0.0610], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:51:19,271 INFO [train.py:904] (5/8) Epoch 16, batch 3450, loss[loss=0.1859, simple_loss=0.2813, pruned_loss=0.04526, over 17026.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2611, pruned_loss=0.04509, over 3321892.63 frames. ], batch size: 50, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:51:41,377 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.296e+02 2.647e+02 3.295e+02 8.789e+02, threshold=5.295e+02, percent-clipped=1.0 2023-04-30 07:51:41,697 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1277, 5.5402, 5.6974, 5.4559, 5.5158, 6.1245, 5.6805, 5.4122], device='cuda:5'), covar=tensor([0.0947, 0.2161, 0.2509, 0.2097, 0.2829, 0.1028, 0.1411, 0.2238], device='cuda:5'), in_proj_covar=tensor([0.0401, 0.0575, 0.0620, 0.0483, 0.0645, 0.0650, 0.0490, 0.0641], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 07:51:41,865 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155718.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:51:52,579 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:52:22,826 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 07:52:24,297 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2266, 5.6963, 5.8286, 5.6036, 5.6458, 6.1941, 5.7356, 5.4475], device='cuda:5'), covar=tensor([0.0886, 0.2115, 0.2243, 0.1946, 0.2503, 0.0895, 0.1589, 0.2511], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0574, 0.0619, 0.0483, 0.0644, 0.0649, 0.0489, 0.0640], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 07:52:28,787 INFO [train.py:904] (5/8) Epoch 16, batch 3500, loss[loss=0.1623, simple_loss=0.2539, pruned_loss=0.03531, over 17224.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2596, pruned_loss=0.04404, over 3329850.58 frames. ], batch size: 46, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:52:52,151 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 07:52:58,254 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:14,818 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:19,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9193, 1.9224, 2.2873, 3.4806, 1.9952, 2.1515, 2.0920, 2.0351], device='cuda:5'), covar=tensor([0.1529, 0.4082, 0.2636, 0.0702, 0.4607, 0.2990, 0.3664, 0.4052], device='cuda:5'), in_proj_covar=tensor([0.0389, 0.0426, 0.0355, 0.0327, 0.0428, 0.0491, 0.0394, 0.0498], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:53:40,590 INFO [train.py:904] (5/8) Epoch 16, batch 3550, loss[loss=0.1813, simple_loss=0.2582, pruned_loss=0.05216, over 16889.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2576, pruned_loss=0.04399, over 3316290.49 frames. ], batch size: 109, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:53:56,153 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:54:04,796 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.310e+02 2.739e+02 3.224e+02 5.856e+02, threshold=5.477e+02, percent-clipped=2.0 2023-04-30 07:54:42,441 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155845.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:54:52,023 INFO [train.py:904] (5/8) Epoch 16, batch 3600, loss[loss=0.1678, simple_loss=0.2357, pruned_loss=0.04994, over 16794.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2567, pruned_loss=0.04426, over 3305468.84 frames. ], batch size: 83, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:55:02,356 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-30 07:55:12,383 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4380, 5.8115, 5.5544, 5.6147, 5.2899, 5.2339, 5.2152, 5.9439], device='cuda:5'), covar=tensor([0.1232, 0.0837, 0.0974, 0.0750, 0.0829, 0.0645, 0.1109, 0.0940], device='cuda:5'), in_proj_covar=tensor([0.0642, 0.0799, 0.0644, 0.0575, 0.0499, 0.0508, 0.0662, 0.0611], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:56:03,898 INFO [train.py:904] (5/8) Epoch 16, batch 3650, loss[loss=0.1748, simple_loss=0.248, pruned_loss=0.0508, over 16887.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.256, pruned_loss=0.04523, over 3292291.66 frames. ], batch size: 96, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:56:28,569 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.258e+02 2.651e+02 3.350e+02 6.145e+02, threshold=5.302e+02, percent-clipped=1.0 2023-04-30 07:56:43,918 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0420, 3.9198, 4.0948, 4.1980, 4.2622, 3.8424, 4.0839, 4.2868], device='cuda:5'), covar=tensor([0.1371, 0.1034, 0.1162, 0.0649, 0.0645, 0.1570, 0.2019, 0.0659], device='cuda:5'), in_proj_covar=tensor([0.0625, 0.0777, 0.0916, 0.0790, 0.0587, 0.0616, 0.0621, 0.0729], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:57:17,743 INFO [train.py:904] (5/8) Epoch 16, batch 3700, loss[loss=0.1798, simple_loss=0.2495, pruned_loss=0.05506, over 16737.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2552, pruned_loss=0.04668, over 3270547.92 frames. ], batch size: 134, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:57:32,577 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:57:41,819 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6042, 4.5710, 4.5977, 4.1530, 4.5756, 1.8036, 4.3775, 4.3511], device='cuda:5'), covar=tensor([0.0112, 0.0093, 0.0154, 0.0267, 0.0088, 0.2532, 0.0142, 0.0169], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0144, 0.0191, 0.0176, 0.0164, 0.0202, 0.0180, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 07:58:18,426 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-30 07:58:36,932 INFO [train.py:904] (5/8) Epoch 16, batch 3750, loss[loss=0.1837, simple_loss=0.2509, pruned_loss=0.05827, over 16902.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2558, pruned_loss=0.04772, over 3267879.24 frames. ], batch size: 109, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:58:53,071 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:59:02,684 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.313e+02 2.777e+02 3.166e+02 7.362e+02, threshold=5.555e+02, percent-clipped=3.0 2023-04-30 07:59:31,662 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 07:59:51,248 INFO [train.py:904] (5/8) Epoch 16, batch 3800, loss[loss=0.2027, simple_loss=0.2719, pruned_loss=0.06678, over 16260.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2579, pruned_loss=0.04938, over 3257152.71 frames. ], batch size: 165, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 08:00:17,863 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156070.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:03,123 INFO [train.py:904] (5/8) Epoch 16, batch 3850, loss[loss=0.1917, simple_loss=0.26, pruned_loss=0.06172, over 16756.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.258, pruned_loss=0.05011, over 3258765.76 frames. ], batch size: 134, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:01:20,129 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156113.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:28,705 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.283e+02 2.734e+02 3.408e+02 7.535e+02, threshold=5.468e+02, percent-clipped=3.0 2023-04-30 08:01:47,413 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156131.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:01:59,099 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156140.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:02:16,231 INFO [train.py:904] (5/8) Epoch 16, batch 3900, loss[loss=0.1597, simple_loss=0.2359, pruned_loss=0.04179, over 16786.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2576, pruned_loss=0.05063, over 3267855.67 frames. ], batch size: 83, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:02:29,442 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:15,689 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-30 08:03:26,074 INFO [train.py:904] (5/8) Epoch 16, batch 3950, loss[loss=0.1915, simple_loss=0.2696, pruned_loss=0.05671, over 17124.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2576, pruned_loss=0.05132, over 3275577.91 frames. ], batch size: 48, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:03:47,392 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:50,344 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.349e+02 2.934e+02 3.426e+02 7.028e+02, threshold=5.868e+02, percent-clipped=1.0 2023-04-30 08:03:55,578 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-30 08:04:16,489 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-30 08:04:37,103 INFO [train.py:904] (5/8) Epoch 16, batch 4000, loss[loss=0.1571, simple_loss=0.2441, pruned_loss=0.03501, over 16795.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2573, pruned_loss=0.05105, over 3272079.71 frames. ], batch size: 102, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:04:49,896 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:12,231 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:13,491 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3523, 2.2428, 1.7525, 1.9365, 2.5414, 2.2851, 2.4839, 2.6954], device='cuda:5'), covar=tensor([0.0226, 0.0341, 0.0492, 0.0453, 0.0205, 0.0315, 0.0191, 0.0254], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0222, 0.0213, 0.0214, 0.0223, 0.0224, 0.0229, 0.0221], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:05:31,859 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:49,252 INFO [train.py:904] (5/8) Epoch 16, batch 4050, loss[loss=0.1782, simple_loss=0.2621, pruned_loss=0.04713, over 16832.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.258, pruned_loss=0.05024, over 3258291.64 frames. ], batch size: 83, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:05:59,639 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:04,441 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:14,230 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.383e+02 1.898e+02 2.203e+02 2.661e+02 4.268e+02, threshold=4.405e+02, percent-clipped=0.0 2023-04-30 08:07:01,335 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:01,992 INFO [train.py:904] (5/8) Epoch 16, batch 4100, loss[loss=0.1753, simple_loss=0.262, pruned_loss=0.04427, over 16699.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2594, pruned_loss=0.04984, over 3253611.83 frames. ], batch size: 76, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:07:15,200 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:36,933 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 08:08:15,976 INFO [train.py:904] (5/8) Epoch 16, batch 4150, loss[loss=0.1857, simple_loss=0.2796, pruned_loss=0.04592, over 16712.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2661, pruned_loss=0.05204, over 3222009.06 frames. ], batch size: 89, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:08:40,345 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.118e+02 2.582e+02 3.252e+02 7.284e+02, threshold=5.164e+02, percent-clipped=7.0 2023-04-30 08:08:52,197 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156426.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:09:13,297 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156440.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:09:30,685 INFO [train.py:904] (5/8) Epoch 16, batch 4200, loss[loss=0.203, simple_loss=0.3019, pruned_loss=0.05209, over 16527.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2735, pruned_loss=0.05372, over 3197591.46 frames. ], batch size: 146, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:10:24,405 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156488.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:10:35,444 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2349, 2.1825, 2.1734, 3.9727, 2.1084, 2.5009, 2.2434, 2.3696], device='cuda:5'), covar=tensor([0.1174, 0.3561, 0.2687, 0.0438, 0.3961, 0.2371, 0.3427, 0.3108], device='cuda:5'), in_proj_covar=tensor([0.0383, 0.0423, 0.0352, 0.0324, 0.0423, 0.0488, 0.0390, 0.0494], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:10:37,794 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8359, 3.9496, 2.2428, 4.5844, 2.9574, 4.5620, 2.6734, 3.1284], device='cuda:5'), covar=tensor([0.0236, 0.0312, 0.1725, 0.0243, 0.0804, 0.0390, 0.1316, 0.0698], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0173, 0.0191, 0.0151, 0.0173, 0.0213, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 08:10:38,788 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3283, 3.5080, 3.6256, 3.6053, 3.6296, 3.4720, 3.4438, 3.4798], device='cuda:5'), covar=tensor([0.0371, 0.0540, 0.0488, 0.0485, 0.0491, 0.0488, 0.0904, 0.0564], device='cuda:5'), in_proj_covar=tensor([0.0380, 0.0409, 0.0398, 0.0376, 0.0452, 0.0425, 0.0518, 0.0336], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 08:10:43,274 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4054, 4.3233, 4.3368, 2.9578, 3.6635, 4.2916, 3.7994, 2.5026], device='cuda:5'), covar=tensor([0.0468, 0.0029, 0.0044, 0.0295, 0.0077, 0.0085, 0.0074, 0.0363], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0075, 0.0076, 0.0130, 0.0089, 0.0099, 0.0088, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 08:10:43,962 INFO [train.py:904] (5/8) Epoch 16, batch 4250, loss[loss=0.1905, simple_loss=0.2825, pruned_loss=0.04921, over 16678.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2768, pruned_loss=0.05375, over 3181189.48 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:11:09,165 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.212e+02 2.674e+02 3.089e+02 6.979e+02, threshold=5.347e+02, percent-clipped=2.0 2023-04-30 08:11:56,451 INFO [train.py:904] (5/8) Epoch 16, batch 4300, loss[loss=0.2036, simple_loss=0.2984, pruned_loss=0.05434, over 16638.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2785, pruned_loss=0.05305, over 3174390.78 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:12:27,178 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:12:41,079 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1757, 1.9661, 2.5798, 3.1970, 2.8183, 3.5664, 2.0647, 3.5430], device='cuda:5'), covar=tensor([0.0154, 0.0488, 0.0289, 0.0214, 0.0286, 0.0131, 0.0470, 0.0092], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0185, 0.0172, 0.0175, 0.0187, 0.0142, 0.0184, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:13:09,409 INFO [train.py:904] (5/8) Epoch 16, batch 4350, loss[loss=0.197, simple_loss=0.2865, pruned_loss=0.05377, over 15218.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2819, pruned_loss=0.05408, over 3170493.47 frames. ], batch size: 190, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:13:22,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 08:13:34,597 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 2.270e+02 2.623e+02 3.147e+02 7.539e+02, threshold=5.245e+02, percent-clipped=2.0 2023-04-30 08:13:56,912 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3062, 4.4130, 4.2383, 3.9779, 3.9202, 4.3347, 3.9815, 4.0419], device='cuda:5'), covar=tensor([0.0533, 0.0294, 0.0250, 0.0234, 0.0727, 0.0295, 0.0687, 0.0549], device='cuda:5'), in_proj_covar=tensor([0.0271, 0.0376, 0.0320, 0.0309, 0.0335, 0.0356, 0.0217, 0.0383], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:14:14,564 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156646.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:14:22,116 INFO [train.py:904] (5/8) Epoch 16, batch 4400, loss[loss=0.2024, simple_loss=0.2895, pruned_loss=0.05766, over 16878.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2842, pruned_loss=0.05491, over 3178344.63 frames. ], batch size: 109, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:14:44,141 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2852, 5.2546, 5.0362, 4.2847, 5.2697, 1.5885, 4.9211, 4.6746], device='cuda:5'), covar=tensor([0.0047, 0.0037, 0.0125, 0.0337, 0.0043, 0.3039, 0.0079, 0.0204], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0139, 0.0186, 0.0172, 0.0159, 0.0198, 0.0175, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:15:06,839 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6892, 1.6879, 2.2744, 2.6368, 2.5140, 3.0748, 1.7715, 2.9965], device='cuda:5'), covar=tensor([0.0169, 0.0485, 0.0279, 0.0261, 0.0278, 0.0122, 0.0481, 0.0106], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0174, 0.0185, 0.0140, 0.0183, 0.0134], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:15:32,101 INFO [train.py:904] (5/8) Epoch 16, batch 4450, loss[loss=0.2193, simple_loss=0.3076, pruned_loss=0.06554, over 17104.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2873, pruned_loss=0.056, over 3190496.07 frames. ], batch size: 53, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:15:57,565 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 2.034e+02 2.328e+02 2.907e+02 4.699e+02, threshold=4.656e+02, percent-clipped=0.0 2023-04-30 08:16:04,255 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5959, 3.6762, 2.1918, 4.2792, 2.7710, 4.2181, 2.3678, 2.8574], device='cuda:5'), covar=tensor([0.0272, 0.0368, 0.1718, 0.0118, 0.0850, 0.0421, 0.1498, 0.0808], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0172, 0.0190, 0.0149, 0.0172, 0.0212, 0.0199, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 08:16:09,709 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156726.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:16:41,452 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:16:45,647 INFO [train.py:904] (5/8) Epoch 16, batch 4500, loss[loss=0.228, simple_loss=0.2973, pruned_loss=0.07933, over 11759.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2871, pruned_loss=0.05657, over 3186604.15 frames. ], batch size: 246, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:16:47,444 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4595, 2.1478, 1.7788, 1.8748, 2.4553, 2.0753, 2.3972, 2.6306], device='cuda:5'), covar=tensor([0.0138, 0.0317, 0.0438, 0.0399, 0.0195, 0.0338, 0.0175, 0.0207], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0217, 0.0210, 0.0211, 0.0220, 0.0221, 0.0224, 0.0217], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:16:49,149 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-30 08:17:16,194 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:17:19,762 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 08:17:54,760 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7900, 2.8919, 2.7672, 4.7888, 3.7830, 4.1204, 1.6349, 3.1817], device='cuda:5'), covar=tensor([0.1224, 0.0695, 0.1060, 0.0124, 0.0322, 0.0355, 0.1519, 0.0739], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0168, 0.0187, 0.0172, 0.0203, 0.0213, 0.0190, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 08:17:56,559 INFO [train.py:904] (5/8) Epoch 16, batch 4550, loss[loss=0.2117, simple_loss=0.2938, pruned_loss=0.06479, over 16626.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2879, pruned_loss=0.0574, over 3203521.76 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:18:07,904 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156810.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:18:16,504 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7026, 3.9910, 2.8807, 2.3543, 2.7638, 2.3447, 4.2921, 3.6114], device='cuda:5'), covar=tensor([0.2626, 0.0592, 0.1674, 0.2353, 0.2379, 0.1941, 0.0413, 0.0992], device='cuda:5'), in_proj_covar=tensor([0.0315, 0.0264, 0.0295, 0.0295, 0.0291, 0.0240, 0.0282, 0.0320], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 08:18:19,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.005e+02 2.274e+02 2.770e+02 4.941e+02, threshold=4.548e+02, percent-clipped=1.0 2023-04-30 08:19:06,329 INFO [train.py:904] (5/8) Epoch 16, batch 4600, loss[loss=0.1743, simple_loss=0.2706, pruned_loss=0.03896, over 16725.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2889, pruned_loss=0.05721, over 3216914.09 frames. ], batch size: 89, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:19:36,311 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:19:55,778 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156885.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:20:19,276 INFO [train.py:904] (5/8) Epoch 16, batch 4650, loss[loss=0.1665, simple_loss=0.2592, pruned_loss=0.03687, over 16865.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2876, pruned_loss=0.05701, over 3210368.54 frames. ], batch size: 96, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:20:37,664 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 08:20:39,034 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 08:20:45,040 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 1.893e+02 2.205e+02 2.654e+02 4.694e+02, threshold=4.410e+02, percent-clipped=1.0 2023-04-30 08:20:46,979 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:20:48,403 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4319, 3.4102, 3.6677, 1.7243, 3.9121, 3.9375, 2.9372, 2.8323], device='cuda:5'), covar=tensor([0.0886, 0.0250, 0.0259, 0.1348, 0.0087, 0.0131, 0.0487, 0.0518], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0106, 0.0093, 0.0139, 0.0074, 0.0119, 0.0126, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 08:21:12,290 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-30 08:21:25,718 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:25,796 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:33,080 INFO [train.py:904] (5/8) Epoch 16, batch 4700, loss[loss=0.1794, simple_loss=0.2681, pruned_loss=0.0453, over 16490.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2851, pruned_loss=0.05586, over 3201316.29 frames. ], batch size: 75, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:36,836 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156994.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:22:47,498 INFO [train.py:904] (5/8) Epoch 16, batch 4750, loss[loss=0.187, simple_loss=0.2638, pruned_loss=0.05509, over 16572.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2816, pruned_loss=0.05431, over 3198602.97 frames. ], batch size: 62, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:59,882 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157010.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:23:11,424 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 1.885e+02 2.255e+02 2.719e+02 5.243e+02, threshold=4.511e+02, percent-clipped=2.0 2023-04-30 08:23:59,616 INFO [train.py:904] (5/8) Epoch 16, batch 4800, loss[loss=0.1892, simple_loss=0.2815, pruned_loss=0.0484, over 16693.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2782, pruned_loss=0.05236, over 3211504.47 frames. ], batch size: 83, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:24:21,366 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 08:24:29,030 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157071.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:00,090 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 08:25:14,793 INFO [train.py:904] (5/8) Epoch 16, batch 4850, loss[loss=0.18, simple_loss=0.2754, pruned_loss=0.04232, over 16797.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2784, pruned_loss=0.05156, over 3194717.78 frames. ], batch size: 83, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:25:20,720 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:41,924 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.885e+02 2.190e+02 2.625e+02 3.825e+02, threshold=4.379e+02, percent-clipped=0.0 2023-04-30 08:26:05,784 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3254, 2.5455, 2.1406, 2.1550, 2.8852, 2.5434, 3.0022, 3.1494], device='cuda:5'), covar=tensor([0.0095, 0.0363, 0.0437, 0.0434, 0.0226, 0.0350, 0.0193, 0.0206], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0217, 0.0211, 0.0210, 0.0220, 0.0220, 0.0222, 0.0215], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:26:17,013 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:26:30,918 INFO [train.py:904] (5/8) Epoch 16, batch 4900, loss[loss=0.1666, simple_loss=0.2591, pruned_loss=0.03707, over 16594.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2776, pruned_loss=0.05043, over 3175577.49 frames. ], batch size: 76, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:21,888 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9869, 5.4429, 5.6739, 5.3445, 5.4576, 6.0080, 5.5574, 5.2844], device='cuda:5'), covar=tensor([0.0862, 0.1605, 0.1802, 0.1807, 0.2210, 0.0846, 0.1114, 0.2126], device='cuda:5'), in_proj_covar=tensor([0.0380, 0.0539, 0.0581, 0.0455, 0.0607, 0.0617, 0.0460, 0.0609], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 08:27:37,717 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1528, 3.4560, 3.4646, 2.1061, 3.0017, 2.3879, 3.5598, 3.6711], device='cuda:5'), covar=tensor([0.0233, 0.0704, 0.0584, 0.1804, 0.0815, 0.0851, 0.0626, 0.0871], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0156, 0.0161, 0.0148, 0.0139, 0.0125, 0.0141, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 08:27:45,030 INFO [train.py:904] (5/8) Epoch 16, batch 4950, loss[loss=0.1817, simple_loss=0.2811, pruned_loss=0.04119, over 16845.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2764, pruned_loss=0.04936, over 3196834.05 frames. ], batch size: 83, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:46,887 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157203.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:28:08,809 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.143e+02 2.414e+02 2.690e+02 6.197e+02, threshold=4.827e+02, percent-clipped=2.0 2023-04-30 08:28:24,992 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-30 08:28:41,078 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:28:57,719 INFO [train.py:904] (5/8) Epoch 16, batch 5000, loss[loss=0.2039, simple_loss=0.2888, pruned_loss=0.05949, over 12027.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2779, pruned_loss=0.04949, over 3199706.34 frames. ], batch size: 248, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:30:10,083 INFO [train.py:904] (5/8) Epoch 16, batch 5050, loss[loss=0.1831, simple_loss=0.2755, pruned_loss=0.0453, over 16804.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2793, pruned_loss=0.04977, over 3189067.52 frames. ], batch size: 102, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:30:10,654 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4240, 3.8410, 3.8206, 2.1641, 3.1732, 2.5632, 3.7576, 3.9419], device='cuda:5'), covar=tensor([0.0232, 0.0660, 0.0539, 0.1840, 0.0793, 0.0852, 0.0650, 0.0828], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0156, 0.0161, 0.0148, 0.0139, 0.0125, 0.0140, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 08:30:33,957 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.188e+02 2.518e+02 3.068e+02 6.843e+02, threshold=5.037e+02, percent-clipped=3.0 2023-04-30 08:31:22,549 INFO [train.py:904] (5/8) Epoch 16, batch 5100, loss[loss=0.1654, simple_loss=0.2449, pruned_loss=0.04298, over 16654.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2768, pruned_loss=0.04852, over 3216643.22 frames. ], batch size: 62, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:31:29,238 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157356.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:31:44,295 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:07,496 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:38,193 INFO [train.py:904] (5/8) Epoch 16, batch 5150, loss[loss=0.204, simple_loss=0.2994, pruned_loss=0.0543, over 15373.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2771, pruned_loss=0.04809, over 3216914.99 frames. ], batch size: 191, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:32:42,310 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157404.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:32:43,942 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157405.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:02,262 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157417.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:04,665 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.981e+02 2.319e+02 2.667e+02 4.266e+02, threshold=4.638e+02, percent-clipped=0.0 2023-04-30 08:33:05,823 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4185, 3.2461, 2.6095, 2.0590, 2.2140, 2.2178, 3.2436, 3.0075], device='cuda:5'), covar=tensor([0.2628, 0.0665, 0.1785, 0.2786, 0.2214, 0.1959, 0.0602, 0.1252], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0261, 0.0292, 0.0293, 0.0286, 0.0237, 0.0279, 0.0315], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 08:33:38,052 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.62 vs. limit=5.0 2023-04-30 08:33:40,829 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157443.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:53,092 INFO [train.py:904] (5/8) Epoch 16, batch 5200, loss[loss=0.2224, simple_loss=0.3085, pruned_loss=0.06811, over 15322.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2764, pruned_loss=0.04792, over 3207434.87 frames. ], batch size: 190, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:33:54,777 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157453.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:34:03,382 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-30 08:34:12,620 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157465.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:35:01,946 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:35:08,194 INFO [train.py:904] (5/8) Epoch 16, batch 5250, loss[loss=0.1989, simple_loss=0.2774, pruned_loss=0.06026, over 12540.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2733, pruned_loss=0.04733, over 3216425.87 frames. ], batch size: 247, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:35:14,360 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157506.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:35:32,957 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.109e+02 2.306e+02 2.689e+02 3.834e+02, threshold=4.611e+02, percent-clipped=0.0 2023-04-30 08:36:06,357 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157541.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:36:21,506 INFO [train.py:904] (5/8) Epoch 16, batch 5300, loss[loss=0.1724, simple_loss=0.2581, pruned_loss=0.0433, over 16632.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2693, pruned_loss=0.04603, over 3235215.32 frames. ], batch size: 57, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:36:45,030 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157567.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:36:46,165 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:37:17,230 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:37:34,906 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0430, 2.4922, 2.6487, 1.8498, 2.6604, 2.8585, 2.4501, 2.4153], device='cuda:5'), covar=tensor([0.0732, 0.0233, 0.0207, 0.1033, 0.0119, 0.0213, 0.0428, 0.0454], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0105, 0.0091, 0.0137, 0.0073, 0.0116, 0.0123, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 08:37:35,576 INFO [train.py:904] (5/8) Epoch 16, batch 5350, loss[loss=0.2167, simple_loss=0.2827, pruned_loss=0.07537, over 12344.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2691, pruned_loss=0.04589, over 3231874.71 frames. ], batch size: 246, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:38:00,214 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 1.990e+02 2.338e+02 2.844e+02 4.721e+02, threshold=4.676e+02, percent-clipped=1.0 2023-04-30 08:38:02,580 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4644, 2.2131, 1.9132, 1.9347, 2.4871, 2.1902, 2.4094, 2.6947], device='cuda:5'), covar=tensor([0.0155, 0.0375, 0.0453, 0.0419, 0.0219, 0.0318, 0.0173, 0.0232], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0220, 0.0213, 0.0213, 0.0221, 0.0223, 0.0224, 0.0216], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:38:16,618 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:38:49,044 INFO [train.py:904] (5/8) Epoch 16, batch 5400, loss[loss=0.186, simple_loss=0.2787, pruned_loss=0.04661, over 17120.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2716, pruned_loss=0.04659, over 3226916.03 frames. ], batch size: 47, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:38:50,607 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6078, 4.7707, 4.9052, 4.7301, 4.6972, 5.3365, 4.8460, 4.6044], device='cuda:5'), covar=tensor([0.1120, 0.1445, 0.1740, 0.1741, 0.2436, 0.0843, 0.1299, 0.2223], device='cuda:5'), in_proj_covar=tensor([0.0384, 0.0539, 0.0582, 0.0452, 0.0609, 0.0619, 0.0460, 0.0610], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 08:39:01,214 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8662, 3.7784, 4.4682, 2.0581, 4.6884, 4.6853, 3.1324, 3.2753], device='cuda:5'), covar=tensor([0.0717, 0.0241, 0.0133, 0.1141, 0.0041, 0.0099, 0.0395, 0.0427], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0104, 0.0090, 0.0136, 0.0072, 0.0115, 0.0123, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 08:39:09,613 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:02,202 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 08:40:04,943 INFO [train.py:904] (5/8) Epoch 16, batch 5450, loss[loss=0.2233, simple_loss=0.2958, pruned_loss=0.0754, over 11915.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2745, pruned_loss=0.04837, over 3205618.90 frames. ], batch size: 248, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:40:20,463 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:23,499 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157714.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:29,845 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.390e+02 2.375e+02 2.743e+02 3.382e+02 6.991e+02, threshold=5.486e+02, percent-clipped=5.0 2023-04-30 08:40:57,854 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 08:41:00,228 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:41:20,979 INFO [train.py:904] (5/8) Epoch 16, batch 5500, loss[loss=0.207, simple_loss=0.3003, pruned_loss=0.05682, over 16709.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2817, pruned_loss=0.05306, over 3171349.50 frames. ], batch size: 89, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:41:25,671 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 08:41:30,798 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9780, 1.8862, 2.6036, 2.8234, 2.7578, 3.3550, 2.0638, 3.3844], device='cuda:5'), covar=tensor([0.0184, 0.0445, 0.0269, 0.0272, 0.0263, 0.0122, 0.0439, 0.0098], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0187, 0.0172, 0.0176, 0.0187, 0.0142, 0.0186, 0.0136], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:41:33,693 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157760.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 08:42:30,918 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157798.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:42:36,603 INFO [train.py:904] (5/8) Epoch 16, batch 5550, loss[loss=0.2187, simple_loss=0.3068, pruned_loss=0.06532, over 16528.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2892, pruned_loss=0.05855, over 3139743.50 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:42:46,037 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1536, 3.2208, 1.7732, 3.4143, 2.3817, 3.4653, 2.0453, 2.5486], device='cuda:5'), covar=tensor([0.0291, 0.0403, 0.1758, 0.0198, 0.0837, 0.0572, 0.1569, 0.0761], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0172, 0.0192, 0.0148, 0.0172, 0.0210, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 08:43:04,452 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.202e+02 3.935e+02 4.988e+02 8.755e+02, threshold=7.870e+02, percent-clipped=18.0 2023-04-30 08:43:47,866 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157846.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:43:57,048 INFO [train.py:904] (5/8) Epoch 16, batch 5600, loss[loss=0.2321, simple_loss=0.3115, pruned_loss=0.0764, over 16389.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2949, pruned_loss=0.06361, over 3096278.34 frames. ], batch size: 146, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:44:00,068 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:44:14,254 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157862.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 08:44:33,449 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4847, 4.4952, 4.9119, 4.8647, 4.9018, 4.5783, 4.5283, 4.4181], device='cuda:5'), covar=tensor([0.0310, 0.0594, 0.0369, 0.0430, 0.0477, 0.0407, 0.1026, 0.0536], device='cuda:5'), in_proj_covar=tensor([0.0374, 0.0401, 0.0391, 0.0371, 0.0442, 0.0417, 0.0513, 0.0331], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 08:45:21,488 INFO [train.py:904] (5/8) Epoch 16, batch 5650, loss[loss=0.2449, simple_loss=0.3213, pruned_loss=0.08429, over 16347.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3, pruned_loss=0.06798, over 3065871.16 frames. ], batch size: 146, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:45:36,755 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 08:45:41,569 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157914.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:45:50,674 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.188e+02 3.549e+02 4.142e+02 4.836e+02 1.033e+03, threshold=8.283e+02, percent-clipped=2.0 2023-04-30 08:45:56,958 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:46:39,310 INFO [train.py:904] (5/8) Epoch 16, batch 5700, loss[loss=0.2079, simple_loss=0.2944, pruned_loss=0.06072, over 16703.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3013, pruned_loss=0.06948, over 3062336.31 frames. ], batch size: 57, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:47:12,251 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 08:48:01,948 INFO [train.py:904] (5/8) Epoch 16, batch 5750, loss[loss=0.2096, simple_loss=0.2987, pruned_loss=0.06029, over 16657.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3041, pruned_loss=0.07073, over 3054552.98 frames. ], batch size: 76, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:48:17,327 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158012.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:48:30,522 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 3.002e+02 3.593e+02 4.729e+02 9.320e+02, threshold=7.185e+02, percent-clipped=2.0 2023-04-30 08:49:00,641 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:04,848 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9451, 1.8908, 2.1101, 3.3723, 1.8978, 2.1474, 2.0287, 1.9748], device='cuda:5'), covar=tensor([0.1446, 0.4368, 0.2954, 0.0751, 0.5253, 0.3011, 0.3891, 0.4151], device='cuda:5'), in_proj_covar=tensor([0.0376, 0.0417, 0.0347, 0.0317, 0.0420, 0.0481, 0.0385, 0.0486], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:49:22,780 INFO [train.py:904] (5/8) Epoch 16, batch 5800, loss[loss=0.1941, simple_loss=0.2933, pruned_loss=0.04749, over 16752.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3031, pruned_loss=0.06864, over 3063282.01 frames. ], batch size: 83, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:49:35,950 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:36,011 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158060.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:49:57,249 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9001, 5.1897, 4.9515, 4.9498, 4.7021, 4.6804, 4.6178, 5.2910], device='cuda:5'), covar=tensor([0.1119, 0.0802, 0.0919, 0.0810, 0.0728, 0.0882, 0.1102, 0.0797], device='cuda:5'), in_proj_covar=tensor([0.0600, 0.0740, 0.0608, 0.0539, 0.0465, 0.0473, 0.0616, 0.0565], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:50:16,665 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158086.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:50:41,303 INFO [train.py:904] (5/8) Epoch 16, batch 5850, loss[loss=0.238, simple_loss=0.3309, pruned_loss=0.07251, over 16238.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3019, pruned_loss=0.06773, over 3055159.49 frames. ], batch size: 165, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:50:51,681 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158108.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:51:08,531 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.851e+02 3.457e+02 4.318e+02 8.019e+02, threshold=6.915e+02, percent-clipped=1.0 2023-04-30 08:51:13,012 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158122.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:51:36,465 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 08:52:03,766 INFO [train.py:904] (5/8) Epoch 16, batch 5900, loss[loss=0.2095, simple_loss=0.3022, pruned_loss=0.0584, over 16298.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3018, pruned_loss=0.06726, over 3081666.24 frames. ], batch size: 35, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:52:24,462 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158162.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:52:56,047 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:25,270 INFO [train.py:904] (5/8) Epoch 16, batch 5950, loss[loss=0.2039, simple_loss=0.29, pruned_loss=0.05891, over 16770.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.3019, pruned_loss=0.06548, over 3099346.71 frames. ], batch size: 83, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:53:36,992 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158209.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:38,264 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158210.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:53:55,274 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.924e+02 3.226e+02 4.367e+02 1.058e+03, threshold=6.453e+02, percent-clipped=4.0 2023-04-30 08:54:01,906 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158224.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:54:45,444 INFO [train.py:904] (5/8) Epoch 16, batch 6000, loss[loss=0.1686, simple_loss=0.2612, pruned_loss=0.038, over 16866.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.3008, pruned_loss=0.06544, over 3075578.65 frames. ], batch size: 96, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:54:45,445 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 08:54:56,482 INFO [train.py:938] (5/8) Epoch 16, validation: loss=0.1553, simple_loss=0.2682, pruned_loss=0.0212, over 944034.00 frames. 2023-04-30 08:54:56,483 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 08:55:01,060 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 08:55:27,068 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:55:30,811 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4479, 4.2187, 4.1568, 2.8918, 3.6530, 4.1336, 3.8251, 2.3194], device='cuda:5'), covar=tensor([0.0499, 0.0034, 0.0042, 0.0333, 0.0090, 0.0094, 0.0072, 0.0411], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0074, 0.0075, 0.0128, 0.0088, 0.0097, 0.0086, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 08:55:54,124 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7990, 5.1255, 4.8538, 4.8461, 4.6554, 4.5908, 4.5088, 5.1982], device='cuda:5'), covar=tensor([0.1139, 0.0809, 0.0943, 0.0831, 0.0697, 0.0970, 0.1114, 0.0818], device='cuda:5'), in_proj_covar=tensor([0.0597, 0.0739, 0.0604, 0.0539, 0.0462, 0.0471, 0.0613, 0.0565], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:56:13,220 INFO [train.py:904] (5/8) Epoch 16, batch 6050, loss[loss=0.2496, simple_loss=0.3151, pruned_loss=0.09207, over 11593.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2994, pruned_loss=0.06479, over 3075881.16 frames. ], batch size: 248, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:56:40,249 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.935e+02 2.678e+02 3.324e+02 4.253e+02 8.310e+02, threshold=6.647e+02, percent-clipped=4.0 2023-04-30 08:56:57,676 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.8244, 6.2314, 5.7839, 6.0144, 5.5892, 5.4298, 5.6380, 6.2875], device='cuda:5'), covar=tensor([0.1219, 0.0742, 0.1055, 0.0823, 0.0710, 0.0560, 0.1119, 0.0837], device='cuda:5'), in_proj_covar=tensor([0.0603, 0.0744, 0.0610, 0.0544, 0.0465, 0.0475, 0.0618, 0.0569], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:57:24,857 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-30 08:57:31,996 INFO [train.py:904] (5/8) Epoch 16, batch 6100, loss[loss=0.1996, simple_loss=0.2901, pruned_loss=0.05456, over 16864.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2983, pruned_loss=0.06335, over 3099992.77 frames. ], batch size: 109, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:58:24,076 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:58:37,311 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 08:58:46,000 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6586, 3.9886, 3.5587, 3.8709, 3.5535, 3.6577, 3.6058, 3.9646], device='cuda:5'), covar=tensor([0.2329, 0.1637, 0.2753, 0.1527, 0.1788, 0.2761, 0.2288, 0.1890], device='cuda:5'), in_proj_covar=tensor([0.0598, 0.0738, 0.0604, 0.0540, 0.0462, 0.0472, 0.0613, 0.0566], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 08:58:49,721 INFO [train.py:904] (5/8) Epoch 16, batch 6150, loss[loss=0.1943, simple_loss=0.2826, pruned_loss=0.05302, over 16710.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2965, pruned_loss=0.06294, over 3083488.72 frames. ], batch size: 83, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:59:11,325 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8910, 3.3412, 3.2125, 2.0363, 2.8919, 2.2298, 3.4002, 3.4543], device='cuda:5'), covar=tensor([0.0266, 0.0692, 0.0620, 0.1903, 0.0812, 0.0965, 0.0618, 0.0877], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0155, 0.0161, 0.0148, 0.0139, 0.0125, 0.0140, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 08:59:18,319 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.901e+02 2.793e+02 3.279e+02 3.951e+02 7.811e+02, threshold=6.558e+02, percent-clipped=1.0 2023-04-30 08:59:57,777 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:00:08,005 INFO [train.py:904] (5/8) Epoch 16, batch 6200, loss[loss=0.1875, simple_loss=0.2782, pruned_loss=0.04841, over 16615.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2939, pruned_loss=0.06165, over 3100961.56 frames. ], batch size: 57, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:00:48,501 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158478.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:12,910 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 09:01:21,949 INFO [train.py:904] (5/8) Epoch 16, batch 6250, loss[loss=0.1917, simple_loss=0.2877, pruned_loss=0.04785, over 16729.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2935, pruned_loss=0.06143, over 3115809.47 frames. ], batch size: 76, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:01:34,532 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:44,911 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9050, 3.0741, 2.7845, 5.2108, 3.7285, 4.4467, 1.8980, 2.8825], device='cuda:5'), covar=tensor([0.1277, 0.0715, 0.1135, 0.0134, 0.0471, 0.0435, 0.1468, 0.0995], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0166, 0.0187, 0.0170, 0.0201, 0.0212, 0.0190, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:01:50,903 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 2.618e+02 3.109e+02 4.242e+02 1.066e+03, threshold=6.218e+02, percent-clipped=4.0 2023-04-30 09:02:20,288 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1240, 4.1836, 4.5094, 4.4735, 4.4963, 4.1826, 4.2375, 4.1012], device='cuda:5'), covar=tensor([0.0303, 0.0538, 0.0388, 0.0445, 0.0402, 0.0415, 0.0861, 0.0510], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0405, 0.0395, 0.0376, 0.0447, 0.0419, 0.0519, 0.0334], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 09:02:24,164 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0074, 4.0605, 3.8972, 3.6507, 3.6180, 4.0123, 3.6760, 3.7497], device='cuda:5'), covar=tensor([0.0672, 0.0618, 0.0295, 0.0284, 0.0773, 0.0490, 0.1010, 0.0660], device='cuda:5'), in_proj_covar=tensor([0.0266, 0.0370, 0.0312, 0.0300, 0.0326, 0.0350, 0.0215, 0.0371], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:02:38,759 INFO [train.py:904] (5/8) Epoch 16, batch 6300, loss[loss=0.1855, simple_loss=0.2817, pruned_loss=0.04459, over 16815.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2932, pruned_loss=0.0609, over 3119686.72 frames. ], batch size: 102, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:02:45,579 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:03:28,847 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 09:03:55,408 INFO [train.py:904] (5/8) Epoch 16, batch 6350, loss[loss=0.1879, simple_loss=0.2809, pruned_loss=0.04744, over 16848.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2941, pruned_loss=0.06256, over 3097416.03 frames. ], batch size: 102, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:04:16,688 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9729, 2.3922, 1.9183, 2.0603, 2.7689, 2.3982, 2.7997, 2.9650], device='cuda:5'), covar=tensor([0.0148, 0.0368, 0.0492, 0.0465, 0.0209, 0.0351, 0.0231, 0.0233], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0217, 0.0211, 0.0210, 0.0218, 0.0219, 0.0221, 0.0213], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:04:24,045 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 2.970e+02 3.577e+02 4.444e+02 9.031e+02, threshold=7.154e+02, percent-clipped=4.0 2023-04-30 09:04:46,153 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4034, 4.4273, 4.6029, 4.3994, 4.4773, 5.0126, 4.5504, 4.2879], device='cuda:5'), covar=tensor([0.1421, 0.1983, 0.2382, 0.2010, 0.2433, 0.1010, 0.1521, 0.2555], device='cuda:5'), in_proj_covar=tensor([0.0386, 0.0550, 0.0601, 0.0464, 0.0622, 0.0630, 0.0474, 0.0624], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:05:11,923 INFO [train.py:904] (5/8) Epoch 16, batch 6400, loss[loss=0.201, simple_loss=0.2908, pruned_loss=0.05556, over 16763.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2953, pruned_loss=0.06433, over 3081897.13 frames. ], batch size: 89, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:06:27,736 INFO [train.py:904] (5/8) Epoch 16, batch 6450, loss[loss=0.1997, simple_loss=0.2943, pruned_loss=0.05261, over 16310.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2955, pruned_loss=0.0639, over 3074926.28 frames. ], batch size: 35, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:06:56,389 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.821e+02 3.663e+02 4.597e+02 9.189e+02, threshold=7.326e+02, percent-clipped=6.0 2023-04-30 09:07:26,484 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:07:34,109 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0298, 4.0288, 3.9652, 3.2552, 3.9553, 1.8191, 3.7722, 3.5163], device='cuda:5'), covar=tensor([0.0104, 0.0087, 0.0158, 0.0281, 0.0082, 0.2525, 0.0118, 0.0212], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0137, 0.0184, 0.0171, 0.0156, 0.0195, 0.0170, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:07:43,547 INFO [train.py:904] (5/8) Epoch 16, batch 6500, loss[loss=0.1851, simple_loss=0.27, pruned_loss=0.05012, over 17106.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2933, pruned_loss=0.06346, over 3048471.12 frames. ], batch size: 47, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:08:21,101 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:22,812 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158778.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:24,837 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7955, 3.8757, 2.3771, 4.4126, 2.9589, 4.3083, 2.5373, 3.0704], device='cuda:5'), covar=tensor([0.0264, 0.0370, 0.1676, 0.0241, 0.0799, 0.0585, 0.1498, 0.0754], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0170, 0.0193, 0.0148, 0.0173, 0.0210, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:08:31,634 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:57,312 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5640, 2.6380, 2.3768, 4.0001, 2.7117, 3.9979, 1.3998, 2.8512], device='cuda:5'), covar=tensor([0.1407, 0.0763, 0.1290, 0.0139, 0.0216, 0.0404, 0.1723, 0.0806], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0166, 0.0188, 0.0170, 0.0202, 0.0212, 0.0191, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:09:01,984 INFO [train.py:904] (5/8) Epoch 16, batch 6550, loss[loss=0.1922, simple_loss=0.2965, pruned_loss=0.04397, over 16874.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2959, pruned_loss=0.06433, over 3056567.03 frames. ], batch size: 102, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:09:33,176 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.836e+02 3.301e+02 3.944e+02 7.483e+02, threshold=6.603e+02, percent-clipped=1.0 2023-04-30 09:09:39,691 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158826.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:09:53,288 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158835.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:09:58,244 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158838.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:06,344 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:18,022 INFO [train.py:904] (5/8) Epoch 16, batch 6600, loss[loss=0.2044, simple_loss=0.3003, pruned_loss=0.05422, over 16908.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.298, pruned_loss=0.06437, over 3057333.81 frames. ], batch size: 90, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:10:52,638 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9059, 4.0963, 3.0538, 2.4414, 2.8830, 2.5736, 4.4892, 3.7479], device='cuda:5'), covar=tensor([0.2676, 0.0698, 0.1812, 0.2481, 0.2566, 0.1879, 0.0452, 0.1076], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0261, 0.0294, 0.0296, 0.0287, 0.0239, 0.0281, 0.0315], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:11:25,981 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158896.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 09:11:34,214 INFO [train.py:904] (5/8) Epoch 16, batch 6650, loss[loss=0.2112, simple_loss=0.2983, pruned_loss=0.06204, over 16396.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2979, pruned_loss=0.06448, over 3071310.57 frames. ], batch size: 146, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:12:04,650 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 3.031e+02 3.577e+02 4.481e+02 9.334e+02, threshold=7.154e+02, percent-clipped=3.0 2023-04-30 09:12:50,518 INFO [train.py:904] (5/8) Epoch 16, batch 6700, loss[loss=0.1882, simple_loss=0.2818, pruned_loss=0.0473, over 16643.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2969, pruned_loss=0.06482, over 3074892.87 frames. ], batch size: 62, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:14:04,091 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9494, 2.7956, 2.7900, 2.0635, 2.6450, 2.1317, 2.7224, 2.8709], device='cuda:5'), covar=tensor([0.0281, 0.0728, 0.0536, 0.1722, 0.0817, 0.0926, 0.0548, 0.0678], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0157, 0.0163, 0.0149, 0.0140, 0.0126, 0.0141, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:14:07,730 INFO [train.py:904] (5/8) Epoch 16, batch 6750, loss[loss=0.1791, simple_loss=0.2706, pruned_loss=0.04382, over 16902.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2954, pruned_loss=0.06441, over 3078149.68 frames. ], batch size: 96, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:14:31,311 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-30 09:14:37,801 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.101e+02 2.962e+02 3.496e+02 4.058e+02 1.383e+03, threshold=6.992e+02, percent-clipped=2.0 2023-04-30 09:15:05,826 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159040.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:15:23,310 INFO [train.py:904] (5/8) Epoch 16, batch 6800, loss[loss=0.1948, simple_loss=0.2878, pruned_loss=0.05087, over 16687.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2956, pruned_loss=0.06422, over 3093550.76 frames. ], batch size: 134, lr: 4.22e-03, grad_scale: 8.0 2023-04-30 09:15:50,659 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:16:21,091 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:16:40,969 INFO [train.py:904] (5/8) Epoch 16, batch 6850, loss[loss=0.2007, simple_loss=0.3025, pruned_loss=0.04946, over 16438.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2968, pruned_loss=0.06457, over 3075387.14 frames. ], batch size: 75, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:17:12,962 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.851e+02 3.389e+02 4.147e+02 9.414e+02, threshold=6.778e+02, percent-clipped=4.0 2023-04-30 09:17:22,480 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:26,686 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159133.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:35,438 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:54,578 INFO [train.py:904] (5/8) Epoch 16, batch 6900, loss[loss=0.2115, simple_loss=0.3006, pruned_loss=0.06117, over 16853.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2984, pruned_loss=0.06322, over 3102288.27 frames. ], batch size: 116, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:18:07,178 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:18:56,512 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159191.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 09:19:13,878 INFO [train.py:904] (5/8) Epoch 16, batch 6950, loss[loss=0.1858, simple_loss=0.2825, pruned_loss=0.04453, over 16858.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.3004, pruned_loss=0.06536, over 3083041.37 frames. ], batch size: 102, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:19:42,693 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:19:48,722 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 3.353e+02 4.093e+02 5.038e+02 1.287e+03, threshold=8.186e+02, percent-clipped=11.0 2023-04-30 09:20:08,566 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3712, 3.6631, 3.7220, 2.2069, 3.2396, 2.5194, 3.8450, 3.8155], device='cuda:5'), covar=tensor([0.0205, 0.0719, 0.0538, 0.1808, 0.0727, 0.0908, 0.0488, 0.0736], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0156, 0.0162, 0.0149, 0.0141, 0.0126, 0.0141, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:20:29,902 INFO [train.py:904] (5/8) Epoch 16, batch 7000, loss[loss=0.2061, simple_loss=0.3, pruned_loss=0.0561, over 16754.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.3002, pruned_loss=0.06402, over 3111612.13 frames. ], batch size: 124, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:21:27,995 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:21:45,229 INFO [train.py:904] (5/8) Epoch 16, batch 7050, loss[loss=0.2356, simple_loss=0.3124, pruned_loss=0.07939, over 16660.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.301, pruned_loss=0.06414, over 3105986.40 frames. ], batch size: 57, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:22:18,901 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.988e+02 3.653e+02 4.362e+02 9.002e+02, threshold=7.306e+02, percent-clipped=3.0 2023-04-30 09:23:00,666 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:23:01,361 INFO [train.py:904] (5/8) Epoch 16, batch 7100, loss[loss=0.1938, simple_loss=0.2842, pruned_loss=0.05175, over 16716.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2994, pruned_loss=0.06382, over 3096837.07 frames. ], batch size: 124, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:23:54,855 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9980, 3.6082, 3.6340, 2.2800, 3.3349, 3.6403, 3.3569, 1.9865], device='cuda:5'), covar=tensor([0.0584, 0.0049, 0.0053, 0.0419, 0.0097, 0.0111, 0.0089, 0.0458], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0075, 0.0076, 0.0130, 0.0088, 0.0098, 0.0087, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:24:17,481 INFO [train.py:904] (5/8) Epoch 16, batch 7150, loss[loss=0.2008, simple_loss=0.2808, pruned_loss=0.06038, over 16706.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2974, pruned_loss=0.06398, over 3094780.07 frames. ], batch size: 134, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:24:49,382 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 3.206e+02 4.048e+02 4.680e+02 7.501e+02, threshold=8.096e+02, percent-clipped=1.0 2023-04-30 09:24:51,554 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:02,874 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:11,797 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:30,663 INFO [train.py:904] (5/8) Epoch 16, batch 7200, loss[loss=0.1877, simple_loss=0.2775, pruned_loss=0.04895, over 16654.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2951, pruned_loss=0.06265, over 3063617.11 frames. ], batch size: 134, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:26:15,871 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:16,031 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9967, 3.1984, 3.2088, 2.1823, 2.9740, 3.1874, 3.0474, 1.9300], device='cuda:5'), covar=tensor([0.0460, 0.0053, 0.0055, 0.0363, 0.0097, 0.0107, 0.0089, 0.0412], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0074, 0.0075, 0.0129, 0.0088, 0.0098, 0.0087, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:26:26,898 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159487.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:33,439 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159491.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 09:26:50,670 INFO [train.py:904] (5/8) Epoch 16, batch 7250, loss[loss=0.1812, simple_loss=0.2795, pruned_loss=0.04145, over 16914.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2929, pruned_loss=0.06154, over 3069149.03 frames. ], batch size: 102, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:27:11,565 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:27:23,392 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.928e+02 2.845e+02 3.560e+02 4.449e+02 7.164e+02, threshold=7.119e+02, percent-clipped=0.0 2023-04-30 09:27:47,289 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159539.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:28:06,282 INFO [train.py:904] (5/8) Epoch 16, batch 7300, loss[loss=0.2262, simple_loss=0.2956, pruned_loss=0.07846, over 11275.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.292, pruned_loss=0.06082, over 3088800.06 frames. ], batch size: 248, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:28:18,645 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8542, 2.0668, 2.4119, 3.1139, 2.1443, 2.3126, 2.2566, 2.1967], device='cuda:5'), covar=tensor([0.1122, 0.3005, 0.2072, 0.0595, 0.3950, 0.2034, 0.2844, 0.3110], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0418, 0.0346, 0.0315, 0.0424, 0.0481, 0.0388, 0.0488], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:28:36,367 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7231, 3.7475, 3.8996, 3.6771, 3.8481, 4.1788, 3.8829, 3.6254], device='cuda:5'), covar=tensor([0.2082, 0.2015, 0.2092, 0.2395, 0.2248, 0.1626, 0.1581, 0.2716], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0541, 0.0590, 0.0456, 0.0604, 0.0623, 0.0466, 0.0612], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:29:22,401 INFO [train.py:904] (5/8) Epoch 16, batch 7350, loss[loss=0.2222, simple_loss=0.3029, pruned_loss=0.07071, over 16835.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2932, pruned_loss=0.06223, over 3041199.98 frames. ], batch size: 116, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:56,735 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 3.169e+02 3.768e+02 4.450e+02 9.358e+02, threshold=7.536e+02, percent-clipped=3.0 2023-04-30 09:30:00,021 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 09:30:07,757 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:30:31,263 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159646.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:30:39,224 INFO [train.py:904] (5/8) Epoch 16, batch 7400, loss[loss=0.2029, simple_loss=0.2936, pruned_loss=0.05606, over 16280.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2941, pruned_loss=0.06213, over 3064723.51 frames. ], batch size: 165, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:31:03,188 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 09:31:21,341 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5456, 2.4147, 2.4118, 3.8334, 2.8232, 3.8855, 1.4466, 2.7708], device='cuda:5'), covar=tensor([0.1476, 0.0833, 0.1264, 0.0161, 0.0253, 0.0388, 0.1749, 0.0879], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0168, 0.0190, 0.0171, 0.0205, 0.0214, 0.0194, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:31:41,276 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:31:57,480 INFO [train.py:904] (5/8) Epoch 16, batch 7450, loss[loss=0.2249, simple_loss=0.3047, pruned_loss=0.07257, over 17047.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.295, pruned_loss=0.06295, over 3074825.70 frames. ], batch size: 53, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:32:33,423 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 3.124e+02 3.664e+02 4.558e+02 8.463e+02, threshold=7.327e+02, percent-clipped=4.0 2023-04-30 09:32:36,606 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:33:15,090 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4156, 4.2589, 4.4579, 4.6086, 4.7638, 4.3626, 4.6977, 4.7527], device='cuda:5'), covar=tensor([0.1804, 0.1246, 0.1451, 0.0658, 0.0544, 0.0981, 0.0718, 0.0611], device='cuda:5'), in_proj_covar=tensor([0.0576, 0.0709, 0.0835, 0.0715, 0.0540, 0.0566, 0.0578, 0.0673], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:33:17,751 INFO [train.py:904] (5/8) Epoch 16, batch 7500, loss[loss=0.192, simple_loss=0.2778, pruned_loss=0.05306, over 16685.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2953, pruned_loss=0.06239, over 3057949.64 frames. ], batch size: 89, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:33:51,104 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:34:21,872 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7701, 3.6332, 3.8082, 3.5810, 3.7669, 4.1809, 3.8202, 3.5908], device='cuda:5'), covar=tensor([0.1990, 0.2537, 0.2662, 0.2818, 0.2998, 0.2020, 0.1762, 0.2737], device='cuda:5'), in_proj_covar=tensor([0.0383, 0.0547, 0.0598, 0.0461, 0.0612, 0.0629, 0.0471, 0.0618], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:34:37,025 INFO [train.py:904] (5/8) Epoch 16, batch 7550, loss[loss=0.1869, simple_loss=0.2769, pruned_loss=0.04844, over 16892.00 frames. ], tot_loss[loss=0.211, simple_loss=0.295, pruned_loss=0.0635, over 3026273.81 frames. ], batch size: 90, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:34:58,295 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:35:10,144 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.970e+02 2.719e+02 3.178e+02 3.820e+02 7.109e+02, threshold=6.356e+02, percent-clipped=0.0 2023-04-30 09:35:53,838 INFO [train.py:904] (5/8) Epoch 16, batch 7600, loss[loss=0.2278, simple_loss=0.3075, pruned_loss=0.07402, over 16311.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2948, pruned_loss=0.06398, over 3031408.20 frames. ], batch size: 165, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:35:59,187 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:36:10,414 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:09,870 INFO [train.py:904] (5/8) Epoch 16, batch 7650, loss[loss=0.2675, simple_loss=0.3181, pruned_loss=0.1085, over 11538.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2952, pruned_loss=0.06424, over 3054006.65 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:37:21,788 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159910.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:30,402 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159916.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:43,265 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 3.207e+02 4.068e+02 5.310e+02 2.281e+03, threshold=8.135e+02, percent-clipped=15.0 2023-04-30 09:38:13,619 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:38:22,090 INFO [train.py:904] (5/8) Epoch 16, batch 7700, loss[loss=0.2067, simple_loss=0.2907, pruned_loss=0.06136, over 16257.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2953, pruned_loss=0.06444, over 3071741.57 frames. ], batch size: 165, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:38:37,689 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4258, 2.9806, 2.5616, 2.2731, 2.3442, 2.1446, 2.9286, 2.8528], device='cuda:5'), covar=tensor([0.2426, 0.1027, 0.1801, 0.2526, 0.2650, 0.2389, 0.0557, 0.1354], device='cuda:5'), in_proj_covar=tensor([0.0319, 0.0266, 0.0299, 0.0300, 0.0292, 0.0242, 0.0284, 0.0322], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:38:51,191 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159971.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:07,081 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 09:39:15,528 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:26,410 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159994.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:41,050 INFO [train.py:904] (5/8) Epoch 16, batch 7750, loss[loss=0.1994, simple_loss=0.2848, pruned_loss=0.05698, over 16676.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2953, pruned_loss=0.06444, over 3064100.48 frames. ], batch size: 57, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:39:52,490 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8166, 2.6748, 2.6592, 1.9390, 2.5184, 2.6452, 2.5387, 1.9056], device='cuda:5'), covar=tensor([0.0387, 0.0074, 0.0069, 0.0322, 0.0118, 0.0126, 0.0113, 0.0361], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0073, 0.0074, 0.0128, 0.0087, 0.0098, 0.0085, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:40:13,571 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 3.102e+02 3.551e+02 4.185e+02 8.391e+02, threshold=7.102e+02, percent-clipped=1.0 2023-04-30 09:40:53,478 INFO [train.py:904] (5/8) Epoch 16, batch 7800, loss[loss=0.2055, simple_loss=0.2924, pruned_loss=0.05929, over 16414.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2965, pruned_loss=0.06529, over 3058423.14 frames. ], batch size: 68, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:02,932 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6822, 2.6073, 2.4354, 3.9494, 2.8277, 3.9561, 1.5085, 2.8272], device='cuda:5'), covar=tensor([0.1318, 0.0774, 0.1199, 0.0179, 0.0245, 0.0385, 0.1591, 0.0818], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0166, 0.0188, 0.0169, 0.0202, 0.0211, 0.0192, 0.0188], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:42:08,849 INFO [train.py:904] (5/8) Epoch 16, batch 7850, loss[loss=0.2213, simple_loss=0.2939, pruned_loss=0.07436, over 11485.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2972, pruned_loss=0.06495, over 3062892.74 frames. ], batch size: 248, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:27,844 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4008, 3.9362, 3.9815, 2.5909, 3.5290, 3.9089, 3.5836, 2.2420], device='cuda:5'), covar=tensor([0.0452, 0.0041, 0.0036, 0.0351, 0.0087, 0.0093, 0.0076, 0.0382], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0073, 0.0074, 0.0129, 0.0087, 0.0097, 0.0086, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:42:43,407 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.758e+02 3.305e+02 4.051e+02 1.069e+03, threshold=6.609e+02, percent-clipped=3.0 2023-04-30 09:43:25,082 INFO [train.py:904] (5/8) Epoch 16, batch 7900, loss[loss=0.1923, simple_loss=0.2848, pruned_loss=0.04986, over 16876.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2958, pruned_loss=0.06383, over 3076791.07 frames. ], batch size: 90, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:44:10,830 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7402, 3.7163, 3.8684, 3.6862, 3.8001, 4.2140, 3.8481, 3.6161], device='cuda:5'), covar=tensor([0.2321, 0.2377, 0.2474, 0.2412, 0.2689, 0.1819, 0.1741, 0.2666], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0554, 0.0604, 0.0465, 0.0618, 0.0635, 0.0477, 0.0627], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 09:44:43,672 INFO [train.py:904] (5/8) Epoch 16, batch 7950, loss[loss=0.2192, simple_loss=0.2989, pruned_loss=0.06975, over 16869.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.296, pruned_loss=0.0644, over 3066935.22 frames. ], batch size: 116, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:44:56,927 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:45:16,542 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.700e+02 3.245e+02 4.105e+02 1.062e+03, threshold=6.489e+02, percent-clipped=6.0 2023-04-30 09:45:30,209 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 09:45:56,368 INFO [train.py:904] (5/8) Epoch 16, batch 8000, loss[loss=0.2024, simple_loss=0.3, pruned_loss=0.05236, over 17127.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.297, pruned_loss=0.06526, over 3068420.80 frames. ], batch size: 47, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:46:17,935 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:46:33,968 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6748, 3.7379, 2.3312, 4.3173, 2.8755, 4.2644, 2.5003, 2.9439], device='cuda:5'), covar=tensor([0.0261, 0.0364, 0.1636, 0.0238, 0.0790, 0.0527, 0.1524, 0.0791], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0170, 0.0193, 0.0147, 0.0173, 0.0211, 0.0202, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:46:43,528 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 09:46:50,255 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:47:11,929 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3104, 4.1509, 4.3757, 4.5135, 4.6646, 4.2730, 4.5823, 4.6549], device='cuda:5'), covar=tensor([0.1762, 0.1215, 0.1535, 0.0672, 0.0575, 0.0973, 0.0790, 0.0709], device='cuda:5'), in_proj_covar=tensor([0.0581, 0.0716, 0.0842, 0.0718, 0.0543, 0.0570, 0.0584, 0.0678], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:47:12,601 INFO [train.py:904] (5/8) Epoch 16, batch 8050, loss[loss=0.2115, simple_loss=0.298, pruned_loss=0.06249, over 16908.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.297, pruned_loss=0.06515, over 3054017.80 frames. ], batch size: 109, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:47:21,027 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160307.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:47:47,809 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.763e+02 2.953e+02 3.727e+02 4.663e+02 8.321e+02, threshold=7.453e+02, percent-clipped=3.0 2023-04-30 09:48:04,064 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:48:29,612 INFO [train.py:904] (5/8) Epoch 16, batch 8100, loss[loss=0.2236, simple_loss=0.306, pruned_loss=0.07057, over 16467.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.296, pruned_loss=0.06452, over 3050830.25 frames. ], batch size: 68, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:48:54,842 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160368.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:49:46,005 INFO [train.py:904] (5/8) Epoch 16, batch 8150, loss[loss=0.1913, simple_loss=0.2717, pruned_loss=0.05551, over 16987.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2933, pruned_loss=0.06309, over 3072886.86 frames. ], batch size: 55, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:50:21,736 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 3.128e+02 3.748e+02 4.647e+02 9.176e+02, threshold=7.497e+02, percent-clipped=1.0 2023-04-30 09:50:22,537 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-30 09:51:05,074 INFO [train.py:904] (5/8) Epoch 16, batch 8200, loss[loss=0.1883, simple_loss=0.2683, pruned_loss=0.05414, over 16567.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2911, pruned_loss=0.06239, over 3071932.11 frames. ], batch size: 57, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:52:27,315 INFO [train.py:904] (5/8) Epoch 16, batch 8250, loss[loss=0.1969, simple_loss=0.2878, pruned_loss=0.05298, over 16588.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.29, pruned_loss=0.05949, over 3076368.59 frames. ], batch size: 62, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:52:42,422 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:52:45,209 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-30 09:53:04,608 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.122e+02 2.811e+02 3.275e+02 4.165e+02 1.278e+03, threshold=6.551e+02, percent-clipped=3.0 2023-04-30 09:53:40,905 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 09:53:49,473 INFO [train.py:904] (5/8) Epoch 16, batch 8300, loss[loss=0.1822, simple_loss=0.2838, pruned_loss=0.04026, over 16869.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2875, pruned_loss=0.05642, over 3082937.03 frames. ], batch size: 96, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:54:01,473 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:54:13,142 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:54:55,784 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-30 09:55:10,572 INFO [train.py:904] (5/8) Epoch 16, batch 8350, loss[loss=0.1894, simple_loss=0.2865, pruned_loss=0.04614, over 15174.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2865, pruned_loss=0.05461, over 3079457.69 frames. ], batch size: 190, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:55:27,791 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 09:55:30,927 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:55:48,786 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.217e+02 2.608e+02 3.326e+02 6.898e+02, threshold=5.216e+02, percent-clipped=1.0 2023-04-30 09:55:58,260 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5796, 1.9017, 2.1747, 2.6628, 2.6482, 2.9304, 1.9991, 2.9169], device='cuda:5'), covar=tensor([0.0182, 0.0433, 0.0307, 0.0239, 0.0242, 0.0171, 0.0411, 0.0126], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0183, 0.0168, 0.0171, 0.0182, 0.0140, 0.0183, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:56:33,068 INFO [train.py:904] (5/8) Epoch 16, batch 8400, loss[loss=0.1864, simple_loss=0.2713, pruned_loss=0.05074, over 12533.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2837, pruned_loss=0.05253, over 3062802.38 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:56:34,031 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 09:56:51,287 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160663.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 09:57:54,572 INFO [train.py:904] (5/8) Epoch 16, batch 8450, loss[loss=0.1753, simple_loss=0.2585, pruned_loss=0.04609, over 12575.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2822, pruned_loss=0.05108, over 3066991.24 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:58:25,795 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5218, 3.4993, 3.4642, 2.6374, 3.4157, 1.9462, 3.2137, 2.7839], device='cuda:5'), covar=tensor([0.0152, 0.0129, 0.0184, 0.0249, 0.0116, 0.2602, 0.0155, 0.0265], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0132, 0.0178, 0.0162, 0.0150, 0.0189, 0.0164, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:58:31,820 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 2.257e+02 2.696e+02 3.215e+02 7.413e+02, threshold=5.391e+02, percent-clipped=3.0 2023-04-30 09:58:40,511 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9810, 5.3002, 5.0997, 5.0741, 4.8304, 4.7869, 4.7620, 5.3958], device='cuda:5'), covar=tensor([0.1184, 0.0900, 0.0983, 0.0777, 0.0743, 0.0860, 0.1146, 0.0801], device='cuda:5'), in_proj_covar=tensor([0.0598, 0.0728, 0.0596, 0.0536, 0.0458, 0.0475, 0.0608, 0.0561], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:59:15,555 INFO [train.py:904] (5/8) Epoch 16, batch 8500, loss[loss=0.1846, simple_loss=0.2577, pruned_loss=0.05581, over 11491.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2783, pruned_loss=0.04871, over 3058023.02 frames. ], batch size: 246, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:59:26,747 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0757, 2.8107, 2.8931, 2.0855, 2.6496, 2.0596, 2.8581, 3.0256], device='cuda:5'), covar=tensor([0.0256, 0.0877, 0.0501, 0.1935, 0.0928, 0.1132, 0.0551, 0.0690], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0151, 0.0159, 0.0144, 0.0137, 0.0123, 0.0137, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 09:59:42,508 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4332, 4.4408, 4.7945, 4.7650, 4.7574, 4.5024, 4.4741, 4.3552], device='cuda:5'), covar=tensor([0.0371, 0.0759, 0.0431, 0.0444, 0.0504, 0.0417, 0.1032, 0.0509], device='cuda:5'), in_proj_covar=tensor([0.0370, 0.0401, 0.0391, 0.0369, 0.0437, 0.0413, 0.0504, 0.0328], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 09:59:49,866 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3821, 2.0400, 1.6190, 1.6032, 2.2034, 1.8663, 2.0631, 2.3435], device='cuda:5'), covar=tensor([0.0198, 0.0386, 0.0548, 0.0475, 0.0257, 0.0376, 0.0201, 0.0287], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0213, 0.0207, 0.0207, 0.0212, 0.0213, 0.0215, 0.0208], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 09:59:50,029 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 10:00:26,474 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9185, 4.2646, 3.2815, 2.2645, 2.7668, 2.6270, 4.4937, 3.7049], device='cuda:5'), covar=tensor([0.2498, 0.0420, 0.1402, 0.2849, 0.2749, 0.1878, 0.0296, 0.1043], device='cuda:5'), in_proj_covar=tensor([0.0309, 0.0255, 0.0288, 0.0290, 0.0280, 0.0235, 0.0274, 0.0310], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:00:42,457 INFO [train.py:904] (5/8) Epoch 16, batch 8550, loss[loss=0.2315, simple_loss=0.3167, pruned_loss=0.07319, over 16703.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2755, pruned_loss=0.04767, over 3031300.17 frames. ], batch size: 134, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:01:27,142 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 2.208e+02 2.559e+02 3.052e+02 5.214e+02, threshold=5.118e+02, percent-clipped=0.0 2023-04-30 10:01:43,107 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 10:01:44,758 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.54 vs. limit=5.0 2023-04-30 10:02:22,624 INFO [train.py:904] (5/8) Epoch 16, batch 8600, loss[loss=0.2086, simple_loss=0.2846, pruned_loss=0.06628, over 12260.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2758, pruned_loss=0.04675, over 3029289.09 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:02:57,168 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5676, 3.6657, 3.4605, 3.1785, 3.2186, 3.5717, 3.2936, 3.3481], device='cuda:5'), covar=tensor([0.0637, 0.0620, 0.0327, 0.0282, 0.0669, 0.0520, 0.1443, 0.0608], device='cuda:5'), in_proj_covar=tensor([0.0261, 0.0366, 0.0304, 0.0294, 0.0316, 0.0344, 0.0211, 0.0365], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:04:02,533 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160901.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:04:03,226 INFO [train.py:904] (5/8) Epoch 16, batch 8650, loss[loss=0.1495, simple_loss=0.2533, pruned_loss=0.02283, over 16868.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.273, pruned_loss=0.04475, over 3025331.07 frames. ], batch size: 102, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:04:05,259 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 10:04:26,884 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6019, 4.8770, 4.6416, 4.6302, 4.4110, 4.3571, 4.2966, 4.9032], device='cuda:5'), covar=tensor([0.1127, 0.0782, 0.0973, 0.0765, 0.0794, 0.1184, 0.1228, 0.0898], device='cuda:5'), in_proj_covar=tensor([0.0589, 0.0716, 0.0586, 0.0529, 0.0451, 0.0468, 0.0599, 0.0553], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:04:55,480 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6739, 5.0021, 4.7625, 4.7646, 4.5567, 4.4920, 4.4351, 5.0316], device='cuda:5'), covar=tensor([0.1105, 0.0761, 0.0920, 0.0727, 0.0767, 0.1035, 0.1205, 0.0901], device='cuda:5'), in_proj_covar=tensor([0.0589, 0.0716, 0.0585, 0.0528, 0.0451, 0.0468, 0.0599, 0.0552], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:04:56,267 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.325e+02 2.764e+02 3.397e+02 5.575e+02, threshold=5.528e+02, percent-clipped=1.0 2023-04-30 10:05:09,134 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0963, 2.6315, 2.7506, 1.8716, 2.7993, 2.8814, 2.5430, 2.4570], device='cuda:5'), covar=tensor([0.0658, 0.0215, 0.0200, 0.0972, 0.0084, 0.0209, 0.0413, 0.0414], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0101, 0.0087, 0.0132, 0.0070, 0.0111, 0.0120, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 10:05:52,489 INFO [train.py:904] (5/8) Epoch 16, batch 8700, loss[loss=0.1581, simple_loss=0.2545, pruned_loss=0.03088, over 16754.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2702, pruned_loss=0.04317, over 3028552.30 frames. ], batch size: 83, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:06:13,653 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160962.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:06:15,213 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160963.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 10:06:18,658 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:06:39,379 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3194, 3.4910, 3.6897, 3.6661, 3.6772, 3.4958, 3.5230, 3.5512], device='cuda:5'), covar=tensor([0.0414, 0.0695, 0.0508, 0.0523, 0.0562, 0.0509, 0.0771, 0.0491], device='cuda:5'), in_proj_covar=tensor([0.0363, 0.0393, 0.0386, 0.0363, 0.0431, 0.0404, 0.0497, 0.0322], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 10:07:29,347 INFO [train.py:904] (5/8) Epoch 16, batch 8750, loss[loss=0.1532, simple_loss=0.2416, pruned_loss=0.03241, over 12389.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2695, pruned_loss=0.0425, over 3037065.65 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 4.0 2023-04-30 10:07:53,660 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161011.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 10:08:15,035 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6342, 2.5969, 2.4590, 3.9245, 2.5905, 3.9661, 1.4323, 3.0320], device='cuda:5'), covar=tensor([0.1320, 0.0732, 0.1122, 0.0121, 0.0124, 0.0325, 0.1652, 0.0634], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0162, 0.0185, 0.0164, 0.0196, 0.0206, 0.0189, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:5') 2023-04-30 10:08:27,684 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.172e+02 2.627e+02 3.227e+02 5.767e+02, threshold=5.254e+02, percent-clipped=1.0 2023-04-30 10:08:28,325 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:08:29,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2245, 2.3530, 1.9771, 2.0599, 2.7086, 2.4220, 2.7697, 2.9712], device='cuda:5'), covar=tensor([0.0126, 0.0409, 0.0521, 0.0520, 0.0255, 0.0404, 0.0203, 0.0258], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0214, 0.0208, 0.0208, 0.0212, 0.0213, 0.0214, 0.0207], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:09:03,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 10:09:11,661 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3320, 2.1368, 2.1560, 3.9644, 1.9995, 2.5239, 2.2126, 2.2697], device='cuda:5'), covar=tensor([0.1119, 0.3627, 0.2917, 0.0469, 0.4547, 0.2595, 0.3600, 0.3529], device='cuda:5'), in_proj_covar=tensor([0.0370, 0.0411, 0.0342, 0.0309, 0.0417, 0.0470, 0.0380, 0.0476], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:09:20,788 INFO [train.py:904] (5/8) Epoch 16, batch 8800, loss[loss=0.1738, simple_loss=0.2613, pruned_loss=0.04321, over 12714.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2683, pruned_loss=0.04183, over 3055429.67 frames. ], batch size: 246, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:11:05,650 INFO [train.py:904] (5/8) Epoch 16, batch 8850, loss[loss=0.1744, simple_loss=0.2814, pruned_loss=0.03365, over 16223.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2704, pruned_loss=0.04152, over 3036651.42 frames. ], batch size: 165, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:11:55,955 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.310e+02 2.876e+02 3.561e+02 5.593e+02, threshold=5.752e+02, percent-clipped=2.0 2023-04-30 10:12:53,427 INFO [train.py:904] (5/8) Epoch 16, batch 8900, loss[loss=0.1741, simple_loss=0.2681, pruned_loss=0.04011, over 16251.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2716, pruned_loss=0.04127, over 3052115.13 frames. ], batch size: 165, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:13:21,514 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3145, 3.4808, 3.6708, 3.6497, 3.6845, 3.4971, 3.5371, 3.5365], device='cuda:5'), covar=tensor([0.0386, 0.0680, 0.0484, 0.0474, 0.0412, 0.0501, 0.0683, 0.0415], device='cuda:5'), in_proj_covar=tensor([0.0355, 0.0383, 0.0378, 0.0355, 0.0423, 0.0396, 0.0484, 0.0315], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 10:13:51,341 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4123, 4.4502, 4.2876, 3.9524, 3.9577, 4.3409, 4.1045, 4.0593], device='cuda:5'), covar=tensor([0.0531, 0.0512, 0.0293, 0.0299, 0.0859, 0.0523, 0.0562, 0.0710], device='cuda:5'), in_proj_covar=tensor([0.0255, 0.0357, 0.0298, 0.0289, 0.0308, 0.0337, 0.0207, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:14:59,663 INFO [train.py:904] (5/8) Epoch 16, batch 8950, loss[loss=0.1641, simple_loss=0.2572, pruned_loss=0.03556, over 15491.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2712, pruned_loss=0.04134, over 3074553.79 frames. ], batch size: 191, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:15:30,155 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:15:47,455 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.276e+02 2.813e+02 3.338e+02 5.711e+02, threshold=5.626e+02, percent-clipped=0.0 2023-04-30 10:16:01,849 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:16:46,893 INFO [train.py:904] (5/8) Epoch 16, batch 9000, loss[loss=0.1855, simple_loss=0.2638, pruned_loss=0.05354, over 12082.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2683, pruned_loss=0.03989, over 3067443.95 frames. ], batch size: 250, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:16:46,894 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 10:16:56,916 INFO [train.py:938] (5/8) Epoch 16, validation: loss=0.1491, simple_loss=0.2531, pruned_loss=0.02259, over 944034.00 frames. 2023-04-30 10:16:56,917 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 10:17:08,106 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:17:13,611 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-04-30 10:17:49,258 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:18:20,849 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:18:40,788 INFO [train.py:904] (5/8) Epoch 16, batch 9050, loss[loss=0.1976, simple_loss=0.2822, pruned_loss=0.05651, over 16329.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.27, pruned_loss=0.04055, over 3104854.40 frames. ], batch size: 146, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:19:13,366 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0983, 3.1077, 1.8313, 3.3093, 2.2496, 3.2907, 2.0588, 2.6903], device='cuda:5'), covar=tensor([0.0281, 0.0337, 0.1549, 0.0224, 0.0869, 0.0586, 0.1481, 0.0651], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0185, 0.0140, 0.0166, 0.0200, 0.0194, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 10:19:16,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5867, 4.5792, 4.9693, 4.9415, 4.9598, 4.7029, 4.6901, 4.4859], device='cuda:5'), covar=tensor([0.0312, 0.0597, 0.0406, 0.0390, 0.0394, 0.0362, 0.0786, 0.0418], device='cuda:5'), in_proj_covar=tensor([0.0358, 0.0386, 0.0380, 0.0358, 0.0425, 0.0399, 0.0486, 0.0317], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 10:19:19,445 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161321.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:19:27,692 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.317e+02 2.661e+02 3.212e+02 6.846e+02, threshold=5.322e+02, percent-clipped=5.0 2023-04-30 10:20:20,452 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1829, 3.6083, 3.5346, 2.3776, 3.1935, 3.5204, 3.3946, 2.0336], device='cuda:5'), covar=tensor([0.0490, 0.0039, 0.0048, 0.0368, 0.0096, 0.0086, 0.0076, 0.0453], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0073, 0.0074, 0.0128, 0.0087, 0.0096, 0.0085, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 10:20:22,819 INFO [train.py:904] (5/8) Epoch 16, batch 9100, loss[loss=0.1942, simple_loss=0.2866, pruned_loss=0.05085, over 15523.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2694, pruned_loss=0.0409, over 3100778.90 frames. ], batch size: 192, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:22:19,307 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:22:19,979 INFO [train.py:904] (5/8) Epoch 16, batch 9150, loss[loss=0.1729, simple_loss=0.2676, pruned_loss=0.03912, over 16418.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2699, pruned_loss=0.04075, over 3088097.82 frames. ], batch size: 146, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:22:47,602 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7879, 1.7678, 2.2506, 2.6364, 2.6013, 2.9561, 1.9053, 3.0043], device='cuda:5'), covar=tensor([0.0178, 0.0508, 0.0333, 0.0291, 0.0283, 0.0171, 0.0492, 0.0126], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0180, 0.0167, 0.0168, 0.0180, 0.0137, 0.0182, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:23:13,976 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.279e+02 2.699e+02 3.193e+02 4.519e+02, threshold=5.398e+02, percent-clipped=0.0 2023-04-30 10:24:04,529 INFO [train.py:904] (5/8) Epoch 16, batch 9200, loss[loss=0.1887, simple_loss=0.288, pruned_loss=0.04472, over 16323.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2658, pruned_loss=0.04011, over 3067086.38 frames. ], batch size: 146, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:24:07,029 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5595, 4.5677, 5.0000, 4.9491, 4.9478, 4.6860, 4.6199, 4.5140], device='cuda:5'), covar=tensor([0.0331, 0.0613, 0.0342, 0.0384, 0.0440, 0.0369, 0.0886, 0.0406], device='cuda:5'), in_proj_covar=tensor([0.0356, 0.0385, 0.0380, 0.0355, 0.0424, 0.0398, 0.0486, 0.0316], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 10:24:10,158 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8387, 3.7454, 3.9181, 4.0086, 4.0806, 3.7100, 4.0498, 4.0941], device='cuda:5'), covar=tensor([0.1400, 0.1059, 0.1208, 0.0649, 0.0538, 0.1597, 0.0679, 0.0719], device='cuda:5'), in_proj_covar=tensor([0.0556, 0.0686, 0.0806, 0.0693, 0.0523, 0.0547, 0.0558, 0.0653], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:24:21,004 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:24:24,506 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:25:38,249 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9548, 3.8733, 4.1086, 2.0699, 4.2962, 4.3468, 3.2974, 3.2300], device='cuda:5'), covar=tensor([0.0613, 0.0178, 0.0173, 0.1088, 0.0048, 0.0093, 0.0356, 0.0384], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0101, 0.0086, 0.0132, 0.0070, 0.0110, 0.0120, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 10:25:42,690 INFO [train.py:904] (5/8) Epoch 16, batch 9250, loss[loss=0.1661, simple_loss=0.2483, pruned_loss=0.04193, over 12413.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2652, pruned_loss=0.04029, over 3030039.87 frames. ], batch size: 248, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:25:50,712 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3598, 4.1434, 4.3157, 4.5855, 4.6921, 4.3131, 4.7116, 4.7115], device='cuda:5'), covar=tensor([0.1722, 0.1339, 0.1807, 0.0832, 0.0779, 0.1161, 0.0798, 0.0865], device='cuda:5'), in_proj_covar=tensor([0.0554, 0.0682, 0.0802, 0.0690, 0.0520, 0.0545, 0.0556, 0.0649], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:26:23,341 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161521.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:26:36,798 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.444e+02 2.938e+02 3.576e+02 6.685e+02, threshold=5.875e+02, percent-clipped=5.0 2023-04-30 10:27:21,117 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6848, 2.6559, 1.8915, 2.8413, 2.1083, 2.8342, 2.2158, 2.5116], device='cuda:5'), covar=tensor([0.0249, 0.0325, 0.1128, 0.0263, 0.0626, 0.0441, 0.0983, 0.0488], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0163, 0.0185, 0.0140, 0.0166, 0.0200, 0.0194, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 10:27:34,805 INFO [train.py:904] (5/8) Epoch 16, batch 9300, loss[loss=0.1705, simple_loss=0.2514, pruned_loss=0.04484, over 12642.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2641, pruned_loss=0.03979, over 3021747.49 frames. ], batch size: 250, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:27:45,438 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:27:58,222 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3468, 3.4417, 3.6776, 3.6475, 3.6543, 3.4595, 3.5113, 3.5634], device='cuda:5'), covar=tensor([0.0383, 0.0821, 0.0490, 0.0467, 0.0541, 0.0525, 0.0756, 0.0430], device='cuda:5'), in_proj_covar=tensor([0.0354, 0.0382, 0.0377, 0.0352, 0.0421, 0.0396, 0.0482, 0.0314], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 10:28:23,209 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:54,781 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:29:21,685 INFO [train.py:904] (5/8) Epoch 16, batch 9350, loss[loss=0.1734, simple_loss=0.2614, pruned_loss=0.04271, over 16749.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2636, pruned_loss=0.03961, over 3048895.44 frames. ], batch size: 124, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:29:28,431 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161605.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:30:03,264 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:30:09,752 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5669, 3.5376, 3.5065, 2.8849, 3.4026, 1.9817, 3.2451, 2.8902], device='cuda:5'), covar=tensor([0.0138, 0.0144, 0.0174, 0.0221, 0.0111, 0.2261, 0.0141, 0.0204], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0127, 0.0172, 0.0155, 0.0147, 0.0186, 0.0160, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:30:12,564 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.307e+02 2.790e+02 3.265e+02 7.040e+02, threshold=5.579e+02, percent-clipped=1.0 2023-04-30 10:31:04,000 INFO [train.py:904] (5/8) Epoch 16, batch 9400, loss[loss=0.154, simple_loss=0.2406, pruned_loss=0.03365, over 12321.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2631, pruned_loss=0.0393, over 3028369.05 frames. ], batch size: 249, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:31:10,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0158, 1.7896, 1.6147, 1.4722, 1.9701, 1.5654, 1.6217, 1.9340], device='cuda:5'), covar=tensor([0.0138, 0.0296, 0.0398, 0.0336, 0.0223, 0.0269, 0.0156, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0216, 0.0210, 0.0209, 0.0215, 0.0216, 0.0216, 0.0207], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:31:12,976 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3956, 2.9749, 2.6483, 2.2424, 2.1477, 2.2206, 2.9603, 2.7846], device='cuda:5'), covar=tensor([0.2523, 0.0767, 0.1582, 0.2844, 0.2548, 0.2079, 0.0503, 0.1396], device='cuda:5'), in_proj_covar=tensor([0.0308, 0.0254, 0.0286, 0.0288, 0.0273, 0.0233, 0.0272, 0.0307], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:31:26,526 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-04-30 10:31:33,975 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4817, 3.5122, 2.7726, 2.1148, 2.2327, 2.3516, 3.7788, 3.0842], device='cuda:5'), covar=tensor([0.2746, 0.0656, 0.1620, 0.2713, 0.2746, 0.2015, 0.0396, 0.1329], device='cuda:5'), in_proj_covar=tensor([0.0307, 0.0254, 0.0286, 0.0288, 0.0273, 0.0233, 0.0272, 0.0307], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:31:39,354 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161669.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:31:50,832 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5444, 3.5281, 3.5016, 2.8966, 3.4480, 2.0309, 3.2844, 2.8947], device='cuda:5'), covar=tensor([0.0113, 0.0095, 0.0153, 0.0175, 0.0087, 0.2217, 0.0115, 0.0178], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0128, 0.0173, 0.0156, 0.0147, 0.0186, 0.0160, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:32:44,169 INFO [train.py:904] (5/8) Epoch 16, batch 9450, loss[loss=0.1649, simple_loss=0.2624, pruned_loss=0.03372, over 15302.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2647, pruned_loss=0.03915, over 3037151.97 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:33:33,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.284e+02 2.576e+02 3.195e+02 6.155e+02, threshold=5.152e+02, percent-clipped=3.0 2023-04-30 10:33:48,849 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 10:33:58,509 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4028, 4.5134, 4.6646, 4.4567, 4.5356, 5.0341, 4.5253, 4.2736], device='cuda:5'), covar=tensor([0.1305, 0.1902, 0.1855, 0.1976, 0.2349, 0.1020, 0.1499, 0.2320], device='cuda:5'), in_proj_covar=tensor([0.0357, 0.0520, 0.0566, 0.0435, 0.0579, 0.0603, 0.0447, 0.0583], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 10:34:23,908 INFO [train.py:904] (5/8) Epoch 16, batch 9500, loss[loss=0.1658, simple_loss=0.2655, pruned_loss=0.03307, over 16913.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2642, pruned_loss=0.03869, over 3047988.51 frames. ], batch size: 102, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:34:37,549 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161757.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:08,216 INFO [train.py:904] (5/8) Epoch 16, batch 9550, loss[loss=0.1828, simple_loss=0.2803, pruned_loss=0.04261, over 16865.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2641, pruned_loss=0.03906, over 3051955.28 frames. ], batch size: 90, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:36:38,449 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:40,709 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161816.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:37:00,503 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.146e+02 2.559e+02 3.093e+02 5.125e+02, threshold=5.118e+02, percent-clipped=0.0 2023-04-30 10:37:02,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 10:37:51,451 INFO [train.py:904] (5/8) Epoch 16, batch 9600, loss[loss=0.1899, simple_loss=0.2917, pruned_loss=0.04408, over 16270.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2657, pruned_loss=0.03974, over 3054947.01 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:38:29,836 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:38:38,456 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:39:02,487 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:39:04,823 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 10:39:37,497 INFO [train.py:904] (5/8) Epoch 16, batch 9650, loss[loss=0.1767, simple_loss=0.2728, pruned_loss=0.04026, over 16404.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2675, pruned_loss=0.04002, over 3053539.94 frames. ], batch size: 146, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:40:20,093 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7358, 4.8857, 5.0486, 4.8404, 4.9560, 5.4324, 4.9625, 4.7090], device='cuda:5'), covar=tensor([0.1063, 0.1781, 0.1648, 0.1741, 0.2362, 0.0982, 0.1428, 0.2388], device='cuda:5'), in_proj_covar=tensor([0.0360, 0.0521, 0.0569, 0.0436, 0.0579, 0.0604, 0.0450, 0.0585], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 10:40:22,072 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:40:29,706 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5641, 4.5529, 4.4239, 3.8600, 4.4028, 1.6655, 4.1938, 4.3175], device='cuda:5'), covar=tensor([0.0099, 0.0080, 0.0163, 0.0321, 0.0107, 0.2583, 0.0140, 0.0206], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0128, 0.0173, 0.0156, 0.0147, 0.0187, 0.0160, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:40:36,071 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.311e+02 2.824e+02 3.460e+02 5.730e+02, threshold=5.647e+02, percent-clipped=1.0 2023-04-30 10:40:49,472 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:41:23,674 INFO [train.py:904] (5/8) Epoch 16, batch 9700, loss[loss=0.1673, simple_loss=0.2634, pruned_loss=0.03557, over 15265.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2663, pruned_loss=0.03977, over 3047349.68 frames. ], batch size: 190, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:42:01,860 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9944, 2.2731, 1.8737, 2.0497, 2.6419, 2.2973, 2.6249, 2.8411], device='cuda:5'), covar=tensor([0.0122, 0.0392, 0.0492, 0.0442, 0.0231, 0.0373, 0.0224, 0.0224], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0215, 0.0209, 0.0208, 0.0214, 0.0215, 0.0213, 0.0206], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:42:48,165 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 10:43:08,535 INFO [train.py:904] (5/8) Epoch 16, batch 9750, loss[loss=0.1582, simple_loss=0.2574, pruned_loss=0.02952, over 16784.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2648, pruned_loss=0.03968, over 3050094.06 frames. ], batch size: 83, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:43:44,137 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 10:43:58,496 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.156e+02 2.615e+02 3.105e+02 5.589e+02, threshold=5.230e+02, percent-clipped=0.0 2023-04-30 10:44:46,292 INFO [train.py:904] (5/8) Epoch 16, batch 9800, loss[loss=0.1561, simple_loss=0.2438, pruned_loss=0.03422, over 12303.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.265, pruned_loss=0.03871, over 3069802.59 frames. ], batch size: 247, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:44:56,994 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162057.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:45:34,286 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:31,263 INFO [train.py:904] (5/8) Epoch 16, batch 9850, loss[loss=0.1712, simple_loss=0.2651, pruned_loss=0.03868, over 16473.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2662, pruned_loss=0.03832, over 3086564.58 frames. ], batch size: 68, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:46:38,615 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:59,782 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:47:23,364 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 2.292e+02 2.812e+02 3.288e+02 4.754e+02, threshold=5.623e+02, percent-clipped=0.0 2023-04-30 10:47:51,584 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162138.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:48:22,264 INFO [train.py:904] (5/8) Epoch 16, batch 9900, loss[loss=0.1857, simple_loss=0.2707, pruned_loss=0.05037, over 12647.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2661, pruned_loss=0.03805, over 3079110.03 frames. ], batch size: 248, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:48:36,579 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 10:48:52,744 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:49:08,597 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:49:42,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0533, 3.3900, 3.3822, 2.3899, 3.0375, 3.3633, 3.2593, 1.7900], device='cuda:5'), covar=tensor([0.0506, 0.0040, 0.0047, 0.0322, 0.0101, 0.0079, 0.0065, 0.0508], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0073, 0.0073, 0.0128, 0.0087, 0.0095, 0.0084, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 10:50:22,156 INFO [train.py:904] (5/8) Epoch 16, batch 9950, loss[loss=0.1756, simple_loss=0.2743, pruned_loss=0.03845, over 16716.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2678, pruned_loss=0.03829, over 3073700.70 frames. ], batch size: 134, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:51:26,524 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.161e+02 2.630e+02 3.300e+02 8.357e+02, threshold=5.260e+02, percent-clipped=1.0 2023-04-30 10:51:36,427 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:52:23,902 INFO [train.py:904] (5/8) Epoch 16, batch 10000, loss[loss=0.1863, simple_loss=0.2813, pruned_loss=0.04562, over 16687.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2666, pruned_loss=0.03793, over 3086706.59 frames. ], batch size: 134, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:52:31,743 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1247, 4.0065, 4.2109, 4.3480, 4.4718, 4.0745, 4.4274, 4.4889], device='cuda:5'), covar=tensor([0.1667, 0.1163, 0.1434, 0.0705, 0.0498, 0.1107, 0.0620, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0556, 0.0685, 0.0803, 0.0692, 0.0523, 0.0546, 0.0559, 0.0652], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:52:38,201 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 10:53:46,685 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:54:04,077 INFO [train.py:904] (5/8) Epoch 16, batch 10050, loss[loss=0.1847, simple_loss=0.2821, pruned_loss=0.04363, over 15314.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2668, pruned_loss=0.038, over 3100563.26 frames. ], batch size: 191, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:54:44,331 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 10:54:54,661 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.225e+02 2.711e+02 3.158e+02 7.314e+02, threshold=5.421e+02, percent-clipped=6.0 2023-04-30 10:55:38,484 INFO [train.py:904] (5/8) Epoch 16, batch 10100, loss[loss=0.1535, simple_loss=0.2463, pruned_loss=0.03035, over 16415.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2675, pruned_loss=0.03833, over 3089372.69 frames. ], batch size: 146, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:57:23,030 INFO [train.py:904] (5/8) Epoch 17, batch 0, loss[loss=0.2713, simple_loss=0.3183, pruned_loss=0.1121, over 16915.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3183, pruned_loss=0.1121, over 16915.00 frames. ], batch size: 109, lr: 4.05e-03, grad_scale: 8.0 2023-04-30 10:57:23,030 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 10:57:30,747 INFO [train.py:938] (5/8) Epoch 17, validation: loss=0.1481, simple_loss=0.2518, pruned_loss=0.02217, over 944034.00 frames. 2023-04-30 10:57:30,749 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 10:57:39,313 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5032, 5.8976, 5.6179, 5.6693, 5.2932, 5.1935, 5.3201, 5.9423], device='cuda:5'), covar=tensor([0.1372, 0.0899, 0.0980, 0.0830, 0.0925, 0.0665, 0.1058, 0.0925], device='cuda:5'), in_proj_covar=tensor([0.0582, 0.0715, 0.0578, 0.0526, 0.0452, 0.0464, 0.0595, 0.0553], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 10:58:09,503 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.232e+02 2.721e+02 3.359e+02 8.921e+02, threshold=5.442e+02, percent-clipped=3.0 2023-04-30 10:58:14,199 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:58:39,642 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 10:58:39,855 INFO [train.py:904] (5/8) Epoch 17, batch 50, loss[loss=0.2424, simple_loss=0.3214, pruned_loss=0.08169, over 15598.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2778, pruned_loss=0.05539, over 741854.88 frames. ], batch size: 190, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 10:59:06,524 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:59:41,404 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7328, 2.8255, 2.8496, 4.8788, 3.9541, 4.4047, 1.7208, 3.2105], device='cuda:5'), covar=tensor([0.1518, 0.0782, 0.1151, 0.0169, 0.0242, 0.0352, 0.1622, 0.0783], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0162, 0.0185, 0.0164, 0.0190, 0.0206, 0.0191, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 10:59:47,960 INFO [train.py:904] (5/8) Epoch 17, batch 100, loss[loss=0.1998, simple_loss=0.2817, pruned_loss=0.0589, over 16284.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2703, pruned_loss=0.04996, over 1311505.50 frames. ], batch size: 165, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 10:59:58,039 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 11:00:00,627 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:00:12,329 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:00:22,623 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9815, 3.9977, 4.4659, 2.1670, 4.6346, 4.7455, 3.4478, 3.6822], device='cuda:5'), covar=tensor([0.0766, 0.0230, 0.0219, 0.1277, 0.0070, 0.0139, 0.0389, 0.0379], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0105, 0.0090, 0.0138, 0.0072, 0.0115, 0.0125, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 11:00:26,387 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.355e+02 2.682e+02 3.198e+02 6.473e+02, threshold=5.364e+02, percent-clipped=1.0 2023-04-30 11:00:56,546 INFO [train.py:904] (5/8) Epoch 17, batch 150, loss[loss=0.1834, simple_loss=0.2576, pruned_loss=0.05459, over 16798.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2679, pruned_loss=0.04927, over 1745239.31 frames. ], batch size: 102, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:01:23,585 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162572.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 11:01:44,475 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:02:05,945 INFO [train.py:904] (5/8) Epoch 17, batch 200, loss[loss=0.1736, simple_loss=0.252, pruned_loss=0.04765, over 16465.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2675, pruned_loss=0.048, over 2104726.50 frames. ], batch size: 146, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:02:25,609 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2895, 3.1981, 3.4208, 2.4786, 3.1086, 3.5262, 3.2438, 2.0448], device='cuda:5'), covar=tensor([0.0461, 0.0160, 0.0057, 0.0343, 0.0109, 0.0090, 0.0092, 0.0436], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0077, 0.0076, 0.0132, 0.0089, 0.0099, 0.0087, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 11:02:43,591 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.407e+02 2.768e+02 3.167e+02 5.394e+02, threshold=5.535e+02, percent-clipped=0.0 2023-04-30 11:03:12,330 INFO [train.py:904] (5/8) Epoch 17, batch 250, loss[loss=0.1494, simple_loss=0.2411, pruned_loss=0.02886, over 16850.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2657, pruned_loss=0.0487, over 2372457.81 frames. ], batch size: 42, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:03:22,244 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7446, 5.1099, 4.8853, 4.8794, 4.6422, 4.5917, 4.5810, 5.1976], device='cuda:5'), covar=tensor([0.1391, 0.1030, 0.1203, 0.0869, 0.1005, 0.1049, 0.1200, 0.0980], device='cuda:5'), in_proj_covar=tensor([0.0608, 0.0750, 0.0609, 0.0551, 0.0473, 0.0482, 0.0626, 0.0577], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:03:39,322 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1676, 2.0540, 2.2389, 3.8150, 2.1457, 2.4095, 2.1580, 2.2244], device='cuda:5'), covar=tensor([0.1263, 0.3677, 0.2835, 0.0606, 0.3732, 0.2539, 0.3499, 0.3014], device='cuda:5'), in_proj_covar=tensor([0.0376, 0.0415, 0.0350, 0.0316, 0.0423, 0.0477, 0.0387, 0.0484], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:04:20,338 INFO [train.py:904] (5/8) Epoch 17, batch 300, loss[loss=0.1797, simple_loss=0.2662, pruned_loss=0.0466, over 12515.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2639, pruned_loss=0.04798, over 2579301.54 frames. ], batch size: 246, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:04:59,736 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.237e+02 2.646e+02 3.133e+02 7.879e+02, threshold=5.293e+02, percent-clipped=2.0 2023-04-30 11:05:03,773 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162733.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:05:29,589 INFO [train.py:904] (5/8) Epoch 17, batch 350, loss[loss=0.1537, simple_loss=0.255, pruned_loss=0.02621, over 17254.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.261, pruned_loss=0.04685, over 2743156.52 frames. ], batch size: 52, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:06:07,975 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162781.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:06:36,764 INFO [train.py:904] (5/8) Epoch 17, batch 400, loss[loss=0.1452, simple_loss=0.2326, pruned_loss=0.02894, over 16865.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2598, pruned_loss=0.04632, over 2872149.12 frames. ], batch size: 42, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:07:06,310 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6444, 2.5828, 2.1350, 2.4282, 3.0269, 2.8130, 3.3296, 3.2040], device='cuda:5'), covar=tensor([0.0129, 0.0412, 0.0536, 0.0450, 0.0247, 0.0334, 0.0247, 0.0266], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0227, 0.0218, 0.0218, 0.0225, 0.0226, 0.0228, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:07:10,989 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:07:15,901 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.145e+02 2.685e+02 3.328e+02 1.079e+03, threshold=5.370e+02, percent-clipped=6.0 2023-04-30 11:07:46,592 INFO [train.py:904] (5/8) Epoch 17, batch 450, loss[loss=0.143, simple_loss=0.2342, pruned_loss=0.02596, over 17199.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2583, pruned_loss=0.04542, over 2974444.95 frames. ], batch size: 44, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:07:54,505 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7472, 3.1636, 3.3122, 5.0368, 4.3979, 4.6372, 1.5255, 3.4510], device='cuda:5'), covar=tensor([0.1353, 0.0638, 0.0860, 0.0146, 0.0200, 0.0306, 0.1577, 0.0657], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0164, 0.0186, 0.0168, 0.0195, 0.0209, 0.0192, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:08:06,782 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162867.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 11:08:34,336 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:35,641 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:55,342 INFO [train.py:904] (5/8) Epoch 17, batch 500, loss[loss=0.182, simple_loss=0.2557, pruned_loss=0.0541, over 16852.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2569, pruned_loss=0.04504, over 3050139.47 frames. ], batch size: 102, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:08:57,315 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 11:09:32,595 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.133e+02 2.499e+02 3.161e+02 5.064e+02, threshold=4.998e+02, percent-clipped=0.0 2023-04-30 11:09:39,867 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:10:01,809 INFO [train.py:904] (5/8) Epoch 17, batch 550, loss[loss=0.1745, simple_loss=0.2634, pruned_loss=0.04284, over 17018.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2562, pruned_loss=0.04494, over 3091203.59 frames. ], batch size: 55, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:11:10,226 INFO [train.py:904] (5/8) Epoch 17, batch 600, loss[loss=0.1463, simple_loss=0.2355, pruned_loss=0.02858, over 16982.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2561, pruned_loss=0.04492, over 3148437.55 frames. ], batch size: 41, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:11:47,297 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.256e+02 2.602e+02 3.226e+02 5.329e+02, threshold=5.204e+02, percent-clipped=2.0 2023-04-30 11:12:08,798 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-30 11:12:16,985 INFO [train.py:904] (5/8) Epoch 17, batch 650, loss[loss=0.1495, simple_loss=0.2321, pruned_loss=0.03346, over 16856.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2552, pruned_loss=0.0443, over 3186572.90 frames. ], batch size: 102, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:12:17,491 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9549, 3.0423, 2.7894, 4.4253, 3.7489, 4.3012, 1.6993, 3.0185], device='cuda:5'), covar=tensor([0.1259, 0.0609, 0.1042, 0.0189, 0.0198, 0.0378, 0.1511, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0164, 0.0186, 0.0169, 0.0196, 0.0210, 0.0192, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:13:05,976 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6239, 4.7844, 4.9615, 4.8039, 4.8358, 5.4393, 4.9568, 4.6384], device='cuda:5'), covar=tensor([0.1477, 0.2249, 0.2545, 0.2166, 0.2871, 0.1200, 0.1758, 0.2543], device='cuda:5'), in_proj_covar=tensor([0.0389, 0.0564, 0.0619, 0.0474, 0.0634, 0.0653, 0.0490, 0.0635], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 11:13:09,421 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:13:17,145 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-30 11:13:25,536 INFO [train.py:904] (5/8) Epoch 17, batch 700, loss[loss=0.1591, simple_loss=0.249, pruned_loss=0.03461, over 17210.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2551, pruned_loss=0.0441, over 3209662.95 frames. ], batch size: 45, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:14:04,490 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.287e+02 2.752e+02 3.270e+02 5.446e+02, threshold=5.503e+02, percent-clipped=2.0 2023-04-30 11:14:34,419 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1939, 4.1094, 4.5260, 2.0528, 4.6750, 4.7817, 3.5140, 3.6154], device='cuda:5'), covar=tensor([0.0643, 0.0234, 0.0211, 0.1225, 0.0078, 0.0153, 0.0367, 0.0384], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0106, 0.0093, 0.0139, 0.0074, 0.0118, 0.0126, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 11:14:34,450 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163151.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:14:35,151 INFO [train.py:904] (5/8) Epoch 17, batch 750, loss[loss=0.1381, simple_loss=0.2227, pruned_loss=0.02674, over 16990.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2549, pruned_loss=0.04352, over 3237550.64 frames. ], batch size: 41, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:14:57,150 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163167.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:15:18,768 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:15:44,399 INFO [train.py:904] (5/8) Epoch 17, batch 800, loss[loss=0.164, simple_loss=0.2551, pruned_loss=0.0365, over 17065.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.255, pruned_loss=0.04392, over 3258922.68 frames. ], batch size: 53, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:16:03,279 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163215.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:16:23,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 2.209e+02 2.648e+02 3.284e+02 7.557e+02, threshold=5.297e+02, percent-clipped=2.0 2023-04-30 11:16:36,586 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163240.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:16:53,807 INFO [train.py:904] (5/8) Epoch 17, batch 850, loss[loss=0.1716, simple_loss=0.2663, pruned_loss=0.03843, over 17144.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2547, pruned_loss=0.04337, over 3273339.02 frames. ], batch size: 48, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:18:01,143 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163301.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:18:01,854 INFO [train.py:904] (5/8) Epoch 17, batch 900, loss[loss=0.149, simple_loss=0.2329, pruned_loss=0.03257, over 16760.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2539, pruned_loss=0.04323, over 3283723.63 frames. ], batch size: 39, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:18:26,602 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5375, 4.5958, 4.9258, 4.9159, 4.9577, 4.6227, 4.6376, 4.4426], device='cuda:5'), covar=tensor([0.0350, 0.0599, 0.0376, 0.0383, 0.0413, 0.0391, 0.0800, 0.0530], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0418, 0.0408, 0.0383, 0.0456, 0.0431, 0.0525, 0.0341], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 11:18:40,398 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.201e+02 2.666e+02 3.227e+02 4.456e+02, threshold=5.332e+02, percent-clipped=0.0 2023-04-30 11:18:42,642 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6225, 2.5335, 2.1900, 2.5598, 2.9029, 2.7085, 3.2401, 3.1432], device='cuda:5'), covar=tensor([0.0119, 0.0400, 0.0500, 0.0424, 0.0287, 0.0381, 0.0281, 0.0270], device='cuda:5'), in_proj_covar=tensor([0.0190, 0.0230, 0.0221, 0.0220, 0.0228, 0.0230, 0.0234, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:18:52,645 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 11:19:09,575 INFO [train.py:904] (5/8) Epoch 17, batch 950, loss[loss=0.1855, simple_loss=0.2561, pruned_loss=0.05743, over 16736.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2541, pruned_loss=0.04347, over 3300884.79 frames. ], batch size: 134, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:19:33,950 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0927, 5.0840, 5.5740, 5.4987, 5.5544, 5.2129, 5.1823, 4.9375], device='cuda:5'), covar=tensor([0.0338, 0.0507, 0.0331, 0.0472, 0.0472, 0.0347, 0.0918, 0.0422], device='cuda:5'), in_proj_covar=tensor([0.0388, 0.0419, 0.0411, 0.0385, 0.0459, 0.0432, 0.0528, 0.0344], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 11:20:00,149 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6898, 3.8563, 2.3792, 4.5369, 2.9835, 4.5225, 2.5575, 3.1254], device='cuda:5'), covar=tensor([0.0305, 0.0393, 0.1574, 0.0216, 0.0865, 0.0430, 0.1481, 0.0748], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0171, 0.0192, 0.0152, 0.0173, 0.0212, 0.0201, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:20:13,044 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 11:20:17,969 INFO [train.py:904] (5/8) Epoch 17, batch 1000, loss[loss=0.1592, simple_loss=0.2493, pruned_loss=0.03455, over 16651.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.254, pruned_loss=0.04349, over 3305266.88 frames. ], batch size: 62, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:20:54,829 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.252e+02 2.741e+02 3.139e+02 5.661e+02, threshold=5.483e+02, percent-clipped=2.0 2023-04-30 11:20:59,960 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7618, 4.2710, 3.0974, 2.2541, 2.6960, 2.4807, 4.5960, 3.6167], device='cuda:5'), covar=tensor([0.2865, 0.0508, 0.1603, 0.2753, 0.2703, 0.1922, 0.0334, 0.1217], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0262, 0.0295, 0.0298, 0.0286, 0.0241, 0.0281, 0.0320], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 11:21:18,978 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:21:26,451 INFO [train.py:904] (5/8) Epoch 17, batch 1050, loss[loss=0.1766, simple_loss=0.2688, pruned_loss=0.04218, over 17057.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2542, pruned_loss=0.04311, over 3318460.51 frames. ], batch size: 55, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:22:10,637 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163483.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:22:36,973 INFO [train.py:904] (5/8) Epoch 17, batch 1100, loss[loss=0.1906, simple_loss=0.2599, pruned_loss=0.06064, over 16855.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2537, pruned_loss=0.04301, over 3326658.82 frames. ], batch size: 116, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:22:40,145 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-30 11:23:08,694 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5175, 4.5637, 4.4527, 4.1650, 3.7396, 4.6622, 4.5520, 4.2534], device='cuda:5'), covar=tensor([0.0926, 0.0846, 0.0515, 0.0479, 0.1873, 0.0571, 0.0536, 0.0759], device='cuda:5'), in_proj_covar=tensor([0.0282, 0.0392, 0.0329, 0.0319, 0.0340, 0.0369, 0.0226, 0.0395], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:23:14,824 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.362e+02 2.061e+02 2.430e+02 2.825e+02 4.071e+02, threshold=4.861e+02, percent-clipped=0.0 2023-04-30 11:23:16,256 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:23:43,929 INFO [train.py:904] (5/8) Epoch 17, batch 1150, loss[loss=0.1763, simple_loss=0.2716, pruned_loss=0.04048, over 17012.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2522, pruned_loss=0.04217, over 3330594.97 frames. ], batch size: 55, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:23:46,872 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6764, 3.7494, 2.2231, 3.9278, 2.9074, 3.9129, 2.4429, 3.0596], device='cuda:5'), covar=tensor([0.0211, 0.0329, 0.1424, 0.0309, 0.0663, 0.0756, 0.1187, 0.0541], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0154, 0.0174, 0.0214, 0.0201, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:24:20,084 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 11:24:43,085 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:24:52,267 INFO [train.py:904] (5/8) Epoch 17, batch 1200, loss[loss=0.1631, simple_loss=0.2446, pruned_loss=0.04079, over 16788.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2516, pruned_loss=0.0417, over 3335050.94 frames. ], batch size: 42, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:24:58,768 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9606, 4.0914, 4.3923, 4.3723, 4.3989, 4.0731, 4.1636, 4.0555], device='cuda:5'), covar=tensor([0.0431, 0.0677, 0.0409, 0.0426, 0.0468, 0.0489, 0.0762, 0.0606], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0422, 0.0413, 0.0388, 0.0460, 0.0436, 0.0532, 0.0345], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 11:25:29,956 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.305e+02 2.671e+02 3.522e+02 8.114e+02, threshold=5.342e+02, percent-clipped=7.0 2023-04-30 11:25:58,589 INFO [train.py:904] (5/8) Epoch 17, batch 1250, loss[loss=0.1628, simple_loss=0.2438, pruned_loss=0.04094, over 16910.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2511, pruned_loss=0.04155, over 3328819.85 frames. ], batch size: 90, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:06,159 INFO [train.py:904] (5/8) Epoch 17, batch 1300, loss[loss=0.149, simple_loss=0.2317, pruned_loss=0.03319, over 16986.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.251, pruned_loss=0.04201, over 3331273.90 frames. ], batch size: 41, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:45,002 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.213e+02 2.483e+02 3.083e+02 7.276e+02, threshold=4.967e+02, percent-clipped=1.0 2023-04-30 11:28:08,510 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:28:16,452 INFO [train.py:904] (5/8) Epoch 17, batch 1350, loss[loss=0.1433, simple_loss=0.2247, pruned_loss=0.03099, over 16760.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2507, pruned_loss=0.04219, over 3317896.75 frames. ], batch size: 39, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:28:49,113 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-30 11:29:15,219 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:29:25,096 INFO [train.py:904] (5/8) Epoch 17, batch 1400, loss[loss=0.1586, simple_loss=0.2472, pruned_loss=0.03498, over 16834.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2512, pruned_loss=0.04195, over 3317714.02 frames. ], batch size: 42, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:29:52,521 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4927, 3.3438, 2.7317, 2.1623, 2.2315, 2.2751, 3.4620, 3.0230], device='cuda:5'), covar=tensor([0.2682, 0.0696, 0.1670, 0.2871, 0.2676, 0.2006, 0.0530, 0.1441], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0263, 0.0296, 0.0298, 0.0287, 0.0243, 0.0282, 0.0322], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 11:30:05,130 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.111e+02 2.569e+02 3.362e+02 6.310e+02, threshold=5.138e+02, percent-clipped=5.0 2023-04-30 11:30:36,595 INFO [train.py:904] (5/8) Epoch 17, batch 1450, loss[loss=0.1742, simple_loss=0.2459, pruned_loss=0.05126, over 16513.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2503, pruned_loss=0.04192, over 3327040.16 frames. ], batch size: 75, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:31:26,388 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0943, 5.0269, 4.9851, 4.5015, 4.5130, 5.0150, 4.9884, 4.6324], device='cuda:5'), covar=tensor([0.0590, 0.0478, 0.0310, 0.0359, 0.1261, 0.0443, 0.0374, 0.0768], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0395, 0.0331, 0.0322, 0.0344, 0.0373, 0.0228, 0.0398], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:31:38,089 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163896.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:31:45,966 INFO [train.py:904] (5/8) Epoch 17, batch 1500, loss[loss=0.1693, simple_loss=0.2552, pruned_loss=0.04175, over 16835.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2504, pruned_loss=0.04232, over 3325918.31 frames. ], batch size: 102, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:32:24,626 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 2.188e+02 2.579e+02 3.200e+02 4.820e+02, threshold=5.157e+02, percent-clipped=0.0 2023-04-30 11:32:45,102 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163944.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:32:56,205 INFO [train.py:904] (5/8) Epoch 17, batch 1550, loss[loss=0.1757, simple_loss=0.2663, pruned_loss=0.04255, over 17136.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2515, pruned_loss=0.04329, over 3327816.42 frames. ], batch size: 47, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:33:00,262 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2415, 2.1666, 2.2872, 4.0081, 2.2463, 2.5746, 2.2244, 2.3525], device='cuda:5'), covar=tensor([0.1268, 0.3511, 0.2711, 0.0579, 0.3579, 0.2311, 0.3538, 0.2956], device='cuda:5'), in_proj_covar=tensor([0.0388, 0.0426, 0.0360, 0.0328, 0.0432, 0.0493, 0.0397, 0.0501], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:33:15,038 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 11:34:07,075 INFO [train.py:904] (5/8) Epoch 17, batch 1600, loss[loss=0.1701, simple_loss=0.2473, pruned_loss=0.04651, over 16828.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2524, pruned_loss=0.04342, over 3330001.88 frames. ], batch size: 83, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:34:45,105 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.274e+02 2.815e+02 3.289e+02 5.501e+02, threshold=5.629e+02, percent-clipped=1.0 2023-04-30 11:35:10,884 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9263, 1.9716, 2.4919, 2.9079, 2.6359, 3.3249, 2.2892, 3.3366], device='cuda:5'), covar=tensor([0.0224, 0.0467, 0.0317, 0.0280, 0.0318, 0.0174, 0.0419, 0.0134], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0187, 0.0175, 0.0178, 0.0187, 0.0145, 0.0188, 0.0137], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:35:15,614 INFO [train.py:904] (5/8) Epoch 17, batch 1650, loss[loss=0.1672, simple_loss=0.2479, pruned_loss=0.04327, over 16530.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2541, pruned_loss=0.04389, over 3320218.46 frames. ], batch size: 75, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:35:39,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8272, 4.3116, 4.4229, 3.2385, 3.6983, 4.3303, 3.9409, 2.5876], device='cuda:5'), covar=tensor([0.0421, 0.0073, 0.0036, 0.0287, 0.0115, 0.0073, 0.0076, 0.0402], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0079, 0.0078, 0.0133, 0.0091, 0.0102, 0.0090, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 11:36:08,328 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:36:24,917 INFO [train.py:904] (5/8) Epoch 17, batch 1700, loss[loss=0.1767, simple_loss=0.2529, pruned_loss=0.05025, over 16403.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2567, pruned_loss=0.04483, over 3317613.16 frames. ], batch size: 146, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:36:40,708 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1191, 5.1332, 5.6432, 5.5260, 5.5996, 5.1933, 5.1757, 4.9340], device='cuda:5'), covar=tensor([0.0345, 0.0527, 0.0317, 0.0443, 0.0445, 0.0401, 0.0943, 0.0453], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0425, 0.0415, 0.0390, 0.0462, 0.0438, 0.0535, 0.0346], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 11:36:50,144 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0889, 4.9830, 4.9472, 4.4432, 4.5089, 4.9679, 4.9411, 4.6033], device='cuda:5'), covar=tensor([0.0592, 0.0564, 0.0313, 0.0374, 0.1235, 0.0484, 0.0379, 0.0810], device='cuda:5'), in_proj_covar=tensor([0.0287, 0.0400, 0.0335, 0.0328, 0.0350, 0.0378, 0.0230, 0.0404], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 11:37:01,953 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.209e+02 2.796e+02 3.472e+02 6.333e+02, threshold=5.591e+02, percent-clipped=2.0 2023-04-30 11:37:02,633 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3835, 2.2062, 2.2969, 4.1222, 2.1356, 2.5049, 2.2954, 2.3428], device='cuda:5'), covar=tensor([0.1230, 0.3711, 0.2820, 0.0529, 0.4288, 0.2635, 0.3526, 0.3746], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0427, 0.0360, 0.0327, 0.0433, 0.0493, 0.0397, 0.0501], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:37:06,581 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9718, 4.0453, 2.1809, 4.8277, 3.1036, 4.7419, 2.5164, 3.2959], device='cuda:5'), covar=tensor([0.0285, 0.0375, 0.1864, 0.0244, 0.0784, 0.0321, 0.1707, 0.0739], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0156, 0.0175, 0.0217, 0.0203, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:37:13,387 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9593, 2.4605, 2.6073, 1.8465, 2.6444, 2.7814, 2.4131, 2.3090], device='cuda:5'), covar=tensor([0.0759, 0.0228, 0.0237, 0.0994, 0.0121, 0.0274, 0.0466, 0.0436], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0107, 0.0094, 0.0139, 0.0075, 0.0121, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 11:37:32,303 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164151.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:37:32,940 INFO [train.py:904] (5/8) Epoch 17, batch 1750, loss[loss=0.1778, simple_loss=0.2574, pruned_loss=0.04905, over 16814.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2583, pruned_loss=0.04496, over 3307462.78 frames. ], batch size: 83, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:37:36,767 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:38:05,330 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 11:38:38,321 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5287, 5.9400, 5.6423, 5.7197, 5.3492, 5.2680, 5.3390, 6.0537], device='cuda:5'), covar=tensor([0.1300, 0.0882, 0.1003, 0.0811, 0.0808, 0.0655, 0.1250, 0.0837], device='cuda:5'), in_proj_covar=tensor([0.0646, 0.0792, 0.0647, 0.0588, 0.0503, 0.0506, 0.0660, 0.0613], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:38:41,951 INFO [train.py:904] (5/8) Epoch 17, batch 1800, loss[loss=0.1695, simple_loss=0.2536, pruned_loss=0.04272, over 16790.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2596, pruned_loss=0.04523, over 3317286.69 frames. ], batch size: 102, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:38:58,232 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0756, 4.2409, 4.1819, 3.1475, 3.5231, 4.1433, 3.8989, 2.0841], device='cuda:5'), covar=tensor([0.0410, 0.0086, 0.0059, 0.0349, 0.0150, 0.0147, 0.0123, 0.0588], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0078, 0.0077, 0.0132, 0.0091, 0.0101, 0.0089, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 11:39:00,561 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7823, 4.0261, 2.5080, 4.5444, 3.0387, 4.5548, 2.6741, 3.2187], device='cuda:5'), covar=tensor([0.0292, 0.0354, 0.1465, 0.0250, 0.0798, 0.0507, 0.1295, 0.0719], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0157, 0.0175, 0.0218, 0.0204, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:39:01,668 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:39:19,767 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.179e+02 2.567e+02 3.142e+02 4.810e+02, threshold=5.135e+02, percent-clipped=0.0 2023-04-30 11:39:49,958 INFO [train.py:904] (5/8) Epoch 17, batch 1850, loss[loss=0.2045, simple_loss=0.2692, pruned_loss=0.06994, over 16909.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.261, pruned_loss=0.04561, over 3320052.83 frames. ], batch size: 109, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:41:01,367 INFO [train.py:904] (5/8) Epoch 17, batch 1900, loss[loss=0.1832, simple_loss=0.259, pruned_loss=0.05369, over 16876.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2592, pruned_loss=0.04467, over 3316224.34 frames. ], batch size: 109, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:41:06,092 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 11:41:26,286 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3138, 2.5770, 2.1135, 2.2867, 2.9373, 2.5809, 3.1063, 3.0361], device='cuda:5'), covar=tensor([0.0176, 0.0363, 0.0488, 0.0440, 0.0244, 0.0355, 0.0262, 0.0252], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0231, 0.0222, 0.0221, 0.0230, 0.0232, 0.0236, 0.0225], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:41:40,563 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7522, 2.6589, 2.3956, 2.7204, 3.0600, 2.8379, 3.4749, 3.2644], device='cuda:5'), covar=tensor([0.0115, 0.0395, 0.0442, 0.0388, 0.0256, 0.0358, 0.0193, 0.0245], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0230, 0.0221, 0.0221, 0.0230, 0.0231, 0.0236, 0.0225], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:41:41,206 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.257e+02 2.647e+02 3.150e+02 4.785e+02, threshold=5.294e+02, percent-clipped=0.0 2023-04-30 11:42:09,172 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9584, 2.9492, 2.6245, 2.8616, 3.2645, 3.0458, 3.6736, 3.4572], device='cuda:5'), covar=tensor([0.0107, 0.0325, 0.0401, 0.0347, 0.0234, 0.0308, 0.0203, 0.0222], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0230, 0.0221, 0.0221, 0.0230, 0.0231, 0.0236, 0.0225], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:42:12,329 INFO [train.py:904] (5/8) Epoch 17, batch 1950, loss[loss=0.1726, simple_loss=0.2514, pruned_loss=0.04695, over 16814.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2591, pruned_loss=0.04384, over 3322587.03 frames. ], batch size: 83, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:42:24,813 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:42:48,949 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 11:42:56,755 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2574, 3.0699, 3.3088, 1.7219, 3.4909, 3.4374, 2.7985, 2.5506], device='cuda:5'), covar=tensor([0.0823, 0.0261, 0.0222, 0.1198, 0.0101, 0.0196, 0.0472, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0108, 0.0094, 0.0140, 0.0075, 0.0121, 0.0127, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 11:43:23,965 INFO [train.py:904] (5/8) Epoch 17, batch 2000, loss[loss=0.1797, simple_loss=0.2783, pruned_loss=0.04054, over 16718.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2587, pruned_loss=0.04371, over 3332062.10 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:43:51,039 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:44:02,063 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.498e+02 2.950e+02 3.427e+02 6.648e+02, threshold=5.900e+02, percent-clipped=3.0 2023-04-30 11:44:25,527 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:44:32,498 INFO [train.py:904] (5/8) Epoch 17, batch 2050, loss[loss=0.161, simple_loss=0.2538, pruned_loss=0.03413, over 17121.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2584, pruned_loss=0.04383, over 3340935.40 frames. ], batch size: 47, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:44:53,817 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 11:45:07,638 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-04-30 11:45:41,566 INFO [train.py:904] (5/8) Epoch 17, batch 2100, loss[loss=0.1716, simple_loss=0.2562, pruned_loss=0.04347, over 16844.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2598, pruned_loss=0.04477, over 3341359.89 frames. ], batch size: 96, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:45:55,062 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:46:09,071 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7290, 2.4726, 2.3408, 3.3450, 2.8125, 3.6319, 1.5094, 2.7247], device='cuda:5'), covar=tensor([0.1270, 0.0688, 0.1133, 0.0181, 0.0177, 0.0377, 0.1503, 0.0806], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0165, 0.0187, 0.0172, 0.0198, 0.0211, 0.0191, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:46:20,656 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.320e+02 2.783e+02 3.356e+02 6.919e+02, threshold=5.567e+02, percent-clipped=1.0 2023-04-30 11:46:38,713 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 11:46:50,967 INFO [train.py:904] (5/8) Epoch 17, batch 2150, loss[loss=0.1868, simple_loss=0.2623, pruned_loss=0.05562, over 16337.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2598, pruned_loss=0.04471, over 3336725.82 frames. ], batch size: 165, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:47:13,626 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:27,063 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:31,416 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 11:47:58,405 INFO [train.py:904] (5/8) Epoch 17, batch 2200, loss[loss=0.1655, simple_loss=0.2677, pruned_loss=0.03163, over 17085.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2608, pruned_loss=0.04533, over 3325762.92 frames. ], batch size: 48, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:48:36,644 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:48:37,258 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.246e+02 2.692e+02 3.358e+02 7.856e+02, threshold=5.383e+02, percent-clipped=4.0 2023-04-30 11:48:44,240 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9621, 4.5839, 4.6099, 3.3626, 3.7963, 4.5699, 4.1151, 2.8184], device='cuda:5'), covar=tensor([0.0426, 0.0050, 0.0032, 0.0298, 0.0108, 0.0062, 0.0077, 0.0380], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0077, 0.0077, 0.0132, 0.0091, 0.0101, 0.0089, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 11:48:44,517 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-04-30 11:48:50,039 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164639.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 11:48:51,384 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-04-30 11:48:58,214 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1757, 4.1604, 4.5605, 2.4547, 4.7605, 4.7812, 3.3227, 3.7606], device='cuda:5'), covar=tensor([0.0664, 0.0224, 0.0195, 0.1040, 0.0065, 0.0178, 0.0450, 0.0344], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0108, 0.0095, 0.0139, 0.0075, 0.0122, 0.0127, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 11:49:02,753 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 11:49:06,807 INFO [train.py:904] (5/8) Epoch 17, batch 2250, loss[loss=0.1919, simple_loss=0.2836, pruned_loss=0.05011, over 16972.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2608, pruned_loss=0.04502, over 3320174.18 frames. ], batch size: 55, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:49:15,036 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164658.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:15,843 INFO [train.py:904] (5/8) Epoch 17, batch 2300, loss[loss=0.1778, simple_loss=0.2548, pruned_loss=0.05042, over 16264.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2607, pruned_loss=0.04515, over 3316728.06 frames. ], batch size: 165, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:50:35,156 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:38,862 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:53,239 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.267e+02 2.732e+02 3.248e+02 6.270e+02, threshold=5.465e+02, percent-clipped=1.0 2023-04-30 11:51:16,716 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:16,826 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:24,018 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4114, 2.3233, 2.2584, 4.3392, 2.2325, 2.6971, 2.3751, 2.4318], device='cuda:5'), covar=tensor([0.1258, 0.3671, 0.2962, 0.0558, 0.4059, 0.2633, 0.3484, 0.3651], device='cuda:5'), in_proj_covar=tensor([0.0390, 0.0429, 0.0359, 0.0327, 0.0431, 0.0496, 0.0398, 0.0501], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:51:24,576 INFO [train.py:904] (5/8) Epoch 17, batch 2350, loss[loss=0.1804, simple_loss=0.2737, pruned_loss=0.04353, over 16716.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2616, pruned_loss=0.04522, over 3312900.02 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:23,528 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=164794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:24,230 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 11:52:34,378 INFO [train.py:904] (5/8) Epoch 17, batch 2400, loss[loss=0.1842, simple_loss=0.2644, pruned_loss=0.05199, over 16440.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.262, pruned_loss=0.04574, over 3323097.62 frames. ], batch size: 146, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:41,607 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:46,697 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:53:12,672 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.216e+02 2.538e+02 3.112e+02 6.074e+02, threshold=5.077e+02, percent-clipped=1.0 2023-04-30 11:53:15,528 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7716, 2.9127, 2.6138, 4.7006, 3.7905, 4.1442, 1.7132, 3.1097], device='cuda:5'), covar=tensor([0.1339, 0.0713, 0.1167, 0.0218, 0.0320, 0.0430, 0.1553, 0.0821], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0164, 0.0185, 0.0172, 0.0198, 0.0210, 0.0190, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:53:19,608 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6158, 3.6392, 2.1552, 3.8949, 2.8383, 3.8421, 2.2993, 2.9434], device='cuda:5'), covar=tensor([0.0227, 0.0350, 0.1404, 0.0279, 0.0693, 0.0651, 0.1257, 0.0638], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0157, 0.0173, 0.0217, 0.0202, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:53:41,596 INFO [train.py:904] (5/8) Epoch 17, batch 2450, loss[loss=0.2011, simple_loss=0.2875, pruned_loss=0.05729, over 12358.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2631, pruned_loss=0.04585, over 3314157.05 frames. ], batch size: 246, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:53:47,302 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6452, 6.0666, 5.7690, 5.8398, 5.4258, 5.2965, 5.4568, 6.1640], device='cuda:5'), covar=tensor([0.1205, 0.0788, 0.0909, 0.0812, 0.0922, 0.0780, 0.1070, 0.0939], device='cuda:5'), in_proj_covar=tensor([0.0647, 0.0797, 0.0649, 0.0593, 0.0506, 0.0509, 0.0662, 0.0618], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:53:51,212 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=164859.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:54:46,730 INFO [train.py:904] (5/8) Epoch 17, batch 2500, loss[loss=0.1976, simple_loss=0.2802, pruned_loss=0.05754, over 16665.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2638, pruned_loss=0.04567, over 3319211.16 frames. ], batch size: 62, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:55:12,211 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4819, 2.3774, 2.3293, 4.3205, 2.2810, 2.7103, 2.4140, 2.4776], device='cuda:5'), covar=tensor([0.1185, 0.3285, 0.2714, 0.0472, 0.3879, 0.2438, 0.3149, 0.3400], device='cuda:5'), in_proj_covar=tensor([0.0388, 0.0427, 0.0357, 0.0326, 0.0429, 0.0494, 0.0396, 0.0500], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:55:17,568 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:55:26,644 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.192e+02 2.472e+02 2.956e+02 6.595e+02, threshold=4.945e+02, percent-clipped=3.0 2023-04-30 11:55:29,675 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 11:55:30,897 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164934.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 11:55:55,244 INFO [train.py:904] (5/8) Epoch 17, batch 2550, loss[loss=0.2039, simple_loss=0.2921, pruned_loss=0.05782, over 16094.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2648, pruned_loss=0.04572, over 3307983.25 frames. ], batch size: 35, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:02,096 INFO [train.py:904] (5/8) Epoch 17, batch 2600, loss[loss=0.2059, simple_loss=0.2838, pruned_loss=0.064, over 16761.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2646, pruned_loss=0.04577, over 3314498.64 frames. ], batch size: 124, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:17,929 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165014.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:20,608 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165016.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:31,845 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0000, 3.3404, 3.1342, 5.2292, 4.3149, 4.5532, 2.0163, 3.4500], device='cuda:5'), covar=tensor([0.1254, 0.0644, 0.1015, 0.0172, 0.0272, 0.0388, 0.1425, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0166, 0.0187, 0.0174, 0.0199, 0.0212, 0.0192, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 11:57:41,410 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.187e+02 2.619e+02 3.229e+02 7.768e+02, threshold=5.237e+02, percent-clipped=3.0 2023-04-30 11:57:43,684 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7807, 5.1266, 4.8189, 4.8881, 4.6460, 4.5955, 4.5709, 5.2120], device='cuda:5'), covar=tensor([0.1230, 0.0891, 0.1092, 0.0828, 0.0901, 0.1171, 0.1220, 0.0958], device='cuda:5'), in_proj_covar=tensor([0.0647, 0.0800, 0.0651, 0.0594, 0.0508, 0.0509, 0.0663, 0.0618], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 11:58:08,472 INFO [train.py:904] (5/8) Epoch 17, batch 2650, loss[loss=0.1797, simple_loss=0.2641, pruned_loss=0.04764, over 16772.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2654, pruned_loss=0.04578, over 3323845.78 frames. ], batch size: 134, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:58:26,280 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:58:29,429 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 11:59:18,030 INFO [train.py:904] (5/8) Epoch 17, batch 2700, loss[loss=0.1652, simple_loss=0.2511, pruned_loss=0.03966, over 16985.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2648, pruned_loss=0.04504, over 3316037.26 frames. ], batch size: 41, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:59:18,372 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:47,111 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 11:59:47,944 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:59,035 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 2.045e+02 2.516e+02 2.952e+02 6.082e+02, threshold=5.032e+02, percent-clipped=3.0 2023-04-30 12:00:28,483 INFO [train.py:904] (5/8) Epoch 17, batch 2750, loss[loss=0.1863, simple_loss=0.2663, pruned_loss=0.05319, over 16681.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2636, pruned_loss=0.04444, over 3311519.28 frames. ], batch size: 83, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:01:13,268 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:01:38,387 INFO [train.py:904] (5/8) Epoch 17, batch 2800, loss[loss=0.1808, simple_loss=0.2769, pruned_loss=0.04241, over 16700.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2633, pruned_loss=0.04431, over 3309719.97 frames. ], batch size: 62, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:02:10,260 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165224.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:02:19,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.251e+02 2.601e+02 3.544e+02 6.990e+02, threshold=5.203e+02, percent-clipped=6.0 2023-04-30 12:02:24,182 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165234.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:02:48,658 INFO [train.py:904] (5/8) Epoch 17, batch 2850, loss[loss=0.1662, simple_loss=0.257, pruned_loss=0.03769, over 16432.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2619, pruned_loss=0.04402, over 3321585.72 frames. ], batch size: 68, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:03:18,272 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:31,439 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:58,724 INFO [train.py:904] (5/8) Epoch 17, batch 2900, loss[loss=0.1495, simple_loss=0.2301, pruned_loss=0.03449, over 16774.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2617, pruned_loss=0.04517, over 3307354.56 frames. ], batch size: 83, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:04:17,068 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165314.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:04:28,162 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 12:04:39,536 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.311e+02 2.666e+02 3.214e+02 5.469e+02, threshold=5.331e+02, percent-clipped=1.0 2023-04-30 12:05:09,246 INFO [train.py:904] (5/8) Epoch 17, batch 2950, loss[loss=0.1949, simple_loss=0.2773, pruned_loss=0.05626, over 16739.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2613, pruned_loss=0.04571, over 3310395.39 frames. ], batch size: 134, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:05:23,903 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165362.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:06:20,622 INFO [train.py:904] (5/8) Epoch 17, batch 3000, loss[loss=0.1645, simple_loss=0.2512, pruned_loss=0.03892, over 16826.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2615, pruned_loss=0.04603, over 3318249.77 frames. ], batch size: 42, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:06:20,622 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 12:06:29,129 INFO [train.py:938] (5/8) Epoch 17, validation: loss=0.1364, simple_loss=0.2422, pruned_loss=0.0153, over 944034.00 frames. 2023-04-30 12:06:29,130 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 12:06:29,625 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165402.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:07:09,775 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.395e+02 2.828e+02 3.269e+02 8.319e+02, threshold=5.656e+02, percent-clipped=3.0 2023-04-30 12:07:36,651 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165450.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:07:38,652 INFO [train.py:904] (5/8) Epoch 17, batch 3050, loss[loss=0.2008, simple_loss=0.276, pruned_loss=0.0628, over 16850.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2614, pruned_loss=0.0459, over 3326565.61 frames. ], batch size: 116, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:08:15,880 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165479.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:08:19,800 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5570, 4.3494, 4.5625, 4.7219, 4.8476, 4.3609, 4.6681, 4.8367], device='cuda:5'), covar=tensor([0.1469, 0.1162, 0.1302, 0.0620, 0.0660, 0.1247, 0.1986, 0.0721], device='cuda:5'), in_proj_covar=tensor([0.0637, 0.0787, 0.0939, 0.0796, 0.0597, 0.0634, 0.0638, 0.0743], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:08:45,838 INFO [train.py:904] (5/8) Epoch 17, batch 3100, loss[loss=0.1658, simple_loss=0.2613, pruned_loss=0.03511, over 16875.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2615, pruned_loss=0.04606, over 3322035.54 frames. ], batch size: 42, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:09:27,660 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.216e+02 2.589e+02 3.162e+02 6.656e+02, threshold=5.179e+02, percent-clipped=4.0 2023-04-30 12:09:55,045 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:09:55,707 INFO [train.py:904] (5/8) Epoch 17, batch 3150, loss[loss=0.1927, simple_loss=0.2749, pruned_loss=0.05524, over 12593.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2603, pruned_loss=0.04582, over 3316326.80 frames. ], batch size: 247, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:10:55,215 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-30 12:11:06,275 INFO [train.py:904] (5/8) Epoch 17, batch 3200, loss[loss=0.2114, simple_loss=0.2809, pruned_loss=0.07093, over 16686.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2593, pruned_loss=0.04517, over 3322833.69 frames. ], batch size: 134, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:11:21,069 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165612.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:11:49,642 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.164e+02 2.494e+02 3.096e+02 4.677e+02, threshold=4.988e+02, percent-clipped=0.0 2023-04-30 12:11:59,383 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6193, 2.2760, 2.3032, 4.4704, 2.2739, 2.7609, 2.3265, 2.4716], device='cuda:5'), covar=tensor([0.1143, 0.3766, 0.2894, 0.0430, 0.4161, 0.2421, 0.3586, 0.3725], device='cuda:5'), in_proj_covar=tensor([0.0394, 0.0431, 0.0359, 0.0329, 0.0432, 0.0498, 0.0401, 0.0505], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:12:15,440 INFO [train.py:904] (5/8) Epoch 17, batch 3250, loss[loss=0.1751, simple_loss=0.2676, pruned_loss=0.04131, over 17122.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2605, pruned_loss=0.04558, over 3319385.12 frames. ], batch size: 47, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:13:23,343 INFO [train.py:904] (5/8) Epoch 17, batch 3300, loss[loss=0.1759, simple_loss=0.2615, pruned_loss=0.04515, over 16485.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2618, pruned_loss=0.04625, over 3321094.79 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:13:37,702 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-30 12:14:00,845 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 12:14:06,779 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.306e+02 2.673e+02 3.210e+02 6.560e+02, threshold=5.346e+02, percent-clipped=3.0 2023-04-30 12:14:30,273 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165750.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:14:32,759 INFO [train.py:904] (5/8) Epoch 17, batch 3350, loss[loss=0.2006, simple_loss=0.2777, pruned_loss=0.06175, over 16721.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2618, pruned_loss=0.04561, over 3326464.16 frames. ], batch size: 134, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:11,143 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165779.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:42,364 INFO [train.py:904] (5/8) Epoch 17, batch 3400, loss[loss=0.1689, simple_loss=0.2505, pruned_loss=0.04362, over 16423.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.262, pruned_loss=0.04557, over 3327198.49 frames. ], batch size: 75, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:44,596 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:56,246 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:19,117 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:27,741 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.228e+02 2.640e+02 3.185e+02 6.119e+02, threshold=5.280e+02, percent-clipped=1.0 2023-04-30 12:16:39,568 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 12:16:54,385 INFO [train.py:904] (5/8) Epoch 17, batch 3450, loss[loss=0.1796, simple_loss=0.2785, pruned_loss=0.04033, over 16684.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.261, pruned_loss=0.04498, over 3315339.10 frames. ], batch size: 62, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:17:00,938 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 12:17:11,570 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165864.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:17:14,493 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5530, 4.4747, 4.6869, 4.4725, 4.4764, 5.1327, 4.6618, 4.3738], device='cuda:5'), covar=tensor([0.1617, 0.2230, 0.2277, 0.2224, 0.2937, 0.1148, 0.1634, 0.2668], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0572, 0.0629, 0.0484, 0.0651, 0.0663, 0.0497, 0.0648], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 12:18:05,616 INFO [train.py:904] (5/8) Epoch 17, batch 3500, loss[loss=0.1648, simple_loss=0.2483, pruned_loss=0.04058, over 16402.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2595, pruned_loss=0.04432, over 3312564.41 frames. ], batch size: 146, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:18:10,393 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2275, 4.9370, 5.1638, 5.3723, 5.6180, 4.8112, 5.5941, 5.5795], device='cuda:5'), covar=tensor([0.1848, 0.1124, 0.1811, 0.0815, 0.0511, 0.0841, 0.0455, 0.0586], device='cuda:5'), in_proj_covar=tensor([0.0640, 0.0790, 0.0945, 0.0803, 0.0602, 0.0639, 0.0640, 0.0748], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:18:13,076 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165907.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:18:25,225 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-04-30 12:18:49,911 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.135e+02 2.523e+02 3.067e+02 7.185e+02, threshold=5.046e+02, percent-clipped=1.0 2023-04-30 12:19:16,035 INFO [train.py:904] (5/8) Epoch 17, batch 3550, loss[loss=0.1731, simple_loss=0.2682, pruned_loss=0.03896, over 17079.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2581, pruned_loss=0.04394, over 3314830.73 frames. ], batch size: 55, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:20:27,938 INFO [train.py:904] (5/8) Epoch 17, batch 3600, loss[loss=0.1459, simple_loss=0.238, pruned_loss=0.02695, over 17217.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2567, pruned_loss=0.04322, over 3320422.57 frames. ], batch size: 45, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:20:53,399 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-30 12:21:11,694 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 12:21:12,109 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.087e+02 2.490e+02 2.982e+02 5.501e+02, threshold=4.979e+02, percent-clipped=1.0 2023-04-30 12:21:40,969 INFO [train.py:904] (5/8) Epoch 17, batch 3650, loss[loss=0.1732, simple_loss=0.2442, pruned_loss=0.05106, over 16895.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2564, pruned_loss=0.04424, over 3298424.61 frames. ], batch size: 96, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:21:44,553 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166054.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:22:55,848 INFO [train.py:904] (5/8) Epoch 17, batch 3700, loss[loss=0.1815, simple_loss=0.2567, pruned_loss=0.05316, over 11717.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2553, pruned_loss=0.04563, over 3284699.38 frames. ], batch size: 246, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:23:02,850 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:23:04,288 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2118, 2.6321, 2.2133, 2.3366, 3.0646, 2.7098, 3.2094, 3.1804], device='cuda:5'), covar=tensor([0.0161, 0.0356, 0.0449, 0.0406, 0.0222, 0.0294, 0.0207, 0.0223], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0231, 0.0221, 0.0223, 0.0233, 0.0232, 0.0237, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:23:16,000 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166115.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:23:42,388 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.852e+02 2.227e+02 2.592e+02 3.005e+02 5.986e+02, threshold=5.184e+02, percent-clipped=3.0 2023-04-30 12:24:03,607 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166147.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:24:10,331 INFO [train.py:904] (5/8) Epoch 17, batch 3750, loss[loss=0.1911, simple_loss=0.2725, pruned_loss=0.05484, over 16371.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2559, pruned_loss=0.04686, over 3281040.99 frames. ], batch size: 35, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:24:21,425 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:24:42,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4672, 4.4487, 4.7719, 4.7878, 4.8210, 4.5200, 4.5295, 4.3272], device='cuda:5'), covar=tensor([0.0306, 0.0736, 0.0409, 0.0402, 0.0519, 0.0382, 0.0808, 0.0592], device='cuda:5'), in_proj_covar=tensor([0.0389, 0.0427, 0.0416, 0.0392, 0.0464, 0.0438, 0.0535, 0.0348], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 12:25:22,525 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6912, 2.7553, 2.2248, 2.4583, 3.1599, 2.7925, 3.3783, 3.3503], device='cuda:5'), covar=tensor([0.0068, 0.0327, 0.0477, 0.0432, 0.0200, 0.0312, 0.0171, 0.0189], device='cuda:5'), in_proj_covar=tensor([0.0195, 0.0230, 0.0221, 0.0222, 0.0232, 0.0230, 0.0236, 0.0227], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:25:24,424 INFO [train.py:904] (5/8) Epoch 17, batch 3800, loss[loss=0.1839, simple_loss=0.2599, pruned_loss=0.05394, over 16879.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2563, pruned_loss=0.04803, over 3285692.08 frames. ], batch size: 116, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:25:32,323 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:33,780 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166208.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:48,021 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 12:26:10,619 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.250e+02 2.656e+02 3.137e+02 5.830e+02, threshold=5.312e+02, percent-clipped=1.0 2023-04-30 12:26:11,230 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1866, 1.5371, 1.9623, 2.1664, 2.3100, 2.3122, 1.6498, 2.2926], device='cuda:5'), covar=tensor([0.0206, 0.0436, 0.0254, 0.0274, 0.0261, 0.0261, 0.0452, 0.0137], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0189, 0.0175, 0.0179, 0.0189, 0.0146, 0.0189, 0.0140], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:26:38,506 INFO [train.py:904] (5/8) Epoch 17, batch 3850, loss[loss=0.1813, simple_loss=0.2519, pruned_loss=0.05535, over 16878.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2571, pruned_loss=0.04914, over 3277532.42 frames. ], batch size: 109, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:26:44,004 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166255.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:26:44,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4066, 3.9468, 4.0914, 2.8166, 3.6872, 4.1865, 3.8026, 2.4448], device='cuda:5'), covar=tensor([0.0477, 0.0174, 0.0040, 0.0339, 0.0078, 0.0076, 0.0065, 0.0409], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0076, 0.0075, 0.0129, 0.0089, 0.0099, 0.0087, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 12:27:00,489 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8707, 2.9392, 2.6976, 4.4983, 3.7775, 4.2732, 1.7369, 3.0930], device='cuda:5'), covar=tensor([0.1292, 0.0640, 0.1089, 0.0192, 0.0225, 0.0344, 0.1486, 0.0761], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0167, 0.0189, 0.0178, 0.0203, 0.0213, 0.0194, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 12:27:52,940 INFO [train.py:904] (5/8) Epoch 17, batch 3900, loss[loss=0.1739, simple_loss=0.2542, pruned_loss=0.04678, over 15668.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2563, pruned_loss=0.04935, over 3279672.76 frames. ], batch size: 190, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:27:59,341 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6706, 3.8291, 2.2403, 4.1354, 2.8746, 4.1984, 2.3177, 2.8663], device='cuda:5'), covar=tensor([0.0294, 0.0357, 0.1619, 0.0238, 0.0759, 0.0517, 0.1458, 0.0800], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0174, 0.0192, 0.0158, 0.0172, 0.0217, 0.0201, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 12:28:01,005 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:28:37,864 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.433e+02 2.279e+02 2.722e+02 3.230e+02 5.998e+02, threshold=5.443e+02, percent-clipped=1.0 2023-04-30 12:28:49,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2991, 3.4865, 3.5565, 2.1399, 3.1158, 2.5797, 3.8331, 3.8389], device='cuda:5'), covar=tensor([0.0243, 0.0794, 0.0568, 0.1951, 0.0825, 0.0879, 0.0459, 0.0797], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0160, 0.0164, 0.0150, 0.0141, 0.0127, 0.0140, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 12:29:07,068 INFO [train.py:904] (5/8) Epoch 17, batch 3950, loss[loss=0.1628, simple_loss=0.2356, pruned_loss=0.04502, over 16864.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2564, pruned_loss=0.05012, over 3278030.61 frames. ], batch size: 109, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:29:30,699 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:18,533 INFO [train.py:904] (5/8) Epoch 17, batch 4000, loss[loss=0.1976, simple_loss=0.2824, pruned_loss=0.05643, over 12239.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2563, pruned_loss=0.05015, over 3268733.68 frames. ], batch size: 248, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:30:25,249 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:31,179 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166410.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:31:03,721 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.351e+02 2.185e+02 2.579e+02 2.981e+02 5.548e+02, threshold=5.157e+02, percent-clipped=1.0 2023-04-30 12:31:10,135 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9206, 4.0035, 2.8163, 2.5287, 2.6333, 2.4320, 4.1724, 3.6250], device='cuda:5'), covar=tensor([0.2325, 0.0625, 0.1774, 0.2112, 0.2481, 0.1924, 0.0502, 0.0887], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0263, 0.0297, 0.0300, 0.0292, 0.0244, 0.0283, 0.0325], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 12:31:31,417 INFO [train.py:904] (5/8) Epoch 17, batch 4050, loss[loss=0.1814, simple_loss=0.2646, pruned_loss=0.04915, over 17231.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2568, pruned_loss=0.04921, over 3271099.14 frames. ], batch size: 45, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:31:34,118 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166454.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:31:42,527 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:08,501 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3770, 3.3468, 3.2675, 3.5119, 3.5113, 3.3127, 3.5254, 3.6070], device='cuda:5'), covar=tensor([0.1368, 0.1041, 0.1538, 0.0817, 0.0888, 0.2697, 0.1156, 0.0952], device='cuda:5'), in_proj_covar=tensor([0.0625, 0.0767, 0.0913, 0.0780, 0.0587, 0.0626, 0.0622, 0.0728], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:32:14,630 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:44,118 INFO [train.py:904] (5/8) Epoch 17, batch 4100, loss[loss=0.1998, simple_loss=0.2897, pruned_loss=0.055, over 16520.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2582, pruned_loss=0.0487, over 3256258.51 frames. ], batch size: 75, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:32:46,311 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166503.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:51,063 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166507.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:54,678 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:29,312 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:31,672 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.012e+02 2.257e+02 2.846e+02 6.984e+02, threshold=4.513e+02, percent-clipped=1.0 2023-04-30 12:33:45,806 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:59,692 INFO [train.py:904] (5/8) Epoch 17, batch 4150, loss[loss=0.2533, simple_loss=0.3241, pruned_loss=0.09131, over 11505.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2662, pruned_loss=0.05169, over 3224364.24 frames. ], batch size: 250, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:34:23,794 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-30 12:34:28,393 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:34:37,094 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2279, 2.2247, 2.2301, 3.9682, 2.1355, 2.5859, 2.2573, 2.3909], device='cuda:5'), covar=tensor([0.1183, 0.3357, 0.2684, 0.0480, 0.3766, 0.2215, 0.3206, 0.3023], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0432, 0.0357, 0.0328, 0.0430, 0.0501, 0.0400, 0.0506], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:35:01,103 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166592.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:04,685 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6843, 5.0301, 5.1795, 4.9049, 4.8903, 5.5431, 4.9645, 4.7126], device='cuda:5'), covar=tensor([0.1044, 0.1691, 0.1510, 0.1787, 0.2688, 0.0941, 0.1453, 0.2383], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0562, 0.0615, 0.0474, 0.0633, 0.0653, 0.0488, 0.0634], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 12:35:14,016 INFO [train.py:904] (5/8) Epoch 17, batch 4200, loss[loss=0.2151, simple_loss=0.3118, pruned_loss=0.0592, over 16472.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2737, pruned_loss=0.05327, over 3218345.99 frames. ], batch size: 68, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:00,017 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.403e+02 2.795e+02 3.348e+02 5.640e+02, threshold=5.591e+02, percent-clipped=5.0 2023-04-30 12:36:13,751 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 12:36:27,724 INFO [train.py:904] (5/8) Epoch 17, batch 4250, loss[loss=0.1994, simple_loss=0.2959, pruned_loss=0.05148, over 17130.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2774, pruned_loss=0.05383, over 3184844.94 frames. ], batch size: 48, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:43,559 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166663.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:36:49,768 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6248, 4.7234, 4.9944, 4.9859, 4.9888, 4.6998, 4.6874, 4.4163], device='cuda:5'), covar=tensor([0.0264, 0.0419, 0.0385, 0.0427, 0.0378, 0.0324, 0.0844, 0.0465], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0415, 0.0404, 0.0381, 0.0450, 0.0426, 0.0524, 0.0339], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 12:37:33,340 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4306, 2.1194, 1.6856, 1.8383, 2.4510, 2.0883, 2.3367, 2.6273], device='cuda:5'), covar=tensor([0.0190, 0.0459, 0.0598, 0.0515, 0.0236, 0.0399, 0.0220, 0.0240], device='cuda:5'), in_proj_covar=tensor([0.0191, 0.0225, 0.0216, 0.0218, 0.0227, 0.0226, 0.0230, 0.0222], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:37:39,006 INFO [train.py:904] (5/8) Epoch 17, batch 4300, loss[loss=0.1739, simple_loss=0.2717, pruned_loss=0.03803, over 16296.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2777, pruned_loss=0.05225, over 3196584.44 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:37:51,466 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166710.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 12:38:24,696 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.116e+02 2.589e+02 3.113e+02 5.603e+02, threshold=5.178e+02, percent-clipped=2.0 2023-04-30 12:38:28,889 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:38:52,949 INFO [train.py:904] (5/8) Epoch 17, batch 4350, loss[loss=0.1898, simple_loss=0.2834, pruned_loss=0.04808, over 16776.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2805, pruned_loss=0.05317, over 3185812.78 frames. ], batch size: 83, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:39:01,857 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166758.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:39:10,411 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:39:57,632 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:05,344 INFO [train.py:904] (5/8) Epoch 17, batch 4400, loss[loss=0.1888, simple_loss=0.2762, pruned_loss=0.05068, over 16655.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2829, pruned_loss=0.05436, over 3173937.99 frames. ], batch size: 134, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:40:06,874 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:37,936 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166825.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:49,146 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.232e+02 2.626e+02 3.012e+02 5.313e+02, threshold=5.252e+02, percent-clipped=1.0 2023-04-30 12:40:55,207 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:41:15,722 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166851.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:41:16,590 INFO [train.py:904] (5/8) Epoch 17, batch 4450, loss[loss=0.2033, simple_loss=0.29, pruned_loss=0.05824, over 16786.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2865, pruned_loss=0.05582, over 3179906.08 frames. ], batch size: 39, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:41:36,458 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:07,505 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:28,866 INFO [train.py:904] (5/8) Epoch 17, batch 4500, loss[loss=0.2115, simple_loss=0.2802, pruned_loss=0.07143, over 11307.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2862, pruned_loss=0.05597, over 3180245.29 frames. ], batch size: 247, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:07,239 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8449, 4.6710, 4.8841, 5.0289, 5.1917, 4.6420, 5.1920, 5.2148], device='cuda:5'), covar=tensor([0.1601, 0.1041, 0.1355, 0.0631, 0.0466, 0.0853, 0.0460, 0.0464], device='cuda:5'), in_proj_covar=tensor([0.0601, 0.0741, 0.0878, 0.0754, 0.0566, 0.0602, 0.0597, 0.0701], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:43:07,282 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:43:12,782 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 1.901e+02 2.204e+02 2.572e+02 5.344e+02, threshold=4.409e+02, percent-clipped=1.0 2023-04-30 12:43:31,343 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8912, 1.7130, 2.4283, 2.7770, 2.6749, 3.0614, 1.8421, 3.1086], device='cuda:5'), covar=tensor([0.0174, 0.0450, 0.0249, 0.0237, 0.0237, 0.0161, 0.0476, 0.0107], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0176, 0.0186, 0.0143, 0.0187, 0.0137], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:43:40,965 INFO [train.py:904] (5/8) Epoch 17, batch 4550, loss[loss=0.2147, simple_loss=0.2991, pruned_loss=0.06515, over 16471.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2871, pruned_loss=0.0569, over 3189271.42 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:57,324 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:20,140 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2909, 4.2768, 4.0693, 3.3570, 4.1395, 1.7541, 3.9013, 3.5796], device='cuda:5'), covar=tensor([0.0062, 0.0055, 0.0146, 0.0238, 0.0057, 0.2712, 0.0096, 0.0198], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0141, 0.0187, 0.0172, 0.0162, 0.0197, 0.0177, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:44:35,612 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:53,181 INFO [train.py:904] (5/8) Epoch 17, batch 4600, loss[loss=0.2025, simple_loss=0.2838, pruned_loss=0.0606, over 16265.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2881, pruned_loss=0.05724, over 3187178.02 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:45:07,108 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167011.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:45:38,258 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.923e+02 2.253e+02 2.642e+02 5.543e+02, threshold=4.506e+02, percent-clipped=2.0 2023-04-30 12:46:05,382 INFO [train.py:904] (5/8) Epoch 17, batch 4650, loss[loss=0.1897, simple_loss=0.2706, pruned_loss=0.05442, over 16972.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2867, pruned_loss=0.05672, over 3203887.72 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:46:43,203 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8648, 2.7208, 2.5948, 2.0046, 2.5528, 2.7284, 2.6016, 1.9070], device='cuda:5'), covar=tensor([0.0363, 0.0066, 0.0060, 0.0320, 0.0097, 0.0108, 0.0095, 0.0339], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0075, 0.0076, 0.0129, 0.0090, 0.0099, 0.0088, 0.0122], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 12:46:58,522 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 12:47:00,709 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:13,871 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167100.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:16,415 INFO [train.py:904] (5/8) Epoch 17, batch 4700, loss[loss=0.1821, simple_loss=0.2675, pruned_loss=0.04837, over 16669.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2838, pruned_loss=0.05549, over 3185770.38 frames. ], batch size: 62, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:47:18,972 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167104.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:42,090 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167120.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:00,931 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 1.982e+02 2.244e+02 2.685e+02 3.929e+02, threshold=4.489e+02, percent-clipped=0.0 2023-04-30 12:48:06,780 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:28,395 INFO [train.py:904] (5/8) Epoch 17, batch 4750, loss[loss=0.1856, simple_loss=0.2757, pruned_loss=0.04777, over 16641.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2801, pruned_loss=0.05381, over 3185962.63 frames. ], batch size: 76, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:48:42,071 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:44,943 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8191, 2.8264, 2.6772, 4.4639, 3.1514, 4.0606, 1.7083, 2.9626], device='cuda:5'), covar=tensor([0.1267, 0.0701, 0.1181, 0.0138, 0.0238, 0.0368, 0.1503, 0.0823], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0167, 0.0189, 0.0175, 0.0203, 0.0211, 0.0192, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 12:48:48,026 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:48,114 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:16,408 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:18,966 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167187.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:40,744 INFO [train.py:904] (5/8) Epoch 17, batch 4800, loss[loss=0.188, simple_loss=0.2825, pruned_loss=0.04672, over 16713.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2763, pruned_loss=0.05174, over 3181846.14 frames. ], batch size: 124, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:49:58,977 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:27,976 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.008e+02 2.170e+02 2.594e+02 4.937e+02, threshold=4.340e+02, percent-clipped=1.0 2023-04-30 12:50:32,465 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167235.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:58,082 INFO [train.py:904] (5/8) Epoch 17, batch 4850, loss[loss=0.172, simple_loss=0.2684, pruned_loss=0.03783, over 16437.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.277, pruned_loss=0.051, over 3178057.64 frames. ], batch size: 146, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:51:46,662 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:51:57,430 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3686, 2.3322, 2.3001, 4.1591, 2.2641, 2.7129, 2.3719, 2.5561], device='cuda:5'), covar=tensor([0.1163, 0.3328, 0.2609, 0.0423, 0.3637, 0.2242, 0.3239, 0.2904], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0428, 0.0352, 0.0323, 0.0427, 0.0493, 0.0396, 0.0499], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 12:52:12,196 INFO [train.py:904] (5/8) Epoch 17, batch 4900, loss[loss=0.1696, simple_loss=0.2642, pruned_loss=0.03755, over 16917.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2767, pruned_loss=0.05015, over 3175364.64 frames. ], batch size: 109, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:52:19,516 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3371, 3.9906, 4.0374, 2.6898, 3.5056, 3.9385, 3.6379, 2.1878], device='cuda:5'), covar=tensor([0.0472, 0.0035, 0.0031, 0.0349, 0.0087, 0.0094, 0.0074, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0129, 0.0075, 0.0075, 0.0127, 0.0089, 0.0098, 0.0087, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 12:52:55,968 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.440e+02 1.938e+02 2.305e+02 2.710e+02 4.454e+02, threshold=4.611e+02, percent-clipped=1.0 2023-04-30 12:53:08,397 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6352, 2.6231, 2.3607, 4.2191, 2.9202, 3.9905, 1.4441, 3.0223], device='cuda:5'), covar=tensor([0.1359, 0.0795, 0.1325, 0.0132, 0.0214, 0.0358, 0.1650, 0.0748], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0168, 0.0189, 0.0175, 0.0203, 0.0211, 0.0193, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 12:53:24,992 INFO [train.py:904] (5/8) Epoch 17, batch 4950, loss[loss=0.2012, simple_loss=0.2986, pruned_loss=0.0519, over 16746.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2764, pruned_loss=0.0495, over 3196847.48 frames. ], batch size: 83, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:54:22,606 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167391.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:54:37,760 INFO [train.py:904] (5/8) Epoch 17, batch 5000, loss[loss=0.1598, simple_loss=0.2576, pruned_loss=0.03102, over 16832.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2779, pruned_loss=0.04922, over 3212204.57 frames. ], batch size: 102, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:02,915 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:19,915 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:20,678 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.024e+02 2.500e+02 2.980e+02 7.048e+02, threshold=5.000e+02, percent-clipped=1.0 2023-04-30 12:55:29,658 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:46,693 INFO [train.py:904] (5/8) Epoch 17, batch 5050, loss[loss=0.1646, simple_loss=0.255, pruned_loss=0.03705, over 17227.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2774, pruned_loss=0.04881, over 3230110.23 frames. ], batch size: 52, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:51,613 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167456.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:57,203 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:08,481 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:42,623 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 12:56:43,513 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:56,786 INFO [train.py:904] (5/8) Epoch 17, batch 5100, loss[loss=0.1633, simple_loss=0.2644, pruned_loss=0.03108, over 16795.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2757, pruned_loss=0.04807, over 3219750.47 frames. ], batch size: 89, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:57:39,790 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 2.010e+02 2.231e+02 2.542e+02 5.876e+02, threshold=4.463e+02, percent-clipped=1.0 2023-04-30 12:58:08,925 INFO [train.py:904] (5/8) Epoch 17, batch 5150, loss[loss=0.1702, simple_loss=0.2731, pruned_loss=0.03359, over 16716.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.275, pruned_loss=0.04706, over 3225764.02 frames. ], batch size: 89, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:58:55,899 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 12:58:56,770 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:59:21,214 INFO [train.py:904] (5/8) Epoch 17, batch 5200, loss[loss=0.1601, simple_loss=0.2499, pruned_loss=0.03512, over 16491.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2738, pruned_loss=0.04646, over 3219662.13 frames. ], batch size: 75, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:00:07,264 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.038e+02 2.260e+02 2.754e+02 5.460e+02, threshold=4.520e+02, percent-clipped=3.0 2023-04-30 13:00:08,172 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167633.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:00:36,324 INFO [train.py:904] (5/8) Epoch 17, batch 5250, loss[loss=0.1892, simple_loss=0.2808, pruned_loss=0.04879, over 16685.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2711, pruned_loss=0.04601, over 3226948.86 frames. ], batch size: 134, lr: 3.99e-03, grad_scale: 16.0 2023-04-30 13:01:02,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2179, 3.6783, 3.7873, 2.1388, 3.1828, 2.4355, 3.8136, 3.9238], device='cuda:5'), covar=tensor([0.0220, 0.0674, 0.0511, 0.1924, 0.0770, 0.0919, 0.0491, 0.0756], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0159, 0.0165, 0.0151, 0.0143, 0.0127, 0.0142, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 13:01:48,350 INFO [train.py:904] (5/8) Epoch 17, batch 5300, loss[loss=0.1712, simple_loss=0.2642, pruned_loss=0.0391, over 16571.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2681, pruned_loss=0.04542, over 3223602.03 frames. ], batch size: 75, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:02:12,798 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:02:35,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 2.142e+02 2.405e+02 3.044e+02 5.207e+02, threshold=4.809e+02, percent-clipped=2.0 2023-04-30 13:03:01,108 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5774, 3.8425, 3.9975, 3.9569, 3.9161, 3.7742, 3.4341, 3.7830], device='cuda:5'), covar=tensor([0.0575, 0.0828, 0.0599, 0.0625, 0.0790, 0.0649, 0.1503, 0.0603], device='cuda:5'), in_proj_covar=tensor([0.0376, 0.0411, 0.0402, 0.0377, 0.0448, 0.0421, 0.0521, 0.0336], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 13:03:01,790 INFO [train.py:904] (5/8) Epoch 17, batch 5350, loss[loss=0.2072, simple_loss=0.3003, pruned_loss=0.05705, over 15369.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2661, pruned_loss=0.04458, over 3214833.38 frames. ], batch size: 191, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:03:08,182 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:13,838 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167760.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:23,297 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:41,671 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:53,664 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:04,001 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5393, 3.6132, 3.3603, 3.0052, 3.1915, 3.4883, 3.2694, 3.3298], device='cuda:5'), covar=tensor([0.0563, 0.0533, 0.0266, 0.0251, 0.0516, 0.0429, 0.1328, 0.0468], device='cuda:5'), in_proj_covar=tensor([0.0274, 0.0386, 0.0327, 0.0315, 0.0336, 0.0367, 0.0223, 0.0389], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:04:14,437 INFO [train.py:904] (5/8) Epoch 17, batch 5400, loss[loss=0.1791, simple_loss=0.2718, pruned_loss=0.04319, over 16486.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2687, pruned_loss=0.04538, over 3198097.78 frames. ], batch size: 68, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:04:17,675 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:24,145 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:27,402 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1398, 4.1818, 4.4871, 4.4596, 4.4497, 4.1880, 4.1862, 4.1324], device='cuda:5'), covar=tensor([0.0299, 0.0602, 0.0387, 0.0393, 0.0456, 0.0361, 0.0868, 0.0466], device='cuda:5'), in_proj_covar=tensor([0.0376, 0.0411, 0.0402, 0.0377, 0.0448, 0.0421, 0.0521, 0.0336], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 13:04:43,670 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0564, 4.3650, 3.1898, 2.7096, 3.1159, 2.8386, 4.8133, 3.9203], device='cuda:5'), covar=tensor([0.2469, 0.0608, 0.1677, 0.2342, 0.2416, 0.1650, 0.0394, 0.1023], device='cuda:5'), in_proj_covar=tensor([0.0315, 0.0260, 0.0293, 0.0297, 0.0286, 0.0239, 0.0281, 0.0317], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 13:04:54,017 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167828.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:05:02,681 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.037e+02 2.360e+02 2.749e+02 5.264e+02, threshold=4.720e+02, percent-clipped=1.0 2023-04-30 13:05:31,685 INFO [train.py:904] (5/8) Epoch 17, batch 5450, loss[loss=0.2476, simple_loss=0.3266, pruned_loss=0.08426, over 15369.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2727, pruned_loss=0.04737, over 3199939.80 frames. ], batch size: 191, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:06:48,606 INFO [train.py:904] (5/8) Epoch 17, batch 5500, loss[loss=0.2852, simple_loss=0.3473, pruned_loss=0.1116, over 11834.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2808, pruned_loss=0.05206, over 3171801.94 frames. ], batch size: 246, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:07:39,671 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.878e+02 2.929e+02 3.465e+02 4.184e+02 8.083e+02, threshold=6.929e+02, percent-clipped=17.0 2023-04-30 13:08:05,267 INFO [train.py:904] (5/8) Epoch 17, batch 5550, loss[loss=0.27, simple_loss=0.3316, pruned_loss=0.1043, over 11003.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2876, pruned_loss=0.05676, over 3140733.22 frames. ], batch size: 247, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:08:30,093 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 13:09:31,458 INFO [train.py:904] (5/8) Epoch 17, batch 5600, loss[loss=0.1936, simple_loss=0.274, pruned_loss=0.05663, over 16651.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2934, pruned_loss=0.06175, over 3094261.29 frames. ], batch size: 57, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:10:27,347 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 3.389e+02 4.058e+02 4.967e+02 1.149e+03, threshold=8.116e+02, percent-clipped=3.0 2023-04-30 13:10:27,975 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168035.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:10:54,416 INFO [train.py:904] (5/8) Epoch 17, batch 5650, loss[loss=0.278, simple_loss=0.3384, pruned_loss=0.1088, over 11238.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2976, pruned_loss=0.06513, over 3074680.56 frames. ], batch size: 250, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:11:30,933 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:11:50,872 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:12:04,689 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168096.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:12:12,313 INFO [train.py:904] (5/8) Epoch 17, batch 5700, loss[loss=0.2395, simple_loss=0.327, pruned_loss=0.076, over 16847.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3, pruned_loss=0.06766, over 3048922.46 frames. ], batch size: 116, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:12:46,687 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:12:53,755 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5767, 3.5261, 2.4465, 2.2449, 2.4280, 2.0953, 3.7070, 3.2897], device='cuda:5'), covar=tensor([0.3060, 0.0921, 0.2225, 0.2584, 0.2704, 0.2408, 0.0584, 0.1142], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0260, 0.0296, 0.0299, 0.0289, 0.0242, 0.0284, 0.0319], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 13:13:04,293 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.185e+02 3.066e+02 3.948e+02 4.929e+02 1.585e+03, threshold=7.895e+02, percent-clipped=5.0 2023-04-30 13:13:06,079 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:13:06,176 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4034, 5.7160, 5.4144, 5.4952, 5.0664, 5.0048, 5.2307, 5.8368], device='cuda:5'), covar=tensor([0.1171, 0.0830, 0.1045, 0.0855, 0.0847, 0.0775, 0.1078, 0.0872], device='cuda:5'), in_proj_covar=tensor([0.0626, 0.0770, 0.0625, 0.0566, 0.0483, 0.0494, 0.0637, 0.0595], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:13:31,693 INFO [train.py:904] (5/8) Epoch 17, batch 5750, loss[loss=0.2656, simple_loss=0.3223, pruned_loss=0.1045, over 11273.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3028, pruned_loss=0.06936, over 3019915.90 frames. ], batch size: 246, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:14:17,445 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2104, 2.4783, 2.1160, 2.3164, 2.8326, 2.5538, 2.8810, 3.0461], device='cuda:5'), covar=tensor([0.0140, 0.0384, 0.0490, 0.0417, 0.0246, 0.0345, 0.0235, 0.0233], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0222, 0.0215, 0.0216, 0.0225, 0.0223, 0.0224, 0.0219], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:14:50,312 INFO [train.py:904] (5/8) Epoch 17, batch 5800, loss[loss=0.1889, simple_loss=0.2812, pruned_loss=0.04835, over 16374.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3018, pruned_loss=0.06707, over 3050275.09 frames. ], batch size: 146, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:15:41,127 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168233.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:15:43,220 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.888e+02 3.447e+02 4.230e+02 9.306e+02, threshold=6.893e+02, percent-clipped=1.0 2023-04-30 13:16:07,866 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5741, 2.4821, 2.4485, 4.2604, 2.7771, 4.0393, 1.4688, 2.9901], device='cuda:5'), covar=tensor([0.1544, 0.0930, 0.1317, 0.0186, 0.0311, 0.0394, 0.1856, 0.0807], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0167, 0.0188, 0.0174, 0.0200, 0.0209, 0.0191, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 13:16:09,629 INFO [train.py:904] (5/8) Epoch 17, batch 5850, loss[loss=0.2072, simple_loss=0.2953, pruned_loss=0.05953, over 16472.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2997, pruned_loss=0.06562, over 3054088.36 frames. ], batch size: 75, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:16:46,506 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 13:17:19,307 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168294.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:17:32,956 INFO [train.py:904] (5/8) Epoch 17, batch 5900, loss[loss=0.201, simple_loss=0.2965, pruned_loss=0.05279, over 16745.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2997, pruned_loss=0.06567, over 3066356.04 frames. ], batch size: 76, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:17:53,347 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8662, 2.6765, 2.6427, 1.9791, 2.5890, 2.6861, 2.5369, 1.9158], device='cuda:5'), covar=tensor([0.0418, 0.0075, 0.0071, 0.0341, 0.0111, 0.0111, 0.0108, 0.0361], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0077, 0.0078, 0.0132, 0.0091, 0.0102, 0.0090, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 13:18:28,194 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.786e+02 3.302e+02 4.334e+02 9.819e+02, threshold=6.604e+02, percent-clipped=5.0 2023-04-30 13:18:52,634 INFO [train.py:904] (5/8) Epoch 17, batch 5950, loss[loss=0.2155, simple_loss=0.3033, pruned_loss=0.06386, over 16281.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2997, pruned_loss=0.06459, over 3051990.70 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:19:29,715 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168375.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:19:52,918 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168391.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 13:19:59,839 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8352, 3.6619, 4.2647, 1.8783, 4.4806, 4.4468, 3.1352, 3.2563], device='cuda:5'), covar=tensor([0.0729, 0.0255, 0.0180, 0.1246, 0.0053, 0.0131, 0.0397, 0.0434], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0104, 0.0092, 0.0135, 0.0073, 0.0117, 0.0123, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 13:20:12,029 INFO [train.py:904] (5/8) Epoch 17, batch 6000, loss[loss=0.2129, simple_loss=0.2893, pruned_loss=0.06828, over 16572.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2991, pruned_loss=0.06432, over 3066099.46 frames. ], batch size: 57, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:20:12,029 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 13:20:21,982 INFO [train.py:938] (5/8) Epoch 17, validation: loss=0.1535, simple_loss=0.2667, pruned_loss=0.0202, over 944034.00 frames. 2023-04-30 13:20:21,983 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 13:20:43,926 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 13:20:53,933 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:20:53,975 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:21:15,441 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 2.768e+02 3.388e+02 4.185e+02 7.743e+02, threshold=6.777e+02, percent-clipped=1.0 2023-04-30 13:21:24,685 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.57 vs. limit=5.0 2023-04-30 13:21:39,931 INFO [train.py:904] (5/8) Epoch 17, batch 6050, loss[loss=0.2518, simple_loss=0.3089, pruned_loss=0.09731, over 11174.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2969, pruned_loss=0.06301, over 3079749.22 frames. ], batch size: 246, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:22:09,547 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:22:59,844 INFO [train.py:904] (5/8) Epoch 17, batch 6100, loss[loss=0.2251, simple_loss=0.2954, pruned_loss=0.07738, over 11603.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2961, pruned_loss=0.0619, over 3084049.78 frames. ], batch size: 248, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:23:55,310 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.750e+02 3.352e+02 4.017e+02 8.288e+02, threshold=6.704e+02, percent-clipped=2.0 2023-04-30 13:24:18,369 INFO [train.py:904] (5/8) Epoch 17, batch 6150, loss[loss=0.173, simple_loss=0.2617, pruned_loss=0.04212, over 17245.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2935, pruned_loss=0.06058, over 3096978.28 frames. ], batch size: 45, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:24:53,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3286, 3.2081, 3.1649, 3.4431, 3.4400, 3.2932, 3.3918, 3.4867], device='cuda:5'), covar=tensor([0.1337, 0.1220, 0.1692, 0.0852, 0.0976, 0.2640, 0.1493, 0.1146], device='cuda:5'), in_proj_covar=tensor([0.0585, 0.0721, 0.0860, 0.0738, 0.0554, 0.0585, 0.0591, 0.0688], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:25:17,268 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:25:36,307 INFO [train.py:904] (5/8) Epoch 17, batch 6200, loss[loss=0.218, simple_loss=0.2953, pruned_loss=0.07034, over 11731.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2923, pruned_loss=0.06084, over 3082388.70 frames. ], batch size: 248, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:26:15,026 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.57 vs. limit=5.0 2023-04-30 13:26:31,029 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.813e+02 2.748e+02 3.467e+02 4.001e+02 6.501e+02, threshold=6.934e+02, percent-clipped=0.0 2023-04-30 13:26:53,722 INFO [train.py:904] (5/8) Epoch 17, batch 6250, loss[loss=0.2036, simple_loss=0.2937, pruned_loss=0.05677, over 15215.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2925, pruned_loss=0.06115, over 3073520.09 frames. ], batch size: 190, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:27:53,846 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168691.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:28:10,637 INFO [train.py:904] (5/8) Epoch 17, batch 6300, loss[loss=0.169, simple_loss=0.2656, pruned_loss=0.03618, over 16882.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2918, pruned_loss=0.06046, over 3079731.17 frames. ], batch size: 96, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:28:25,218 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8831, 4.8500, 4.7085, 4.0314, 4.7843, 1.8114, 4.5423, 4.5220], device='cuda:5'), covar=tensor([0.0083, 0.0075, 0.0164, 0.0339, 0.0084, 0.2612, 0.0122, 0.0176], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0139, 0.0186, 0.0171, 0.0159, 0.0196, 0.0173, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:29:06,434 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.898e+02 2.739e+02 3.263e+02 4.043e+02 7.710e+02, threshold=6.525e+02, percent-clipped=1.0 2023-04-30 13:29:09,982 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168739.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:29:29,590 INFO [train.py:904] (5/8) Epoch 17, batch 6350, loss[loss=0.2131, simple_loss=0.2981, pruned_loss=0.06404, over 16770.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2929, pruned_loss=0.06186, over 3065884.90 frames. ], batch size: 124, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:29:31,996 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9095, 2.7262, 2.6622, 1.9750, 2.5306, 2.6549, 2.6140, 1.9519], device='cuda:5'), covar=tensor([0.0392, 0.0077, 0.0074, 0.0344, 0.0118, 0.0127, 0.0105, 0.0361], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0076, 0.0078, 0.0132, 0.0090, 0.0102, 0.0089, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 13:29:42,679 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0742, 5.0413, 4.9167, 4.5519, 4.5242, 4.9561, 4.8298, 4.6773], device='cuda:5'), covar=tensor([0.0589, 0.0546, 0.0305, 0.0303, 0.1063, 0.0458, 0.0308, 0.0695], device='cuda:5'), in_proj_covar=tensor([0.0275, 0.0385, 0.0324, 0.0312, 0.0332, 0.0364, 0.0221, 0.0388], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:30:46,706 INFO [train.py:904] (5/8) Epoch 17, batch 6400, loss[loss=0.1935, simple_loss=0.2768, pruned_loss=0.05511, over 16713.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2928, pruned_loss=0.06238, over 3075690.92 frames. ], batch size: 89, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:31:42,104 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.043e+02 2.957e+02 3.621e+02 4.188e+02 7.946e+02, threshold=7.242e+02, percent-clipped=4.0 2023-04-30 13:32:00,932 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6180, 3.9223, 2.9536, 2.2363, 2.5249, 2.4555, 4.1318, 3.4515], device='cuda:5'), covar=tensor([0.2843, 0.0570, 0.1675, 0.2611, 0.2642, 0.1960, 0.0398, 0.1169], device='cuda:5'), in_proj_covar=tensor([0.0316, 0.0258, 0.0295, 0.0298, 0.0287, 0.0241, 0.0282, 0.0318], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 13:32:03,562 INFO [train.py:904] (5/8) Epoch 17, batch 6450, loss[loss=0.188, simple_loss=0.2796, pruned_loss=0.04819, over 16643.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2929, pruned_loss=0.06166, over 3085030.03 frames. ], batch size: 134, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:33:02,323 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:33:02,378 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5997, 3.5392, 3.9748, 1.7720, 4.1202, 4.1589, 3.0717, 3.0472], device='cuda:5'), covar=tensor([0.0776, 0.0235, 0.0168, 0.1285, 0.0064, 0.0133, 0.0375, 0.0439], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0104, 0.0092, 0.0136, 0.0074, 0.0117, 0.0124, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 13:33:20,860 INFO [train.py:904] (5/8) Epoch 17, batch 6500, loss[loss=0.1847, simple_loss=0.2707, pruned_loss=0.04934, over 16671.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2911, pruned_loss=0.06113, over 3093362.21 frames. ], batch size: 62, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:34:15,423 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:34:16,253 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 2.744e+02 3.257e+02 4.082e+02 7.063e+02, threshold=6.513e+02, percent-clipped=0.0 2023-04-30 13:34:40,491 INFO [train.py:904] (5/8) Epoch 17, batch 6550, loss[loss=0.2383, simple_loss=0.3057, pruned_loss=0.08547, over 11539.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2947, pruned_loss=0.06275, over 3076066.21 frames. ], batch size: 246, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:34:44,924 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3949, 3.2665, 2.7127, 2.1029, 2.2338, 2.2441, 3.3506, 3.0293], device='cuda:5'), covar=tensor([0.2737, 0.0650, 0.1578, 0.2827, 0.2523, 0.2106, 0.0498, 0.1276], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0259, 0.0295, 0.0299, 0.0289, 0.0242, 0.0283, 0.0319], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 13:35:03,835 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168967.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:35:06,344 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168969.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:35:33,576 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8343, 3.0072, 2.6534, 4.8184, 3.6118, 4.1683, 1.6243, 2.9824], device='cuda:5'), covar=tensor([0.1307, 0.0722, 0.1237, 0.0154, 0.0375, 0.0414, 0.1634, 0.0864], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0167, 0.0189, 0.0175, 0.0202, 0.0210, 0.0193, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 13:35:58,163 INFO [train.py:904] (5/8) Epoch 17, batch 6600, loss[loss=0.2334, simple_loss=0.3099, pruned_loss=0.07846, over 15272.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2971, pruned_loss=0.06348, over 3072942.06 frames. ], batch size: 190, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:36:09,706 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0376, 4.8729, 5.1101, 5.2842, 5.4731, 4.8830, 5.4876, 5.4918], device='cuda:5'), covar=tensor([0.1705, 0.1345, 0.1593, 0.0723, 0.0542, 0.0770, 0.0544, 0.0595], device='cuda:5'), in_proj_covar=tensor([0.0586, 0.0722, 0.0859, 0.0740, 0.0555, 0.0588, 0.0591, 0.0690], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:36:37,556 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:36:40,853 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169030.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 13:36:50,874 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.872e+02 3.571e+02 4.497e+02 9.620e+02, threshold=7.142e+02, percent-clipped=5.0 2023-04-30 13:37:13,190 INFO [train.py:904] (5/8) Epoch 17, batch 6650, loss[loss=0.2278, simple_loss=0.3039, pruned_loss=0.07586, over 15363.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2974, pruned_loss=0.06439, over 3075103.54 frames. ], batch size: 190, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:37:47,797 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3132, 4.3267, 4.7134, 4.6741, 4.6828, 4.3824, 4.3947, 4.2723], device='cuda:5'), covar=tensor([0.0341, 0.0570, 0.0363, 0.0419, 0.0503, 0.0371, 0.0878, 0.0521], device='cuda:5'), in_proj_covar=tensor([0.0381, 0.0416, 0.0405, 0.0382, 0.0452, 0.0429, 0.0526, 0.0340], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 13:37:59,908 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:38:30,529 INFO [train.py:904] (5/8) Epoch 17, batch 6700, loss[loss=0.2066, simple_loss=0.2902, pruned_loss=0.06152, over 16618.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2957, pruned_loss=0.06389, over 3083370.22 frames. ], batch size: 134, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:38:53,643 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:39:26,520 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.893e+02 2.876e+02 3.506e+02 4.097e+02 7.829e+02, threshold=7.011e+02, percent-clipped=1.0 2023-04-30 13:39:35,028 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169143.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:39:47,626 INFO [train.py:904] (5/8) Epoch 17, batch 6750, loss[loss=0.1932, simple_loss=0.2819, pruned_loss=0.05229, over 16371.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2937, pruned_loss=0.06284, over 3093686.10 frames. ], batch size: 146, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:39:54,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6389, 4.7081, 4.5032, 4.1943, 4.2014, 4.5878, 4.3974, 4.2935], device='cuda:5'), covar=tensor([0.0561, 0.0357, 0.0280, 0.0290, 0.0865, 0.0412, 0.0472, 0.0656], device='cuda:5'), in_proj_covar=tensor([0.0272, 0.0383, 0.0322, 0.0309, 0.0331, 0.0361, 0.0219, 0.0385], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:39:58,674 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:40:26,268 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:40:37,288 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2919, 2.4852, 1.9503, 2.2586, 2.9012, 2.5131, 2.9607, 3.0981], device='cuda:5'), covar=tensor([0.0126, 0.0423, 0.0539, 0.0454, 0.0257, 0.0369, 0.0234, 0.0239], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0223, 0.0216, 0.0216, 0.0224, 0.0222, 0.0225, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:41:04,394 INFO [train.py:904] (5/8) Epoch 17, batch 6800, loss[loss=0.2135, simple_loss=0.2985, pruned_loss=0.06429, over 17100.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2939, pruned_loss=0.06329, over 3089667.26 frames. ], batch size: 47, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:41:33,631 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:42:02,689 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.174e+02 2.920e+02 3.483e+02 4.404e+02 7.991e+02, threshold=6.967e+02, percent-clipped=1.0 2023-04-30 13:42:23,106 INFO [train.py:904] (5/8) Epoch 17, batch 6850, loss[loss=0.2623, simple_loss=0.3217, pruned_loss=0.1015, over 11537.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2954, pruned_loss=0.06408, over 3072729.92 frames. ], batch size: 248, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:43:37,895 INFO [train.py:904] (5/8) Epoch 17, batch 6900, loss[loss=0.2189, simple_loss=0.3078, pruned_loss=0.06503, over 16887.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2971, pruned_loss=0.06326, over 3073031.43 frames. ], batch size: 116, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:44:11,067 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:44:11,709 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 13:44:14,333 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169325.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 13:44:33,271 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.634e+02 3.268e+02 4.133e+02 7.362e+02, threshold=6.535e+02, percent-clipped=1.0 2023-04-30 13:44:55,624 INFO [train.py:904] (5/8) Epoch 17, batch 6950, loss[loss=0.2126, simple_loss=0.2933, pruned_loss=0.06595, over 16852.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2991, pruned_loss=0.06503, over 3066736.12 frames. ], batch size: 116, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:46:11,849 INFO [train.py:904] (5/8) Epoch 17, batch 7000, loss[loss=0.193, simple_loss=0.2881, pruned_loss=0.04893, over 16763.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2989, pruned_loss=0.06358, over 3092368.47 frames. ], batch size: 124, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:47:01,850 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8547, 1.7574, 2.4738, 2.8030, 2.7344, 3.1839, 1.9742, 3.1136], device='cuda:5'), covar=tensor([0.0182, 0.0504, 0.0286, 0.0265, 0.0258, 0.0148, 0.0510, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0184, 0.0170, 0.0173, 0.0184, 0.0141, 0.0186, 0.0136], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:47:07,100 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.910e+02 3.013e+02 3.558e+02 4.321e+02 1.050e+03, threshold=7.116e+02, percent-clipped=3.0 2023-04-30 13:47:07,447 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169438.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:47:28,670 INFO [train.py:904] (5/8) Epoch 17, batch 7050, loss[loss=0.1949, simple_loss=0.2938, pruned_loss=0.04796, over 16985.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2994, pruned_loss=0.06366, over 3087403.43 frames. ], batch size: 55, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:48:00,528 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:48:47,195 INFO [train.py:904] (5/8) Epoch 17, batch 7100, loss[loss=0.2912, simple_loss=0.3413, pruned_loss=0.1206, over 11540.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2982, pruned_loss=0.06353, over 3071071.80 frames. ], batch size: 246, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:49:08,089 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:49:42,143 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 2.876e+02 3.484e+02 4.198e+02 1.044e+03, threshold=6.968e+02, percent-clipped=3.0 2023-04-30 13:50:02,979 INFO [train.py:904] (5/8) Epoch 17, batch 7150, loss[loss=0.2789, simple_loss=0.3394, pruned_loss=0.1092, over 11714.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2954, pruned_loss=0.06308, over 3066183.61 frames. ], batch size: 248, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:50:35,761 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4493, 3.4065, 3.4077, 2.6962, 3.3087, 2.1450, 3.0752, 2.6760], device='cuda:5'), covar=tensor([0.0154, 0.0121, 0.0172, 0.0221, 0.0102, 0.2086, 0.0138, 0.0202], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0136, 0.0183, 0.0167, 0.0156, 0.0193, 0.0169, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:51:00,538 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 13:51:10,859 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3597, 3.3004, 3.5612, 1.6949, 3.7704, 3.7964, 2.8836, 2.7642], device='cuda:5'), covar=tensor([0.0886, 0.0234, 0.0201, 0.1383, 0.0066, 0.0148, 0.0444, 0.0517], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0105, 0.0094, 0.0139, 0.0075, 0.0119, 0.0126, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 13:51:19,566 INFO [train.py:904] (5/8) Epoch 17, batch 7200, loss[loss=0.1818, simple_loss=0.2716, pruned_loss=0.046, over 16246.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2931, pruned_loss=0.06138, over 3054158.61 frames. ], batch size: 165, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:51:48,669 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:51,927 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169623.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:55,296 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169625.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:52:16,457 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.818e+02 3.396e+02 4.262e+02 6.944e+02, threshold=6.791e+02, percent-clipped=0.0 2023-04-30 13:52:38,972 INFO [train.py:904] (5/8) Epoch 17, batch 7250, loss[loss=0.1786, simple_loss=0.2631, pruned_loss=0.0471, over 16624.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2911, pruned_loss=0.06032, over 3068775.42 frames. ], batch size: 134, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:53:08,594 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169671.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:53:12,520 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169673.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:53:22,684 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-30 13:53:25,541 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:53:34,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7864, 3.7554, 3.9539, 3.7444, 3.8850, 4.2701, 3.9402, 3.6811], device='cuda:5'), covar=tensor([0.2368, 0.2247, 0.2252, 0.2662, 0.2733, 0.2098, 0.1587, 0.2647], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0557, 0.0610, 0.0467, 0.0627, 0.0644, 0.0484, 0.0627], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 13:53:57,120 INFO [train.py:904] (5/8) Epoch 17, batch 7300, loss[loss=0.2062, simple_loss=0.2918, pruned_loss=0.06031, over 16779.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2914, pruned_loss=0.06045, over 3073113.31 frames. ], batch size: 134, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:54:23,978 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-30 13:54:29,170 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6402, 1.7096, 2.2651, 2.6183, 2.5306, 3.0762, 1.8381, 2.9726], device='cuda:5'), covar=tensor([0.0210, 0.0475, 0.0282, 0.0279, 0.0283, 0.0133, 0.0483, 0.0124], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0184, 0.0170, 0.0173, 0.0183, 0.0141, 0.0186, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:54:41,449 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:54:55,066 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 2.965e+02 3.498e+02 4.419e+02 8.107e+02, threshold=6.996e+02, percent-clipped=3.0 2023-04-30 13:54:55,518 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:10,687 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:11,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4489, 4.6684, 4.7989, 4.5987, 4.6380, 5.1626, 4.6312, 4.4570], device='cuda:5'), covar=tensor([0.1415, 0.1804, 0.2074, 0.1917, 0.2390, 0.1019, 0.1770, 0.2507], device='cuda:5'), in_proj_covar=tensor([0.0385, 0.0555, 0.0609, 0.0465, 0.0626, 0.0641, 0.0483, 0.0626], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 13:55:17,636 INFO [train.py:904] (5/8) Epoch 17, batch 7350, loss[loss=0.1754, simple_loss=0.2688, pruned_loss=0.04102, over 16420.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2926, pruned_loss=0.06189, over 3036760.07 frames. ], batch size: 68, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:55:22,508 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-30 13:55:49,889 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:12,136 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:12,259 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:12,764 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-04-30 13:56:19,806 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:37,958 INFO [train.py:904] (5/8) Epoch 17, batch 7400, loss[loss=0.2242, simple_loss=0.2964, pruned_loss=0.07597, over 11094.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2933, pruned_loss=0.06197, over 3056550.82 frames. ], batch size: 246, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:56:48,596 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:59,887 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:07,987 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169820.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:36,686 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.712e+02 2.775e+02 3.366e+02 3.998e+02 8.087e+02, threshold=6.732e+02, percent-clipped=3.0 2023-04-30 13:57:38,214 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 13:57:52,304 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169847.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:59,471 INFO [train.py:904] (5/8) Epoch 17, batch 7450, loss[loss=0.2183, simple_loss=0.311, pruned_loss=0.06277, over 15414.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2948, pruned_loss=0.06284, over 3067451.20 frames. ], batch size: 191, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 13:58:10,706 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8757, 2.0223, 2.3664, 3.1687, 2.1094, 2.2496, 2.2491, 2.1413], device='cuda:5'), covar=tensor([0.1270, 0.3181, 0.2164, 0.0631, 0.4055, 0.2442, 0.2957, 0.3241], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0421, 0.0349, 0.0317, 0.0425, 0.0487, 0.0392, 0.0490], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 13:58:19,039 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:59:20,321 INFO [train.py:904] (5/8) Epoch 17, batch 7500, loss[loss=0.1851, simple_loss=0.278, pruned_loss=0.04612, over 16693.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2943, pruned_loss=0.06226, over 3075950.27 frames. ], batch size: 124, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:00:17,628 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.828e+02 2.669e+02 3.210e+02 3.850e+02 6.683e+02, threshold=6.420e+02, percent-clipped=0.0 2023-04-30 14:00:39,277 INFO [train.py:904] (5/8) Epoch 17, batch 7550, loss[loss=0.2018, simple_loss=0.2876, pruned_loss=0.05799, over 16998.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2932, pruned_loss=0.06183, over 3093090.19 frames. ], batch size: 53, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:01:19,596 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:02:00,846 INFO [train.py:904] (5/8) Epoch 17, batch 7600, loss[loss=0.2129, simple_loss=0.2982, pruned_loss=0.0638, over 16437.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2927, pruned_loss=0.06227, over 3088118.70 frames. ], batch size: 146, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:02:08,692 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0210, 4.1024, 3.9186, 3.6209, 3.6482, 4.0207, 3.7047, 3.7865], device='cuda:5'), covar=tensor([0.0578, 0.0526, 0.0277, 0.0272, 0.0652, 0.0433, 0.0985, 0.0575], device='cuda:5'), in_proj_covar=tensor([0.0266, 0.0378, 0.0316, 0.0302, 0.0324, 0.0352, 0.0215, 0.0375], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:02:25,565 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-30 14:02:45,606 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 14:02:46,665 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 14:02:58,191 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.819e+02 3.511e+02 4.461e+02 7.625e+02, threshold=7.022e+02, percent-clipped=5.0 2023-04-30 14:03:04,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-04-30 14:03:08,008 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170045.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:03:17,370 INFO [train.py:904] (5/8) Epoch 17, batch 7650, loss[loss=0.2546, simple_loss=0.3151, pruned_loss=0.09707, over 11772.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2932, pruned_loss=0.06305, over 3070325.43 frames. ], batch size: 247, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:03:30,595 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 14:04:07,614 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:33,141 INFO [train.py:904] (5/8) Epoch 17, batch 7700, loss[loss=0.2057, simple_loss=0.2915, pruned_loss=0.05995, over 16763.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.293, pruned_loss=0.06279, over 3088788.79 frames. ], batch size: 124, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:35,302 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:39,636 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:41,007 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6695, 2.5708, 2.2654, 4.1605, 3.1202, 4.0067, 1.4817, 2.8434], device='cuda:5'), covar=tensor([0.1464, 0.0890, 0.1457, 0.0187, 0.0324, 0.0408, 0.1790, 0.0916], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0168, 0.0189, 0.0174, 0.0203, 0.0211, 0.0194, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 14:05:29,559 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.959e+02 3.624e+02 4.617e+02 9.999e+02, threshold=7.248e+02, percent-clipped=4.0 2023-04-30 14:05:34,969 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:05:50,030 INFO [train.py:904] (5/8) Epoch 17, batch 7750, loss[loss=0.2218, simple_loss=0.3065, pruned_loss=0.06856, over 16748.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2931, pruned_loss=0.0626, over 3083532.31 frames. ], batch size: 124, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:06:27,803 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9932, 3.2172, 3.2047, 2.1697, 2.9660, 3.1723, 3.0527, 1.9398], device='cuda:5'), covar=tensor([0.0513, 0.0056, 0.0063, 0.0375, 0.0105, 0.0123, 0.0095, 0.0430], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0076, 0.0077, 0.0132, 0.0091, 0.0102, 0.0089, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 14:06:36,796 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2818, 4.3554, 4.1929, 3.9146, 3.8214, 4.2883, 4.0128, 3.9815], device='cuda:5'), covar=tensor([0.0669, 0.0643, 0.0317, 0.0309, 0.0911, 0.0541, 0.0694, 0.0664], device='cuda:5'), in_proj_covar=tensor([0.0265, 0.0376, 0.0315, 0.0302, 0.0323, 0.0351, 0.0216, 0.0375], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:06:42,242 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4642, 3.3309, 3.6562, 1.7032, 3.8321, 3.8578, 2.9092, 2.7803], device='cuda:5'), covar=tensor([0.0756, 0.0225, 0.0184, 0.1258, 0.0060, 0.0159, 0.0410, 0.0471], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0104, 0.0092, 0.0136, 0.0074, 0.0118, 0.0124, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 14:07:07,224 INFO [train.py:904] (5/8) Epoch 17, batch 7800, loss[loss=0.2031, simple_loss=0.2871, pruned_loss=0.05953, over 16690.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2937, pruned_loss=0.06277, over 3105846.24 frames. ], batch size: 134, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:07:07,725 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4033, 5.7362, 5.4593, 5.5454, 5.2059, 5.1298, 5.1585, 5.8331], device='cuda:5'), covar=tensor([0.1272, 0.0815, 0.1037, 0.0826, 0.0837, 0.0728, 0.1142, 0.0873], device='cuda:5'), in_proj_covar=tensor([0.0628, 0.0766, 0.0628, 0.0567, 0.0482, 0.0494, 0.0634, 0.0591], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:07:55,102 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 14:08:03,355 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.635e+02 2.884e+02 3.642e+02 4.850e+02 8.293e+02, threshold=7.285e+02, percent-clipped=4.0 2023-04-30 14:08:23,356 INFO [train.py:904] (5/8) Epoch 17, batch 7850, loss[loss=0.2178, simple_loss=0.3032, pruned_loss=0.0662, over 16550.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2947, pruned_loss=0.06275, over 3097739.30 frames. ], batch size: 62, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:08:24,069 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7082, 3.7710, 2.9293, 2.2247, 2.5289, 2.3724, 4.0466, 3.3656], device='cuda:5'), covar=tensor([0.2715, 0.0724, 0.1682, 0.2603, 0.2604, 0.1997, 0.0452, 0.1258], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0262, 0.0299, 0.0301, 0.0291, 0.0244, 0.0286, 0.0323], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 14:08:35,535 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0800, 2.3517, 2.5775, 1.9449, 2.6681, 2.7955, 2.4282, 2.4023], device='cuda:5'), covar=tensor([0.0656, 0.0213, 0.0228, 0.0901, 0.0095, 0.0264, 0.0412, 0.0419], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0104, 0.0092, 0.0136, 0.0074, 0.0117, 0.0124, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 14:08:44,437 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7715, 3.8162, 3.9395, 3.7239, 3.8658, 4.2544, 3.8572, 3.6578], device='cuda:5'), covar=tensor([0.2250, 0.2062, 0.2414, 0.2497, 0.2713, 0.1807, 0.1825, 0.2606], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0560, 0.0615, 0.0469, 0.0631, 0.0648, 0.0488, 0.0630], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 14:09:00,978 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:09:38,292 INFO [train.py:904] (5/8) Epoch 17, batch 7900, loss[loss=0.1917, simple_loss=0.2831, pruned_loss=0.05014, over 16760.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2932, pruned_loss=0.06197, over 3102408.28 frames. ], batch size: 83, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:10:08,771 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 14:10:13,069 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170325.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:10:36,115 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.088e+02 2.800e+02 3.465e+02 4.175e+02 8.586e+02, threshold=6.929e+02, percent-clipped=4.0 2023-04-30 14:10:37,741 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 14:10:55,572 INFO [train.py:904] (5/8) Epoch 17, batch 7950, loss[loss=0.2068, simple_loss=0.2943, pruned_loss=0.05964, over 16835.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2942, pruned_loss=0.06311, over 3083952.51 frames. ], batch size: 102, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:11:17,290 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6563, 2.3604, 1.9495, 2.0420, 2.6595, 2.2607, 2.5258, 2.8057], device='cuda:5'), covar=tensor([0.0182, 0.0371, 0.0483, 0.0471, 0.0251, 0.0377, 0.0213, 0.0217], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0221, 0.0214, 0.0216, 0.0221, 0.0220, 0.0222, 0.0215], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:11:26,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9782, 5.2981, 5.0810, 5.0656, 4.8203, 4.7684, 4.7323, 5.3860], device='cuda:5'), covar=tensor([0.1248, 0.0839, 0.0961, 0.0868, 0.0779, 0.0846, 0.1083, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0626, 0.0763, 0.0625, 0.0567, 0.0479, 0.0492, 0.0632, 0.0591], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:11:49,341 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:12:13,370 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:12:13,812 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 14:12:14,142 INFO [train.py:904] (5/8) Epoch 17, batch 8000, loss[loss=0.1983, simple_loss=0.285, pruned_loss=0.05581, over 16455.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2949, pruned_loss=0.06353, over 3087722.41 frames. ], batch size: 68, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:12:15,924 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170403.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:00,997 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:11,711 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.905e+02 3.477e+02 4.096e+02 7.787e+02, threshold=6.954e+02, percent-clipped=1.0 2023-04-30 14:13:15,296 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170442.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:29,062 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170451.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:29,857 INFO [train.py:904] (5/8) Epoch 17, batch 8050, loss[loss=0.2275, simple_loss=0.2963, pruned_loss=0.0793, over 11690.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2954, pruned_loss=0.0636, over 3074412.06 frames. ], batch size: 248, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:13:44,297 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170461.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:28,185 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:46,137 INFO [train.py:904] (5/8) Epoch 17, batch 8100, loss[loss=0.184, simple_loss=0.2782, pruned_loss=0.04488, over 16823.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2946, pruned_loss=0.06258, over 3095914.40 frames. ], batch size: 102, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:15:16,599 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170522.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 14:15:42,386 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.570e+02 3.195e+02 3.934e+02 7.205e+02, threshold=6.390e+02, percent-clipped=3.0 2023-04-30 14:16:00,985 INFO [train.py:904] (5/8) Epoch 17, batch 8150, loss[loss=0.2106, simple_loss=0.2878, pruned_loss=0.06672, over 16619.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2929, pruned_loss=0.06257, over 3069831.40 frames. ], batch size: 62, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:16:27,627 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170569.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:17:18,501 INFO [train.py:904] (5/8) Epoch 17, batch 8200, loss[loss=0.2026, simple_loss=0.2936, pruned_loss=0.05581, over 15375.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2908, pruned_loss=0.06209, over 3060092.28 frames. ], batch size: 191, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:18:05,953 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170630.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:18:21,308 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.868e+02 2.599e+02 3.232e+02 3.801e+02 6.958e+02, threshold=6.463e+02, percent-clipped=1.0 2023-04-30 14:18:41,084 INFO [train.py:904] (5/8) Epoch 17, batch 8250, loss[loss=0.1744, simple_loss=0.2703, pruned_loss=0.03929, over 16793.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2894, pruned_loss=0.05927, over 3050192.09 frames. ], batch size: 89, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:20:04,608 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:20:05,306 INFO [train.py:904] (5/8) Epoch 17, batch 8300, loss[loss=0.1796, simple_loss=0.2738, pruned_loss=0.04268, over 15444.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2867, pruned_loss=0.05604, over 3057575.93 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:20:19,795 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-30 14:21:09,572 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.327e+02 2.774e+02 3.132e+02 7.452e+02, threshold=5.547e+02, percent-clipped=1.0 2023-04-30 14:21:26,074 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:21:29,860 INFO [train.py:904] (5/8) Epoch 17, batch 8350, loss[loss=0.2003, simple_loss=0.2965, pruned_loss=0.05204, over 15225.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2862, pruned_loss=0.05404, over 3060797.09 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:21:46,829 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1841, 2.3989, 2.6213, 1.9581, 2.7430, 2.8268, 2.5353, 2.4715], device='cuda:5'), covar=tensor([0.0594, 0.0215, 0.0241, 0.0943, 0.0083, 0.0256, 0.0379, 0.0424], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0101, 0.0090, 0.0133, 0.0072, 0.0114, 0.0120, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 14:21:54,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0234, 2.3587, 1.9353, 2.0518, 2.6515, 2.3401, 2.6774, 2.8306], device='cuda:5'), covar=tensor([0.0151, 0.0370, 0.0477, 0.0448, 0.0263, 0.0363, 0.0206, 0.0260], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0219, 0.0213, 0.0214, 0.0219, 0.0218, 0.0219, 0.0212], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:22:51,865 INFO [train.py:904] (5/8) Epoch 17, batch 8400, loss[loss=0.1941, simple_loss=0.2932, pruned_loss=0.04748, over 16735.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2829, pruned_loss=0.05165, over 3059417.44 frames. ], batch size: 124, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:23:16,991 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170817.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 14:23:52,183 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.262e+02 2.769e+02 3.288e+02 8.219e+02, threshold=5.539e+02, percent-clipped=2.0 2023-04-30 14:24:01,542 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 14:24:10,759 INFO [train.py:904] (5/8) Epoch 17, batch 8450, loss[loss=0.1786, simple_loss=0.2629, pruned_loss=0.04721, over 12260.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2807, pruned_loss=0.04975, over 3053630.09 frames. ], batch size: 246, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:25:32,481 INFO [train.py:904] (5/8) Epoch 17, batch 8500, loss[loss=0.1672, simple_loss=0.2555, pruned_loss=0.03946, over 16804.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2777, pruned_loss=0.04783, over 3076416.64 frames. ], batch size: 124, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:26:10,257 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:10,448 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7867, 1.3305, 1.7067, 1.7456, 1.8498, 1.9140, 1.5915, 1.7990], device='cuda:5'), covar=tensor([0.0220, 0.0348, 0.0201, 0.0237, 0.0238, 0.0156, 0.0376, 0.0117], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0182, 0.0170, 0.0171, 0.0182, 0.0140, 0.0185, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:26:13,395 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3106, 4.6419, 4.4608, 4.4536, 4.1675, 4.1609, 4.1786, 4.6797], device='cuda:5'), covar=tensor([0.1239, 0.1039, 0.0997, 0.0826, 0.0888, 0.1439, 0.1082, 0.0931], device='cuda:5'), in_proj_covar=tensor([0.0613, 0.0750, 0.0610, 0.0556, 0.0471, 0.0485, 0.0622, 0.0583], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:26:14,659 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:34,262 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.118e+02 2.662e+02 3.144e+02 4.893e+02, threshold=5.323e+02, percent-clipped=0.0 2023-04-30 14:26:56,015 INFO [train.py:904] (5/8) Epoch 17, batch 8550, loss[loss=0.1907, simple_loss=0.2899, pruned_loss=0.04571, over 16438.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2747, pruned_loss=0.04684, over 3045233.82 frames. ], batch size: 146, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:28:08,434 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170989.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:28:33,701 INFO [train.py:904] (5/8) Epoch 17, batch 8600, loss[loss=0.1753, simple_loss=0.263, pruned_loss=0.04378, over 12527.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2745, pruned_loss=0.04602, over 3038039.12 frames. ], batch size: 248, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:28:49,434 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8437, 5.1359, 4.9134, 4.8937, 4.6334, 4.5505, 4.5742, 5.2216], device='cuda:5'), covar=tensor([0.1107, 0.0911, 0.0983, 0.0818, 0.0772, 0.1023, 0.1116, 0.0817], device='cuda:5'), in_proj_covar=tensor([0.0607, 0.0744, 0.0605, 0.0551, 0.0466, 0.0481, 0.0616, 0.0578], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:29:52,153 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.454e+02 2.845e+02 3.449e+02 5.171e+02, threshold=5.691e+02, percent-clipped=0.0 2023-04-30 14:30:15,682 INFO [train.py:904] (5/8) Epoch 17, batch 8650, loss[loss=0.1645, simple_loss=0.2687, pruned_loss=0.03016, over 15225.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.273, pruned_loss=0.04466, over 3044614.25 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:31:44,083 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171092.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:32:02,421 INFO [train.py:904] (5/8) Epoch 17, batch 8700, loss[loss=0.1643, simple_loss=0.2631, pruned_loss=0.03272, over 15301.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2699, pruned_loss=0.04305, over 3047365.48 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:32:31,821 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171117.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 14:33:13,275 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.121e+02 2.631e+02 3.199e+02 5.449e+02, threshold=5.263e+02, percent-clipped=0.0 2023-04-30 14:33:38,788 INFO [train.py:904] (5/8) Epoch 17, batch 8750, loss[loss=0.1709, simple_loss=0.2667, pruned_loss=0.03755, over 12263.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2702, pruned_loss=0.04278, over 3053385.39 frames. ], batch size: 248, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:33:43,300 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:34:13,579 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:35:21,584 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171196.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:35:33,200 INFO [train.py:904] (5/8) Epoch 17, batch 8800, loss[loss=0.1862, simple_loss=0.2819, pruned_loss=0.04526, over 16153.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2686, pruned_loss=0.04147, over 3064454.33 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:35:57,147 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-04-30 14:36:22,731 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:36:54,980 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.375e+02 2.916e+02 3.471e+02 5.213e+02, threshold=5.833e+02, percent-clipped=0.0 2023-04-30 14:37:18,440 INFO [train.py:904] (5/8) Epoch 17, batch 8850, loss[loss=0.1832, simple_loss=0.2857, pruned_loss=0.0404, over 15206.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2715, pruned_loss=0.04119, over 3054807.84 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:37:29,119 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:03,247 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:27,527 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:39:01,364 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-04-30 14:39:01,956 INFO [train.py:904] (5/8) Epoch 17, batch 8900, loss[loss=0.1773, simple_loss=0.2711, pruned_loss=0.04173, over 16903.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2715, pruned_loss=0.04054, over 3059569.19 frames. ], batch size: 116, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:39:26,383 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 14:40:38,053 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 14:40:38,366 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 2.247e+02 2.656e+02 3.091e+02 5.659e+02, threshold=5.312e+02, percent-clipped=0.0 2023-04-30 14:40:51,331 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171345.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:41:04,768 INFO [train.py:904] (5/8) Epoch 17, batch 8950, loss[loss=0.1547, simple_loss=0.2542, pruned_loss=0.02758, over 16775.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2714, pruned_loss=0.04115, over 3068020.57 frames. ], batch size: 83, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:53,012 INFO [train.py:904] (5/8) Epoch 17, batch 9000, loss[loss=0.1374, simple_loss=0.235, pruned_loss=0.01991, over 16908.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2682, pruned_loss=0.03973, over 3057292.87 frames. ], batch size: 102, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:53,012 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 14:43:02,949 INFO [train.py:938] (5/8) Epoch 17, validation: loss=0.1478, simple_loss=0.2519, pruned_loss=0.02187, over 944034.00 frames. 2023-04-30 14:43:02,950 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 14:43:11,064 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:43:24,960 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0023, 3.2370, 3.2235, 2.1472, 2.9710, 3.2558, 3.1410, 1.9683], device='cuda:5'), covar=tensor([0.0509, 0.0038, 0.0050, 0.0392, 0.0101, 0.0070, 0.0069, 0.0424], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0074, 0.0075, 0.0129, 0.0089, 0.0099, 0.0087, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 14:43:42,512 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-30 14:44:23,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 2.012e+02 2.627e+02 3.134e+02 6.797e+02, threshold=5.255e+02, percent-clipped=3.0 2023-04-30 14:44:41,571 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171448.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:44:45,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9856, 3.2171, 2.8354, 5.0738, 3.8433, 4.4634, 1.6707, 3.3109], device='cuda:5'), covar=tensor([0.1294, 0.0654, 0.1103, 0.0128, 0.0238, 0.0332, 0.1620, 0.0676], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0164, 0.0186, 0.0169, 0.0196, 0.0207, 0.0191, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 14:44:47,400 INFO [train.py:904] (5/8) Epoch 17, batch 9050, loss[loss=0.1622, simple_loss=0.2534, pruned_loss=0.03551, over 16775.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2692, pruned_loss=0.04028, over 3072629.74 frames. ], batch size: 76, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:44:48,890 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 14:44:59,116 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3515, 3.4189, 1.9918, 3.7334, 2.5132, 3.6972, 2.0467, 2.7156], device='cuda:5'), covar=tensor([0.0263, 0.0361, 0.1676, 0.0199, 0.0874, 0.0524, 0.1668, 0.0771], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0165, 0.0188, 0.0144, 0.0168, 0.0203, 0.0196, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 14:45:37,251 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 14:46:24,682 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:46:30,497 INFO [train.py:904] (5/8) Epoch 17, batch 9100, loss[loss=0.174, simple_loss=0.2596, pruned_loss=0.04423, over 12294.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2691, pruned_loss=0.0408, over 3072580.59 frames. ], batch size: 248, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:46:47,893 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0677, 3.1083, 1.9009, 3.3303, 2.2726, 3.3278, 2.1671, 2.5921], device='cuda:5'), covar=tensor([0.0281, 0.0334, 0.1538, 0.0199, 0.0775, 0.0489, 0.1356, 0.0714], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0165, 0.0188, 0.0144, 0.0167, 0.0202, 0.0195, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 14:47:33,892 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9832, 2.8324, 2.7293, 2.0552, 2.6059, 2.8365, 2.7484, 1.9434], device='cuda:5'), covar=tensor([0.0412, 0.0056, 0.0069, 0.0334, 0.0111, 0.0080, 0.0087, 0.0410], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0075, 0.0076, 0.0130, 0.0089, 0.0100, 0.0087, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 14:47:58,147 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5166, 1.7387, 2.1187, 2.5367, 2.4527, 2.8436, 1.9422, 2.7802], device='cuda:5'), covar=tensor([0.0201, 0.0461, 0.0316, 0.0272, 0.0308, 0.0153, 0.0478, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0182, 0.0170, 0.0171, 0.0182, 0.0139, 0.0185, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 14:48:02,855 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.178e+02 2.662e+02 3.383e+02 6.301e+02, threshold=5.324e+02, percent-clipped=5.0 2023-04-30 14:48:28,390 INFO [train.py:904] (5/8) Epoch 17, batch 9150, loss[loss=0.1649, simple_loss=0.2583, pruned_loss=0.03572, over 16877.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2695, pruned_loss=0.04041, over 3090261.72 frames. ], batch size: 124, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:48:28,920 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171552.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:48:44,365 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:48:44,597 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-04-30 14:49:18,475 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 14:49:37,018 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:49:44,734 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7261, 3.1124, 3.4443, 2.1513, 2.9726, 2.1523, 3.2654, 3.2141], device='cuda:5'), covar=tensor([0.0309, 0.0826, 0.0463, 0.1790, 0.0732, 0.0975, 0.0707, 0.0939], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0149, 0.0157, 0.0144, 0.0136, 0.0123, 0.0136, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 14:50:13,343 INFO [train.py:904] (5/8) Epoch 17, batch 9200, loss[loss=0.1822, simple_loss=0.2711, pruned_loss=0.04662, over 15265.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2657, pruned_loss=0.03963, over 3101437.32 frames. ], batch size: 191, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:51:10,343 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:51:27,839 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.277e+02 2.610e+02 3.229e+02 7.994e+02, threshold=5.220e+02, percent-clipped=2.0 2023-04-30 14:51:50,823 INFO [train.py:904] (5/8) Epoch 17, batch 9250, loss[loss=0.157, simple_loss=0.2504, pruned_loss=0.03178, over 16855.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.265, pruned_loss=0.0392, over 3094286.01 frames. ], batch size: 102, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:52:50,605 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-04-30 14:53:42,090 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:53:42,886 INFO [train.py:904] (5/8) Epoch 17, batch 9300, loss[loss=0.1649, simple_loss=0.2547, pruned_loss=0.03761, over 16675.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2628, pruned_loss=0.03847, over 3061131.76 frames. ], batch size: 134, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:55:09,819 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.061e+02 2.530e+02 2.950e+02 5.223e+02, threshold=5.060e+02, percent-clipped=1.0 2023-04-30 14:55:21,351 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171748.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:55:27,412 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 14:55:27,781 INFO [train.py:904] (5/8) Epoch 17, batch 9350, loss[loss=0.1979, simple_loss=0.2939, pruned_loss=0.051, over 16418.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2628, pruned_loss=0.03856, over 3067077.46 frames. ], batch size: 146, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:55:43,619 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4025, 3.4347, 1.7937, 3.8251, 2.5537, 3.6972, 1.9790, 2.7319], device='cuda:5'), covar=tensor([0.0250, 0.0369, 0.1786, 0.0187, 0.0809, 0.0532, 0.1701, 0.0674], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0165, 0.0187, 0.0145, 0.0167, 0.0202, 0.0195, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 14:56:24,734 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 14:56:58,065 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:57:07,785 INFO [train.py:904] (5/8) Epoch 17, batch 9400, loss[loss=0.1773, simple_loss=0.2744, pruned_loss=0.0401, over 16340.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.263, pruned_loss=0.03826, over 3070923.10 frames. ], batch size: 146, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:29,676 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.242e+02 2.701e+02 3.369e+02 7.708e+02, threshold=5.403e+02, percent-clipped=2.0 2023-04-30 14:58:45,813 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-30 14:58:48,380 INFO [train.py:904] (5/8) Epoch 17, batch 9450, loss[loss=0.1858, simple_loss=0.2766, pruned_loss=0.04745, over 16618.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2653, pruned_loss=0.03834, over 3083970.43 frames. ], batch size: 76, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:49,587 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171852.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:58:52,229 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:59:05,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5763, 3.5300, 3.5042, 2.8749, 3.4271, 1.9674, 3.2399, 2.9427], device='cuda:5'), covar=tensor([0.0118, 0.0106, 0.0157, 0.0190, 0.0099, 0.2376, 0.0126, 0.0241], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0133, 0.0176, 0.0159, 0.0152, 0.0190, 0.0164, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:00:27,218 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171900.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:00:31,026 INFO [train.py:904] (5/8) Epoch 17, batch 9500, loss[loss=0.1704, simple_loss=0.2656, pruned_loss=0.03762, over 16515.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2642, pruned_loss=0.03798, over 3065554.06 frames. ], batch size: 68, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:01:06,821 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2082, 2.0517, 2.0693, 3.8051, 2.0860, 2.3521, 2.1720, 2.1772], device='cuda:5'), covar=tensor([0.1161, 0.3725, 0.3080, 0.0513, 0.4324, 0.2704, 0.3614, 0.3823], device='cuda:5'), in_proj_covar=tensor([0.0370, 0.0413, 0.0345, 0.0309, 0.0416, 0.0473, 0.0384, 0.0479], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:01:19,741 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5536, 3.6463, 3.4424, 3.1446, 3.2467, 3.5616, 3.3364, 3.3722], device='cuda:5'), covar=tensor([0.0584, 0.0558, 0.0315, 0.0277, 0.0568, 0.0447, 0.1305, 0.0482], device='cuda:5'), in_proj_covar=tensor([0.0257, 0.0361, 0.0305, 0.0291, 0.0310, 0.0340, 0.0209, 0.0358], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-30 15:01:41,248 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6822, 2.6966, 1.8271, 2.8395, 2.0941, 2.8367, 2.0862, 2.4182], device='cuda:5'), covar=tensor([0.0290, 0.0345, 0.1340, 0.0255, 0.0693, 0.0638, 0.1227, 0.0601], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0165, 0.0187, 0.0145, 0.0168, 0.0204, 0.0196, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 15:01:41,253 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:01:51,141 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.125e+02 2.584e+02 3.095e+02 6.778e+02, threshold=5.168e+02, percent-clipped=1.0 2023-04-30 15:02:06,944 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 15:02:14,953 INFO [train.py:904] (5/8) Epoch 17, batch 9550, loss[loss=0.187, simple_loss=0.2896, pruned_loss=0.04224, over 15583.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2646, pruned_loss=0.03835, over 3069342.44 frames. ], batch size: 194, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:03:26,197 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 15:03:49,268 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171998.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:03:57,504 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172001.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:03:58,213 INFO [train.py:904] (5/8) Epoch 17, batch 9600, loss[loss=0.1822, simple_loss=0.2845, pruned_loss=0.03999, over 16347.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2655, pruned_loss=0.03906, over 3036212.63 frames. ], batch size: 146, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:05:20,544 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.253e+02 2.956e+02 3.661e+02 6.195e+02, threshold=5.912e+02, percent-clipped=3.0 2023-04-30 15:05:38,180 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172049.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:05:44,197 INFO [train.py:904] (5/8) Epoch 17, batch 9650, loss[loss=0.1702, simple_loss=0.2687, pruned_loss=0.0359, over 16253.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2675, pruned_loss=0.03952, over 3034479.08 frames. ], batch size: 165, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:06:37,762 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7866, 3.7048, 3.8729, 3.9718, 4.0709, 3.6919, 4.0514, 4.0947], device='cuda:5'), covar=tensor([0.1685, 0.1096, 0.1362, 0.0709, 0.0591, 0.1641, 0.0633, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0563, 0.0690, 0.0811, 0.0706, 0.0531, 0.0557, 0.0569, 0.0664], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:07:32,490 INFO [train.py:904] (5/8) Epoch 17, batch 9700, loss[loss=0.1657, simple_loss=0.2764, pruned_loss=0.02754, over 16848.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2667, pruned_loss=0.03923, over 3055772.78 frames. ], batch size: 102, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:08:00,947 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2290, 2.1443, 2.0301, 3.9213, 2.0476, 2.5484, 2.2223, 2.2901], device='cuda:5'), covar=tensor([0.1074, 0.3518, 0.3082, 0.0419, 0.4318, 0.2260, 0.3481, 0.3068], device='cuda:5'), in_proj_covar=tensor([0.0370, 0.0412, 0.0344, 0.0308, 0.0416, 0.0471, 0.0383, 0.0477], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:08:05,726 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172118.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:08:57,403 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.330e+02 2.721e+02 3.165e+02 5.976e+02, threshold=5.442e+02, percent-clipped=1.0 2023-04-30 15:09:14,198 INFO [train.py:904] (5/8) Epoch 17, batch 9750, loss[loss=0.1641, simple_loss=0.2527, pruned_loss=0.03781, over 12291.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2654, pruned_loss=0.03952, over 3026029.17 frames. ], batch size: 247, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:09:18,929 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:09:57,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4463, 3.3897, 3.5020, 3.5462, 3.6057, 3.3318, 3.5874, 3.6485], device='cuda:5'), covar=tensor([0.1247, 0.0887, 0.1050, 0.0637, 0.0587, 0.2275, 0.0796, 0.0702], device='cuda:5'), in_proj_covar=tensor([0.0560, 0.0688, 0.0810, 0.0707, 0.0530, 0.0555, 0.0569, 0.0662], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:10:10,095 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:10:54,105 INFO [train.py:904] (5/8) Epoch 17, batch 9800, loss[loss=0.1825, simple_loss=0.26, pruned_loss=0.05253, over 11982.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2647, pruned_loss=0.0384, over 3037817.62 frames. ], batch size: 247, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:10:54,781 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:12:17,140 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 2.047e+02 2.504e+02 3.250e+02 5.993e+02, threshold=5.009e+02, percent-clipped=1.0 2023-04-30 15:12:39,004 INFO [train.py:904] (5/8) Epoch 17, batch 9850, loss[loss=0.1822, simple_loss=0.2737, pruned_loss=0.04534, over 16947.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2655, pruned_loss=0.03795, over 3043168.90 frames. ], batch size: 109, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:14:14,447 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:14:32,012 INFO [train.py:904] (5/8) Epoch 17, batch 9900, loss[loss=0.1559, simple_loss=0.2496, pruned_loss=0.0311, over 12317.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2659, pruned_loss=0.03796, over 3038005.45 frames. ], batch size: 247, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:14:43,141 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8873, 2.1762, 1.7437, 1.9837, 2.5666, 2.1857, 2.4178, 2.6776], device='cuda:5'), covar=tensor([0.0144, 0.0403, 0.0550, 0.0453, 0.0290, 0.0381, 0.0191, 0.0240], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0221, 0.0212, 0.0213, 0.0219, 0.0218, 0.0214, 0.0209], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:16:10,690 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.496e+02 2.206e+02 2.589e+02 2.968e+02 8.308e+02, threshold=5.178e+02, percent-clipped=1.0 2023-04-30 15:16:31,661 INFO [train.py:904] (5/8) Epoch 17, batch 9950, loss[loss=0.1611, simple_loss=0.2608, pruned_loss=0.03069, over 16617.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2686, pruned_loss=0.03816, over 3069791.61 frames. ], batch size: 68, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:16:55,925 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0759, 2.6404, 2.8756, 1.9724, 2.6772, 2.0235, 2.7144, 2.8807], device='cuda:5'), covar=tensor([0.0268, 0.0878, 0.0540, 0.1917, 0.0722, 0.0966, 0.0614, 0.0843], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0148, 0.0158, 0.0145, 0.0136, 0.0123, 0.0136, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 15:18:33,066 INFO [train.py:904] (5/8) Epoch 17, batch 10000, loss[loss=0.1827, simple_loss=0.2634, pruned_loss=0.05103, over 12585.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2673, pruned_loss=0.03812, over 3066826.12 frames. ], batch size: 250, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:19:56,018 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 2.112e+02 2.435e+02 3.073e+02 5.359e+02, threshold=4.869e+02, percent-clipped=1.0 2023-04-30 15:20:13,527 INFO [train.py:904] (5/8) Epoch 17, batch 10050, loss[loss=0.1941, simple_loss=0.2951, pruned_loss=0.04651, over 16357.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2674, pruned_loss=0.03791, over 3077174.04 frames. ], batch size: 146, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:20:58,577 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:21:17,101 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1891, 1.4918, 1.9273, 2.1387, 2.1698, 2.3534, 1.8167, 2.2507], device='cuda:5'), covar=tensor([0.0226, 0.0463, 0.0266, 0.0297, 0.0297, 0.0198, 0.0408, 0.0129], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0179, 0.0166, 0.0167, 0.0178, 0.0136, 0.0182, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:21:47,158 INFO [train.py:904] (5/8) Epoch 17, batch 10100, loss[loss=0.1591, simple_loss=0.2478, pruned_loss=0.03517, over 12521.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2673, pruned_loss=0.03782, over 3088109.65 frames. ], batch size: 249, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:22:57,876 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.137e+02 2.474e+02 3.171e+02 5.916e+02, threshold=4.949e+02, percent-clipped=1.0 2023-04-30 15:23:33,090 INFO [train.py:904] (5/8) Epoch 18, batch 0, loss[loss=0.2575, simple_loss=0.3178, pruned_loss=0.09858, over 16829.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.3178, pruned_loss=0.09858, over 16829.00 frames. ], batch size: 96, lr: 3.82e-03, grad_scale: 8.0 2023-04-30 15:23:33,091 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 15:23:40,343 INFO [train.py:938] (5/8) Epoch 18, validation: loss=0.148, simple_loss=0.2515, pruned_loss=0.02223, over 944034.00 frames. 2023-04-30 15:23:40,344 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 15:24:36,460 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172593.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:24:44,257 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:24:50,182 INFO [train.py:904] (5/8) Epoch 18, batch 50, loss[loss=0.2055, simple_loss=0.2782, pruned_loss=0.06638, over 16730.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.277, pruned_loss=0.05419, over 747337.61 frames. ], batch size: 134, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:25:21,776 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:42,962 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172641.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:48,588 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.462e+02 2.938e+02 3.654e+02 9.185e+02, threshold=5.877e+02, percent-clipped=7.0 2023-04-30 15:25:50,888 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9101, 4.6031, 4.9514, 5.1101, 5.3257, 4.6567, 5.3011, 5.2805], device='cuda:5'), covar=tensor([0.1804, 0.1324, 0.1531, 0.0755, 0.0556, 0.0861, 0.0545, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0561, 0.0694, 0.0813, 0.0708, 0.0530, 0.0558, 0.0573, 0.0665], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:25:56,045 INFO [train.py:904] (5/8) Epoch 18, batch 100, loss[loss=0.1963, simple_loss=0.2842, pruned_loss=0.0542, over 16730.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2723, pruned_loss=0.05073, over 1318749.65 frames. ], batch size: 57, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:25:59,596 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5470, 2.2519, 2.3572, 4.5078, 2.2083, 2.6260, 2.3179, 2.4247], device='cuda:5'), covar=tensor([0.1091, 0.3622, 0.2760, 0.0378, 0.4129, 0.2600, 0.3397, 0.3533], device='cuda:5'), in_proj_covar=tensor([0.0374, 0.0415, 0.0347, 0.0311, 0.0418, 0.0474, 0.0385, 0.0481], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:26:05,985 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:26:44,057 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:27:02,728 INFO [train.py:904] (5/8) Epoch 18, batch 150, loss[loss=0.1715, simple_loss=0.2628, pruned_loss=0.04006, over 17269.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2702, pruned_loss=0.04868, over 1766337.38 frames. ], batch size: 52, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:27:10,976 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-30 15:27:46,449 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:28:01,450 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.271e+02 2.721e+02 3.424e+02 7.505e+02, threshold=5.441e+02, percent-clipped=3.0 2023-04-30 15:28:09,990 INFO [train.py:904] (5/8) Epoch 18, batch 200, loss[loss=0.1991, simple_loss=0.2853, pruned_loss=0.05645, over 11841.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2688, pruned_loss=0.04881, over 2111925.66 frames. ], batch size: 245, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:28:18,918 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7388, 2.9864, 2.8743, 5.0554, 4.0859, 4.4505, 1.5185, 3.1822], device='cuda:5'), covar=tensor([0.1417, 0.0711, 0.1118, 0.0175, 0.0300, 0.0413, 0.1660, 0.0833], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0165, 0.0187, 0.0169, 0.0194, 0.0208, 0.0192, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 15:28:30,055 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3528, 4.1413, 4.3698, 4.5355, 4.6247, 4.1916, 4.4732, 4.6199], device='cuda:5'), covar=tensor([0.1674, 0.1249, 0.1347, 0.0725, 0.0681, 0.1285, 0.2289, 0.0757], device='cuda:5'), in_proj_covar=tensor([0.0573, 0.0708, 0.0832, 0.0722, 0.0540, 0.0570, 0.0584, 0.0679], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:28:41,325 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:28:43,730 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0437, 4.0557, 2.5193, 4.6013, 3.1468, 4.5145, 2.6502, 3.2881], device='cuda:5'), covar=tensor([0.0253, 0.0376, 0.1539, 0.0249, 0.0824, 0.0515, 0.1495, 0.0752], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0169, 0.0194, 0.0150, 0.0173, 0.0209, 0.0202, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 15:28:52,350 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5320, 4.6346, 4.7830, 4.5896, 4.6050, 5.2274, 4.7142, 4.4529], device='cuda:5'), covar=tensor([0.1588, 0.2133, 0.2692, 0.2404, 0.3364, 0.1444, 0.1914, 0.2657], device='cuda:5'), in_proj_covar=tensor([0.0374, 0.0548, 0.0601, 0.0456, 0.0616, 0.0638, 0.0477, 0.0613], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 15:29:10,039 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:29:10,430 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-30 15:29:18,650 INFO [train.py:904] (5/8) Epoch 18, batch 250, loss[loss=0.1494, simple_loss=0.2369, pruned_loss=0.03101, over 16951.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2658, pruned_loss=0.04824, over 2380263.99 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:29:29,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5131, 3.4470, 3.4715, 2.8687, 3.3124, 2.0840, 3.0917, 2.8049], device='cuda:5'), covar=tensor([0.0133, 0.0121, 0.0171, 0.0215, 0.0095, 0.2087, 0.0130, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0138, 0.0181, 0.0164, 0.0157, 0.0196, 0.0170, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:29:47,729 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:30:18,767 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172845.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:30:19,549 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 2.240e+02 2.726e+02 3.152e+02 7.045e+02, threshold=5.451e+02, percent-clipped=2.0 2023-04-30 15:30:23,651 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7196, 3.6338, 2.5910, 2.1388, 2.2753, 2.0455, 3.5721, 3.0738], device='cuda:5'), covar=tensor([0.2552, 0.0659, 0.1965, 0.2940, 0.2715, 0.2394, 0.0654, 0.1554], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0255, 0.0293, 0.0294, 0.0278, 0.0241, 0.0279, 0.0315], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 15:30:29,005 INFO [train.py:904] (5/8) Epoch 18, batch 300, loss[loss=0.186, simple_loss=0.2597, pruned_loss=0.05609, over 16750.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.264, pruned_loss=0.04737, over 2581179.25 frames. ], batch size: 134, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:39,800 INFO [train.py:904] (5/8) Epoch 18, batch 350, loss[loss=0.1632, simple_loss=0.2466, pruned_loss=0.03989, over 16585.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2609, pruned_loss=0.04618, over 2742459.09 frames. ], batch size: 68, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:45,408 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172906.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:32:27,312 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8523, 3.7087, 4.1691, 1.9993, 4.3738, 4.4010, 3.1859, 3.3386], device='cuda:5'), covar=tensor([0.0730, 0.0244, 0.0213, 0.1263, 0.0061, 0.0162, 0.0427, 0.0405], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0105, 0.0091, 0.0138, 0.0074, 0.0118, 0.0125, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 15:32:40,487 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.138e+02 2.555e+02 2.961e+02 4.551e+02, threshold=5.109e+02, percent-clipped=0.0 2023-04-30 15:32:47,908 INFO [train.py:904] (5/8) Epoch 18, batch 400, loss[loss=0.158, simple_loss=0.2442, pruned_loss=0.0359, over 17202.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2588, pruned_loss=0.04537, over 2869823.07 frames. ], batch size: 44, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:32:51,750 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:10,321 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4417, 4.3229, 4.3317, 4.0404, 4.0479, 4.3698, 4.1290, 4.1326], device='cuda:5'), covar=tensor([0.0583, 0.0681, 0.0331, 0.0283, 0.0772, 0.0461, 0.0668, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0273, 0.0384, 0.0321, 0.0308, 0.0330, 0.0358, 0.0219, 0.0381], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:33:30,498 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:44,173 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-04-30 15:33:59,540 INFO [train.py:904] (5/8) Epoch 18, batch 450, loss[loss=0.1488, simple_loss=0.2329, pruned_loss=0.03236, over 17265.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2575, pruned_loss=0.04476, over 2974637.27 frames. ], batch size: 45, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:35:00,083 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.073e+02 2.463e+02 2.995e+02 6.633e+02, threshold=4.925e+02, percent-clipped=1.0 2023-04-30 15:35:08,312 INFO [train.py:904] (5/8) Epoch 18, batch 500, loss[loss=0.177, simple_loss=0.257, pruned_loss=0.04855, over 16820.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2564, pruned_loss=0.04416, over 3052703.63 frames. ], batch size: 96, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:35:22,637 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0157, 3.9256, 4.4167, 2.2307, 4.6292, 4.6298, 3.2802, 3.6103], device='cuda:5'), covar=tensor([0.0718, 0.0247, 0.0267, 0.1150, 0.0057, 0.0165, 0.0421, 0.0374], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0107, 0.0093, 0.0141, 0.0076, 0.0121, 0.0127, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 15:35:28,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5767, 3.4538, 2.6302, 2.1191, 2.2814, 2.1766, 3.5337, 3.0491], device='cuda:5'), covar=tensor([0.2949, 0.0717, 0.1889, 0.2796, 0.2885, 0.2257, 0.0591, 0.1534], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0258, 0.0296, 0.0298, 0.0282, 0.0243, 0.0281, 0.0319], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 15:35:59,655 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:36:14,716 INFO [train.py:904] (5/8) Epoch 18, batch 550, loss[loss=0.1669, simple_loss=0.244, pruned_loss=0.04488, over 16878.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2556, pruned_loss=0.04336, over 3119206.52 frames. ], batch size: 96, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:36:34,001 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.98 vs. limit=5.0 2023-04-30 15:37:14,574 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.114e+02 2.403e+02 3.161e+02 6.470e+02, threshold=4.805e+02, percent-clipped=1.0 2023-04-30 15:37:20,945 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173150.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:37:22,711 INFO [train.py:904] (5/8) Epoch 18, batch 600, loss[loss=0.1536, simple_loss=0.2406, pruned_loss=0.0333, over 17178.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2553, pruned_loss=0.0436, over 3169009.52 frames. ], batch size: 46, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:37:45,902 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3580, 2.1008, 2.2966, 4.0549, 2.2494, 2.5071, 2.2117, 2.3328], device='cuda:5'), covar=tensor([0.1342, 0.3992, 0.2842, 0.0580, 0.4005, 0.2542, 0.3855, 0.3161], device='cuda:5'), in_proj_covar=tensor([0.0383, 0.0425, 0.0354, 0.0320, 0.0425, 0.0488, 0.0394, 0.0494], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:37:48,696 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5297, 4.6540, 4.7620, 4.5861, 4.6532, 5.2159, 4.7266, 4.3825], device='cuda:5'), covar=tensor([0.1800, 0.2234, 0.2613, 0.2661, 0.3081, 0.1396, 0.1996, 0.3041], device='cuda:5'), in_proj_covar=tensor([0.0386, 0.0564, 0.0619, 0.0468, 0.0634, 0.0653, 0.0492, 0.0631], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 15:38:08,492 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-30 15:38:30,580 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173201.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:38:31,382 INFO [train.py:904] (5/8) Epoch 18, batch 650, loss[loss=0.1447, simple_loss=0.2395, pruned_loss=0.02492, over 17204.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2543, pruned_loss=0.04291, over 3203413.85 frames. ], batch size: 44, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:38:45,556 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:39:32,446 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.266e+02 2.757e+02 3.739e+02 8.562e+02, threshold=5.513e+02, percent-clipped=7.0 2023-04-30 15:39:32,824 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1367, 5.0828, 4.9665, 4.4711, 4.5412, 5.0211, 4.9158, 4.6313], device='cuda:5'), covar=tensor([0.0586, 0.0486, 0.0291, 0.0326, 0.1091, 0.0472, 0.0351, 0.0716], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0396, 0.0329, 0.0317, 0.0339, 0.0368, 0.0225, 0.0393], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:39:40,836 INFO [train.py:904] (5/8) Epoch 18, batch 700, loss[loss=0.1689, simple_loss=0.2611, pruned_loss=0.03833, over 17112.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2543, pruned_loss=0.04301, over 3230760.48 frames. ], batch size: 49, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:39:44,112 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173254.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:39:58,115 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2434, 4.2894, 4.6715, 4.6368, 4.6761, 4.3493, 4.3690, 4.2968], device='cuda:5'), covar=tensor([0.0386, 0.0665, 0.0408, 0.0491, 0.0475, 0.0442, 0.0863, 0.0556], device='cuda:5'), in_proj_covar=tensor([0.0388, 0.0422, 0.0413, 0.0390, 0.0456, 0.0434, 0.0528, 0.0342], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 15:40:22,695 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:50,346 INFO [train.py:904] (5/8) Epoch 18, batch 750, loss[loss=0.1667, simple_loss=0.2441, pruned_loss=0.04469, over 16794.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2546, pruned_loss=0.04314, over 3258639.06 frames. ], batch size: 96, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:40:51,631 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173302.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:00,040 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:26,736 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:50,591 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.162e+02 2.529e+02 2.992e+02 5.194e+02, threshold=5.057e+02, percent-clipped=0.0 2023-04-30 15:41:58,288 INFO [train.py:904] (5/8) Epoch 18, batch 800, loss[loss=0.1789, simple_loss=0.2555, pruned_loss=0.0512, over 16442.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2546, pruned_loss=0.0429, over 3278606.52 frames. ], batch size: 75, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:42:22,708 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:42:49,217 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:43:05,173 INFO [train.py:904] (5/8) Epoch 18, batch 850, loss[loss=0.1448, simple_loss=0.2349, pruned_loss=0.02731, over 17218.00 frames. ], tot_loss[loss=0.17, simple_loss=0.254, pruned_loss=0.04295, over 3290154.17 frames. ], batch size: 45, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:43:55,076 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173438.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:44:07,467 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.076e+02 2.484e+02 3.047e+02 7.232e+02, threshold=4.969e+02, percent-clipped=3.0 2023-04-30 15:44:15,659 INFO [train.py:904] (5/8) Epoch 18, batch 900, loss[loss=0.1767, simple_loss=0.2493, pruned_loss=0.05205, over 16825.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2535, pruned_loss=0.04283, over 3290398.89 frames. ], batch size: 102, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:44:30,300 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 15:44:32,857 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6143, 4.4009, 4.6122, 4.8063, 4.9928, 4.4009, 4.8861, 4.9497], device='cuda:5'), covar=tensor([0.1642, 0.1245, 0.1595, 0.0815, 0.0614, 0.1317, 0.1413, 0.0928], device='cuda:5'), in_proj_covar=tensor([0.0608, 0.0748, 0.0883, 0.0760, 0.0567, 0.0602, 0.0617, 0.0720], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:44:52,461 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-04-30 15:45:11,310 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173491.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 15:45:24,949 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:45:25,651 INFO [train.py:904] (5/8) Epoch 18, batch 950, loss[loss=0.1642, simple_loss=0.2592, pruned_loss=0.03458, over 17173.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2536, pruned_loss=0.04285, over 3297961.17 frames. ], batch size: 46, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:45:31,779 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:46:26,076 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.297e+02 2.589e+02 3.111e+02 6.149e+02, threshold=5.178e+02, percent-clipped=2.0 2023-04-30 15:46:31,946 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173549.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:46:35,245 INFO [train.py:904] (5/8) Epoch 18, batch 1000, loss[loss=0.1535, simple_loss=0.2325, pruned_loss=0.03727, over 16317.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2527, pruned_loss=0.04271, over 3307484.30 frames. ], batch size: 165, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:46:35,731 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173552.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:47:44,531 INFO [train.py:904] (5/8) Epoch 18, batch 1050, loss[loss=0.1875, simple_loss=0.2613, pruned_loss=0.05688, over 16213.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2525, pruned_loss=0.0428, over 3308207.53 frames. ], batch size: 165, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:48:02,489 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:48:46,281 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.302e+02 2.654e+02 3.076e+02 5.985e+02, threshold=5.308e+02, percent-clipped=1.0 2023-04-30 15:48:54,964 INFO [train.py:904] (5/8) Epoch 18, batch 1100, loss[loss=0.157, simple_loss=0.2449, pruned_loss=0.03454, over 16836.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2525, pruned_loss=0.04266, over 3318274.44 frames. ], batch size: 42, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:49:12,461 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:49:26,686 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173675.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:50:01,553 INFO [train.py:904] (5/8) Epoch 18, batch 1150, loss[loss=0.1525, simple_loss=0.228, pruned_loss=0.03849, over 16741.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2519, pruned_loss=0.04206, over 3315588.24 frames. ], batch size: 83, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:51:02,217 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.041e+02 2.391e+02 2.834e+02 4.820e+02, threshold=4.781e+02, percent-clipped=0.0 2023-04-30 15:51:11,442 INFO [train.py:904] (5/8) Epoch 18, batch 1200, loss[loss=0.1847, simple_loss=0.2567, pruned_loss=0.0564, over 16764.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2503, pruned_loss=0.04131, over 3311742.95 frames. ], batch size: 124, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:51:33,556 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 15:52:19,915 INFO [train.py:904] (5/8) Epoch 18, batch 1250, loss[loss=0.2015, simple_loss=0.2677, pruned_loss=0.06768, over 16446.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2507, pruned_loss=0.04188, over 3310592.34 frames. ], batch size: 146, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:52:24,755 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:53:21,079 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.313e+02 2.602e+02 3.076e+02 5.284e+02, threshold=5.204e+02, percent-clipped=3.0 2023-04-30 15:53:22,614 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173847.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 15:53:28,944 INFO [train.py:904] (5/8) Epoch 18, batch 1300, loss[loss=0.1524, simple_loss=0.2336, pruned_loss=0.03562, over 15458.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2502, pruned_loss=0.04154, over 3305591.37 frames. ], batch size: 190, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:53:32,735 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:54:33,391 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6969, 3.5262, 3.9540, 2.0505, 4.1011, 4.0665, 3.2867, 3.1144], device='cuda:5'), covar=tensor([0.0684, 0.0247, 0.0189, 0.1110, 0.0075, 0.0196, 0.0318, 0.0383], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0107, 0.0095, 0.0140, 0.0076, 0.0121, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 15:54:37,702 INFO [train.py:904] (5/8) Epoch 18, batch 1350, loss[loss=0.1553, simple_loss=0.2425, pruned_loss=0.03407, over 16732.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2512, pruned_loss=0.04165, over 3314597.98 frames. ], batch size: 89, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:55:37,175 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 2.229e+02 2.739e+02 3.130e+02 5.090e+02, threshold=5.477e+02, percent-clipped=0.0 2023-04-30 15:55:44,253 INFO [train.py:904] (5/8) Epoch 18, batch 1400, loss[loss=0.1776, simple_loss=0.2518, pruned_loss=0.05173, over 16710.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2503, pruned_loss=0.04115, over 3308554.75 frames. ], batch size: 134, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:56:04,272 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:56:09,768 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173970.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:56:38,747 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8995, 1.9255, 2.4305, 2.7931, 2.5267, 3.3297, 2.2960, 3.3222], device='cuda:5'), covar=tensor([0.0234, 0.0480, 0.0332, 0.0309, 0.0357, 0.0172, 0.0442, 0.0142], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0190, 0.0175, 0.0178, 0.0187, 0.0146, 0.0192, 0.0140], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 15:56:56,570 INFO [train.py:904] (5/8) Epoch 18, batch 1450, loss[loss=0.176, simple_loss=0.2495, pruned_loss=0.05124, over 16753.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2496, pruned_loss=0.04106, over 3308404.89 frames. ], batch size: 83, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:57:11,462 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:14,641 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:16,978 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:54,756 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.047e+02 2.425e+02 3.116e+02 5.904e+02, threshold=4.850e+02, percent-clipped=1.0 2023-04-30 15:58:03,184 INFO [train.py:904] (5/8) Epoch 18, batch 1500, loss[loss=0.1512, simple_loss=0.2424, pruned_loss=0.03002, over 17292.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2488, pruned_loss=0.04095, over 3313939.51 frames. ], batch size: 52, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 15:58:36,104 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174076.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 15:58:38,843 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174078.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:59:10,949 INFO [train.py:904] (5/8) Epoch 18, batch 1550, loss[loss=0.1857, simple_loss=0.2597, pruned_loss=0.05585, over 16684.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2503, pruned_loss=0.04211, over 3302407.53 frames. ], batch size: 134, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 15:59:15,211 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2322, 2.8560, 3.1250, 1.6942, 3.2456, 3.2529, 2.7628, 2.6102], device='cuda:5'), covar=tensor([0.0853, 0.0272, 0.0230, 0.1256, 0.0109, 0.0245, 0.0439, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0107, 0.0094, 0.0140, 0.0076, 0.0121, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 15:59:47,034 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 16:00:00,546 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:00:09,839 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.219e+02 2.746e+02 3.375e+02 7.585e+02, threshold=5.492e+02, percent-clipped=5.0 2023-04-30 16:00:11,918 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174147.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:00:17,422 INFO [train.py:904] (5/8) Epoch 18, batch 1600, loss[loss=0.1629, simple_loss=0.2576, pruned_loss=0.03404, over 17239.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2529, pruned_loss=0.04322, over 3300183.92 frames. ], batch size: 52, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:00:33,457 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 16:01:16,802 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174195.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:01:23,750 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:01:25,629 INFO [train.py:904] (5/8) Epoch 18, batch 1650, loss[loss=0.1748, simple_loss=0.2635, pruned_loss=0.04305, over 16696.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2553, pruned_loss=0.04435, over 3304836.74 frames. ], batch size: 62, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:01:51,006 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6057, 3.9422, 4.2258, 2.0388, 4.5089, 4.6751, 3.3213, 3.4467], device='cuda:5'), covar=tensor([0.1066, 0.0189, 0.0284, 0.1309, 0.0082, 0.0145, 0.0418, 0.0480], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0107, 0.0094, 0.0140, 0.0076, 0.0122, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:02:23,839 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.550e+02 2.934e+02 3.695e+02 6.371e+02, threshold=5.868e+02, percent-clipped=1.0 2023-04-30 16:02:32,762 INFO [train.py:904] (5/8) Epoch 18, batch 1700, loss[loss=0.172, simple_loss=0.2626, pruned_loss=0.04068, over 17000.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2574, pruned_loss=0.04494, over 3307549.48 frames. ], batch size: 41, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:02:48,245 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3040, 4.1404, 4.3839, 4.5345, 4.6545, 4.2525, 4.4413, 4.6144], device='cuda:5'), covar=tensor([0.1516, 0.1171, 0.1364, 0.0703, 0.0570, 0.1052, 0.2533, 0.0660], device='cuda:5'), in_proj_covar=tensor([0.0622, 0.0769, 0.0906, 0.0780, 0.0579, 0.0616, 0.0630, 0.0737], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:02:57,371 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174270.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:03:11,395 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-30 16:03:15,695 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:25,054 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:28,428 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8855, 4.0324, 2.2869, 4.6131, 3.0450, 4.4497, 2.0147, 3.2182], device='cuda:5'), covar=tensor([0.0266, 0.0300, 0.1731, 0.0196, 0.0784, 0.0423, 0.2029, 0.0640], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0175, 0.0196, 0.0159, 0.0176, 0.0218, 0.0204, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:03:40,539 INFO [train.py:904] (5/8) Epoch 18, batch 1750, loss[loss=0.1396, simple_loss=0.2284, pruned_loss=0.0254, over 16827.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2576, pruned_loss=0.04476, over 3310689.28 frames. ], batch size: 42, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:04:01,319 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:04:36,751 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:04:38,536 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.402e+02 2.710e+02 3.242e+02 7.254e+02, threshold=5.419e+02, percent-clipped=1.0 2023-04-30 16:04:45,369 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1051, 5.1708, 5.6556, 5.6217, 5.6470, 5.2741, 5.2297, 5.0259], device='cuda:5'), covar=tensor([0.0320, 0.0602, 0.0387, 0.0467, 0.0483, 0.0341, 0.0913, 0.0406], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0438, 0.0425, 0.0398, 0.0473, 0.0449, 0.0543, 0.0355], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 16:04:46,166 INFO [train.py:904] (5/8) Epoch 18, batch 1800, loss[loss=0.1704, simple_loss=0.2526, pruned_loss=0.04407, over 16731.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.259, pruned_loss=0.04477, over 3305213.44 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:04:46,648 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174352.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 16:05:10,981 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174371.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 16:05:13,955 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174373.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:33,512 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:51,533 INFO [train.py:904] (5/8) Epoch 18, batch 1850, loss[loss=0.1771, simple_loss=0.2745, pruned_loss=0.03987, over 17261.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2593, pruned_loss=0.04388, over 3320110.13 frames. ], batch size: 52, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:05:53,803 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0167, 2.4055, 2.6669, 1.8782, 2.7993, 2.8056, 2.4483, 2.2775], device='cuda:5'), covar=tensor([0.0767, 0.0240, 0.0221, 0.0928, 0.0108, 0.0256, 0.0435, 0.0450], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0107, 0.0094, 0.0140, 0.0076, 0.0122, 0.0126, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:06:08,792 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 16:06:51,033 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.281e+02 2.184e+02 2.475e+02 2.961e+02 4.707e+02, threshold=4.951e+02, percent-clipped=0.0 2023-04-30 16:06:55,545 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174449.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:06:57,617 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2334, 4.1908, 4.1376, 3.6596, 4.1617, 1.7554, 3.9165, 3.7410], device='cuda:5'), covar=tensor([0.0108, 0.0099, 0.0161, 0.0221, 0.0097, 0.2619, 0.0125, 0.0200], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0145, 0.0191, 0.0173, 0.0166, 0.0202, 0.0181, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:06:58,377 INFO [train.py:904] (5/8) Epoch 18, batch 1900, loss[loss=0.1703, simple_loss=0.2567, pruned_loss=0.04191, over 16296.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2581, pruned_loss=0.04348, over 3311444.70 frames. ], batch size: 165, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:07:00,607 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2031, 2.0681, 2.2634, 3.8676, 2.2384, 2.4675, 2.2044, 2.3074], device='cuda:5'), covar=tensor([0.1373, 0.3827, 0.2870, 0.0606, 0.3798, 0.2499, 0.3821, 0.2914], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0431, 0.0359, 0.0326, 0.0430, 0.0496, 0.0400, 0.0503], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:07:16,575 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9349, 4.7373, 4.9924, 5.1789, 5.3808, 4.7687, 5.3469, 5.3203], device='cuda:5'), covar=tensor([0.1769, 0.1250, 0.1655, 0.0772, 0.0584, 0.0947, 0.0578, 0.0616], device='cuda:5'), in_proj_covar=tensor([0.0624, 0.0772, 0.0909, 0.0783, 0.0582, 0.0621, 0.0634, 0.0740], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:07:57,452 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:08:06,710 INFO [train.py:904] (5/8) Epoch 18, batch 1950, loss[loss=0.1942, simple_loss=0.2628, pruned_loss=0.06277, over 16755.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2582, pruned_loss=0.04327, over 3317962.75 frames. ], batch size: 124, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:08:45,964 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3529, 4.6953, 5.0293, 4.9448, 4.9580, 4.6529, 4.3647, 4.5020], device='cuda:5'), covar=tensor([0.0727, 0.0796, 0.0570, 0.0716, 0.0831, 0.0704, 0.1546, 0.0638], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0434, 0.0420, 0.0397, 0.0470, 0.0445, 0.0540, 0.0353], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 16:09:05,470 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.251e+02 2.513e+02 3.165e+02 5.916e+02, threshold=5.027e+02, percent-clipped=3.0 2023-04-30 16:09:11,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9517, 3.0434, 2.7368, 4.4562, 3.7578, 4.3068, 1.7234, 3.0733], device='cuda:5'), covar=tensor([0.1269, 0.0594, 0.1041, 0.0162, 0.0151, 0.0369, 0.1420, 0.0743], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0168, 0.0189, 0.0178, 0.0200, 0.0214, 0.0193, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:09:14,579 INFO [train.py:904] (5/8) Epoch 18, batch 2000, loss[loss=0.1768, simple_loss=0.2457, pruned_loss=0.05394, over 16873.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.258, pruned_loss=0.04334, over 3324374.25 frames. ], batch size: 109, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:10:23,415 INFO [train.py:904] (5/8) Epoch 18, batch 2050, loss[loss=0.1535, simple_loss=0.2465, pruned_loss=0.03024, over 17179.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2587, pruned_loss=0.04384, over 3318171.94 frames. ], batch size: 46, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:11:02,135 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8500, 1.9012, 2.3607, 2.7606, 2.7043, 2.8316, 2.0821, 3.0250], device='cuda:5'), covar=tensor([0.0185, 0.0461, 0.0321, 0.0255, 0.0277, 0.0228, 0.0459, 0.0131], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0190, 0.0176, 0.0181, 0.0188, 0.0148, 0.0193, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:11:17,532 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:11:26,650 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.708e+02 2.224e+02 2.527e+02 3.035e+02 6.825e+02, threshold=5.053e+02, percent-clipped=2.0 2023-04-30 16:11:28,723 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174647.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:11:35,058 INFO [train.py:904] (5/8) Epoch 18, batch 2100, loss[loss=0.1862, simple_loss=0.2705, pruned_loss=0.05092, over 16518.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2597, pruned_loss=0.044, over 3324495.75 frames. ], batch size: 75, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:12:02,072 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174671.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:12:06,174 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174673.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:12:45,817 INFO [train.py:904] (5/8) Epoch 18, batch 2150, loss[loss=0.1882, simple_loss=0.2754, pruned_loss=0.0505, over 17105.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2603, pruned_loss=0.04416, over 3321137.41 frames. ], batch size: 53, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:12:50,366 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174705.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:10,643 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:13,524 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:31,463 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-30 16:13:37,283 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-04-30 16:13:45,479 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174744.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:49,507 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.235e+02 2.673e+02 3.216e+02 6.695e+02, threshold=5.347e+02, percent-clipped=2.0 2023-04-30 16:13:56,941 INFO [train.py:904] (5/8) Epoch 18, batch 2200, loss[loss=0.1962, simple_loss=0.2766, pruned_loss=0.05789, over 16795.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2595, pruned_loss=0.04368, over 3325189.85 frames. ], batch size: 124, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:14:16,326 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:14:56,015 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:15:06,369 INFO [train.py:904] (5/8) Epoch 18, batch 2250, loss[loss=0.1914, simple_loss=0.2722, pruned_loss=0.05527, over 16881.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2611, pruned_loss=0.04484, over 3314366.54 frames. ], batch size: 116, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:16:02,788 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174843.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:16:06,392 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 16:16:07,690 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.164e+02 2.481e+02 3.118e+02 6.907e+02, threshold=4.963e+02, percent-clipped=2.0 2023-04-30 16:16:14,119 INFO [train.py:904] (5/8) Epoch 18, batch 2300, loss[loss=0.1794, simple_loss=0.2646, pruned_loss=0.04715, over 16738.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2604, pruned_loss=0.04482, over 3325271.72 frames. ], batch size: 89, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:17:22,134 INFO [train.py:904] (5/8) Epoch 18, batch 2350, loss[loss=0.1617, simple_loss=0.2505, pruned_loss=0.03644, over 15909.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2611, pruned_loss=0.04497, over 3322470.54 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:18:14,654 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:18:25,283 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.298e+02 2.659e+02 3.331e+02 6.396e+02, threshold=5.319e+02, percent-clipped=2.0 2023-04-30 16:18:25,777 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174947.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:18:32,765 INFO [train.py:904] (5/8) Epoch 18, batch 2400, loss[loss=0.1377, simple_loss=0.2312, pruned_loss=0.02208, over 16805.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2621, pruned_loss=0.04494, over 3329957.32 frames. ], batch size: 39, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:19:21,462 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:19:33,424 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174995.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:19:42,211 INFO [train.py:904] (5/8) Epoch 18, batch 2450, loss[loss=0.1609, simple_loss=0.2539, pruned_loss=0.03396, over 16991.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2623, pruned_loss=0.04436, over 3336402.02 frames. ], batch size: 41, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:20:00,597 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3955, 5.8066, 5.5829, 5.6443, 5.1911, 5.1354, 5.2006, 5.9638], device='cuda:5'), covar=tensor([0.1380, 0.1013, 0.1002, 0.0756, 0.0943, 0.0764, 0.1207, 0.0927], device='cuda:5'), in_proj_covar=tensor([0.0661, 0.0814, 0.0653, 0.0601, 0.0511, 0.0514, 0.0674, 0.0626], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:20:40,568 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:20:43,744 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.297e+02 2.808e+02 3.204e+02 6.791e+02, threshold=5.617e+02, percent-clipped=2.0 2023-04-30 16:20:51,501 INFO [train.py:904] (5/8) Epoch 18, batch 2500, loss[loss=0.1577, simple_loss=0.2591, pruned_loss=0.02814, over 17057.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2622, pruned_loss=0.04443, over 3328475.11 frames. ], batch size: 50, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:21:04,693 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175061.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:21:46,595 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175092.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:22:00,363 INFO [train.py:904] (5/8) Epoch 18, batch 2550, loss[loss=0.1846, simple_loss=0.2662, pruned_loss=0.05146, over 16871.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2633, pruned_loss=0.04484, over 3320044.75 frames. ], batch size: 116, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:22:30,173 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2618, 2.1276, 2.2766, 3.9778, 2.1625, 2.5313, 2.2108, 2.3370], device='cuda:5'), covar=tensor([0.1415, 0.3650, 0.2849, 0.0588, 0.3945, 0.2638, 0.3593, 0.3198], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0432, 0.0361, 0.0327, 0.0432, 0.0500, 0.0401, 0.0505], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:22:34,853 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 16:22:52,005 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:23:03,049 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.483e+02 2.226e+02 2.530e+02 3.220e+02 6.959e+02, threshold=5.059e+02, percent-clipped=1.0 2023-04-30 16:23:09,315 INFO [train.py:904] (5/8) Epoch 18, batch 2600, loss[loss=0.1778, simple_loss=0.2659, pruned_loss=0.04488, over 17174.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2626, pruned_loss=0.04402, over 3329571.44 frames. ], batch size: 46, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:17,710 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:24:19,536 INFO [train.py:904] (5/8) Epoch 18, batch 2650, loss[loss=0.1713, simple_loss=0.2549, pruned_loss=0.0438, over 16748.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2627, pruned_loss=0.04393, over 3324632.70 frames. ], batch size: 89, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:27,539 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:25:22,166 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.085e+02 2.405e+02 2.864e+02 4.837e+02, threshold=4.811e+02, percent-clipped=0.0 2023-04-30 16:25:27,546 INFO [train.py:904] (5/8) Epoch 18, batch 2700, loss[loss=0.1719, simple_loss=0.2581, pruned_loss=0.04285, over 16923.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2634, pruned_loss=0.04356, over 3332404.25 frames. ], batch size: 96, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:25:50,112 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175268.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:26:36,862 INFO [train.py:904] (5/8) Epoch 18, batch 2750, loss[loss=0.1623, simple_loss=0.257, pruned_loss=0.03375, over 16600.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2628, pruned_loss=0.04289, over 3322436.16 frames. ], batch size: 62, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:27:40,026 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.063e+02 2.386e+02 2.912e+02 4.096e+02, threshold=4.773e+02, percent-clipped=0.0 2023-04-30 16:27:45,205 INFO [train.py:904] (5/8) Epoch 18, batch 2800, loss[loss=0.1755, simple_loss=0.2519, pruned_loss=0.04959, over 16855.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2628, pruned_loss=0.04289, over 3323680.88 frames. ], batch size: 116, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:27:58,755 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:11,796 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:44,368 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9214, 4.7999, 4.7870, 4.4642, 4.4766, 4.8434, 4.6616, 4.5692], device='cuda:5'), covar=tensor([0.0588, 0.0703, 0.0273, 0.0288, 0.0856, 0.0484, 0.0429, 0.0618], device='cuda:5'), in_proj_covar=tensor([0.0294, 0.0416, 0.0344, 0.0336, 0.0359, 0.0389, 0.0238, 0.0415], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 16:28:45,648 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:56,413 INFO [train.py:904] (5/8) Epoch 18, batch 2850, loss[loss=0.1537, simple_loss=0.2327, pruned_loss=0.03734, over 16907.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2618, pruned_loss=0.04302, over 3322652.07 frames. ], batch size: 109, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:29:05,435 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175409.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:05,676 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0864, 2.2154, 2.6621, 3.0137, 2.8340, 3.5111, 2.2666, 3.4044], device='cuda:5'), covar=tensor([0.0219, 0.0428, 0.0285, 0.0308, 0.0287, 0.0172, 0.0473, 0.0164], device='cuda:5'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:29:35,293 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175431.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:58,778 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.293e+02 2.737e+02 3.416e+02 1.541e+03, threshold=5.475e+02, percent-clipped=7.0 2023-04-30 16:30:04,928 INFO [train.py:904] (5/8) Epoch 18, batch 2900, loss[loss=0.149, simple_loss=0.2447, pruned_loss=0.02666, over 17120.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2603, pruned_loss=0.0428, over 3325694.36 frames. ], batch size: 47, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:30:09,000 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175455.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:45,709 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:50,764 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7875, 2.5081, 1.9499, 2.3193, 2.8614, 2.6899, 2.9437, 2.9585], device='cuda:5'), covar=tensor([0.0174, 0.0310, 0.0495, 0.0376, 0.0199, 0.0270, 0.0200, 0.0250], device='cuda:5'), in_proj_covar=tensor([0.0202, 0.0235, 0.0225, 0.0225, 0.0236, 0.0234, 0.0240, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:31:04,476 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:31:13,725 INFO [train.py:904] (5/8) Epoch 18, batch 2950, loss[loss=0.1965, simple_loss=0.2832, pruned_loss=0.05494, over 16791.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2598, pruned_loss=0.04363, over 3321600.81 frames. ], batch size: 124, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:31:50,249 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2015, 1.5796, 2.0157, 2.1413, 2.2906, 2.3266, 1.7234, 2.3348], device='cuda:5'), covar=tensor([0.0218, 0.0453, 0.0253, 0.0283, 0.0258, 0.0244, 0.0461, 0.0145], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0192, 0.0179, 0.0185, 0.0192, 0.0150, 0.0195, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:32:09,676 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:32:17,359 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.280e+02 2.587e+02 3.034e+02 5.964e+02, threshold=5.174e+02, percent-clipped=1.0 2023-04-30 16:32:23,317 INFO [train.py:904] (5/8) Epoch 18, batch 3000, loss[loss=0.1817, simple_loss=0.2777, pruned_loss=0.0429, over 17032.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2611, pruned_loss=0.04473, over 3314140.94 frames. ], batch size: 50, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:32:23,317 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 16:32:32,128 INFO [train.py:938] (5/8) Epoch 18, validation: loss=0.1359, simple_loss=0.2415, pruned_loss=0.01516, over 944034.00 frames. 2023-04-30 16:32:32,129 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 16:32:48,742 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175563.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:33:12,080 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 16:33:42,222 INFO [train.py:904] (5/8) Epoch 18, batch 3050, loss[loss=0.1753, simple_loss=0.2724, pruned_loss=0.0391, over 17115.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2605, pruned_loss=0.04472, over 3318988.07 frames. ], batch size: 53, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:34:38,071 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9306, 2.5083, 1.9404, 2.3626, 2.8638, 2.6282, 3.0145, 3.0415], device='cuda:5'), covar=tensor([0.0203, 0.0375, 0.0546, 0.0411, 0.0247, 0.0363, 0.0235, 0.0246], device='cuda:5'), in_proj_covar=tensor([0.0201, 0.0234, 0.0224, 0.0223, 0.0235, 0.0233, 0.0238, 0.0226], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:34:46,186 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.265e+02 2.697e+02 3.394e+02 4.892e+02, threshold=5.395e+02, percent-clipped=0.0 2023-04-30 16:34:51,894 INFO [train.py:904] (5/8) Epoch 18, batch 3100, loss[loss=0.1711, simple_loss=0.2469, pruned_loss=0.0477, over 16916.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.26, pruned_loss=0.04477, over 3308137.77 frames. ], batch size: 109, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:01,378 INFO [train.py:904] (5/8) Epoch 18, batch 3150, loss[loss=0.1683, simple_loss=0.257, pruned_loss=0.03979, over 17258.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.259, pruned_loss=0.04426, over 3309568.98 frames. ], batch size: 45, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:02,184 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 16:36:13,042 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175710.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:36:36,354 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:37:05,953 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.210e+02 2.531e+02 3.035e+02 7.463e+02, threshold=5.061e+02, percent-clipped=4.0 2023-04-30 16:37:06,352 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8944, 4.8470, 4.7572, 4.4650, 4.4078, 4.8479, 4.6567, 4.5379], device='cuda:5'), covar=tensor([0.0751, 0.0723, 0.0354, 0.0339, 0.1066, 0.0570, 0.0494, 0.0787], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0423, 0.0349, 0.0341, 0.0365, 0.0396, 0.0241, 0.0421], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 16:37:09,258 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175750.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:37:11,235 INFO [train.py:904] (5/8) Epoch 18, batch 3200, loss[loss=0.153, simple_loss=0.2491, pruned_loss=0.0284, over 17255.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2588, pruned_loss=0.04426, over 3307989.24 frames. ], batch size: 45, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:37:17,450 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5263, 3.5972, 3.2926, 3.0017, 3.1451, 3.5095, 3.2737, 3.2832], device='cuda:5'), covar=tensor([0.0653, 0.0556, 0.0322, 0.0276, 0.0590, 0.0452, 0.1330, 0.0520], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0422, 0.0349, 0.0341, 0.0365, 0.0395, 0.0241, 0.0421], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 16:37:37,871 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175771.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:37:38,993 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9343, 4.0727, 2.7073, 4.6645, 3.2040, 4.6411, 2.7870, 3.2927], device='cuda:5'), covar=tensor([0.0278, 0.0383, 0.1388, 0.0343, 0.0772, 0.0471, 0.1343, 0.0695], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0178, 0.0196, 0.0163, 0.0177, 0.0222, 0.0205, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:37:56,525 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0552, 4.4798, 4.4311, 3.2518, 3.7212, 4.3980, 3.9653, 2.6026], device='cuda:5'), covar=tensor([0.0407, 0.0061, 0.0045, 0.0320, 0.0129, 0.0082, 0.0091, 0.0424], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0080, 0.0080, 0.0132, 0.0094, 0.0104, 0.0091, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 16:38:11,032 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:38:13,772 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7439, 2.8757, 2.8198, 4.9753, 4.0100, 4.4717, 1.7643, 3.0938], device='cuda:5'), covar=tensor([0.1412, 0.0761, 0.1113, 0.0248, 0.0233, 0.0380, 0.1548, 0.0807], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0169, 0.0189, 0.0180, 0.0203, 0.0216, 0.0193, 0.0188], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:38:20,679 INFO [train.py:904] (5/8) Epoch 18, batch 3250, loss[loss=0.1733, simple_loss=0.2558, pruned_loss=0.0454, over 16739.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2586, pruned_loss=0.04408, over 3311609.28 frames. ], batch size: 89, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:39:10,114 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:17,731 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175843.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:23,830 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.233e+02 2.626e+02 2.962e+02 5.772e+02, threshold=5.253e+02, percent-clipped=1.0 2023-04-30 16:39:29,874 INFO [train.py:904] (5/8) Epoch 18, batch 3300, loss[loss=0.2007, simple_loss=0.2697, pruned_loss=0.06585, over 16857.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2593, pruned_loss=0.0441, over 3320320.93 frames. ], batch size: 116, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:39:31,553 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1997, 3.9465, 4.4057, 2.2064, 4.6372, 4.6367, 3.3738, 3.5202], device='cuda:5'), covar=tensor([0.0638, 0.0233, 0.0191, 0.1151, 0.0069, 0.0151, 0.0380, 0.0380], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0137, 0.0077, 0.0122, 0.0125, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 16:39:45,171 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:59,580 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3669, 4.4402, 4.8223, 4.7802, 4.7950, 4.4661, 4.5068, 4.3050], device='cuda:5'), covar=tensor([0.0380, 0.0635, 0.0443, 0.0433, 0.0514, 0.0425, 0.0849, 0.0609], device='cuda:5'), in_proj_covar=tensor([0.0406, 0.0441, 0.0431, 0.0404, 0.0477, 0.0453, 0.0552, 0.0359], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 16:40:10,965 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:40:26,494 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8036, 4.3485, 4.4206, 3.1852, 3.7657, 4.3793, 3.9090, 2.5740], device='cuda:5'), covar=tensor([0.0461, 0.0057, 0.0037, 0.0343, 0.0101, 0.0086, 0.0083, 0.0421], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0080, 0.0080, 0.0133, 0.0094, 0.0105, 0.0092, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 16:40:38,828 INFO [train.py:904] (5/8) Epoch 18, batch 3350, loss[loss=0.1872, simple_loss=0.2681, pruned_loss=0.05315, over 16455.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2598, pruned_loss=0.04387, over 3319872.95 frames. ], batch size: 75, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:40:51,562 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:34,646 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:38,200 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:41,869 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.233e+02 2.642e+02 3.041e+02 9.151e+02, threshold=5.284e+02, percent-clipped=6.0 2023-04-30 16:41:47,209 INFO [train.py:904] (5/8) Epoch 18, batch 3400, loss[loss=0.1825, simple_loss=0.2543, pruned_loss=0.05537, over 16904.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2593, pruned_loss=0.04365, over 3323787.39 frames. ], batch size: 109, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:43:01,870 INFO [train.py:904] (5/8) Epoch 18, batch 3450, loss[loss=0.1786, simple_loss=0.2506, pruned_loss=0.0533, over 16404.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2574, pruned_loss=0.04305, over 3326915.15 frames. ], batch size: 146, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:43:07,960 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176006.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:43:35,495 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:06,115 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.158e+02 2.512e+02 2.983e+02 6.813e+02, threshold=5.023e+02, percent-clipped=2.0 2023-04-30 16:44:09,531 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:11,310 INFO [train.py:904] (5/8) Epoch 18, batch 3500, loss[loss=0.1494, simple_loss=0.2377, pruned_loss=0.03057, over 16801.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2568, pruned_loss=0.04305, over 3316593.16 frames. ], batch size: 39, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:44:31,796 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176066.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:41,808 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176074.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:03,179 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8072, 2.0001, 2.4129, 2.8641, 2.6479, 3.3382, 2.2571, 3.3208], device='cuda:5'), covar=tensor([0.0238, 0.0442, 0.0309, 0.0281, 0.0278, 0.0150, 0.0421, 0.0145], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0189, 0.0177, 0.0181, 0.0188, 0.0149, 0.0192, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:45:15,800 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176098.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:20,522 INFO [train.py:904] (5/8) Epoch 18, batch 3550, loss[loss=0.1474, simple_loss=0.2394, pruned_loss=0.02768, over 17224.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2565, pruned_loss=0.04244, over 3315459.15 frames. ], batch size: 45, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:46:10,652 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:46:26,486 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.026e+02 2.447e+02 2.938e+02 5.668e+02, threshold=4.895e+02, percent-clipped=1.0 2023-04-30 16:46:32,322 INFO [train.py:904] (5/8) Epoch 18, batch 3600, loss[loss=0.1802, simple_loss=0.2493, pruned_loss=0.05554, over 16781.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2547, pruned_loss=0.04235, over 3308814.63 frames. ], batch size: 124, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:47:02,337 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 16:47:18,726 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:47:37,737 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 16:47:43,325 INFO [train.py:904] (5/8) Epoch 18, batch 3650, loss[loss=0.1751, simple_loss=0.2435, pruned_loss=0.05332, over 16878.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2532, pruned_loss=0.04279, over 3313243.74 frames. ], batch size: 109, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:48:30,694 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8489, 2.8216, 2.6410, 4.2593, 3.6896, 4.2090, 1.6624, 2.9802], device='cuda:5'), covar=tensor([0.1326, 0.0676, 0.1091, 0.0149, 0.0140, 0.0329, 0.1487, 0.0800], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0169, 0.0189, 0.0182, 0.0203, 0.0215, 0.0194, 0.0188], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:48:35,860 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176237.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:43,990 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:47,938 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6354, 3.6892, 2.0499, 3.9010, 2.8676, 3.8436, 2.1522, 2.8740], device='cuda:5'), covar=tensor([0.0221, 0.0288, 0.1532, 0.0270, 0.0690, 0.0664, 0.1405, 0.0657], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0178, 0.0195, 0.0162, 0.0176, 0.0221, 0.0204, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:48:53,039 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.298e+02 2.711e+02 3.292e+02 6.434e+02, threshold=5.422e+02, percent-clipped=3.0 2023-04-30 16:48:58,535 INFO [train.py:904] (5/8) Epoch 18, batch 3700, loss[loss=0.1636, simple_loss=0.2397, pruned_loss=0.04374, over 16715.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2519, pruned_loss=0.04422, over 3307085.62 frames. ], batch size: 134, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:49:07,005 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:10,773 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176301.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:11,692 INFO [train.py:904] (5/8) Epoch 18, batch 3750, loss[loss=0.1662, simple_loss=0.2426, pruned_loss=0.04488, over 16226.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2521, pruned_loss=0.04569, over 3301734.52 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:50:13,721 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:35,152 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:51:17,342 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.173e+02 2.603e+02 3.284e+02 5.784e+02, threshold=5.206e+02, percent-clipped=1.0 2023-04-30 16:51:23,249 INFO [train.py:904] (5/8) Epoch 18, batch 3800, loss[loss=0.1778, simple_loss=0.2567, pruned_loss=0.04947, over 16375.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2538, pruned_loss=0.04713, over 3301889.71 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:51:44,185 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:51:58,885 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1088, 4.0194, 4.1771, 4.3193, 4.3917, 3.9981, 4.1820, 4.3970], device='cuda:5'), covar=tensor([0.1549, 0.0984, 0.1210, 0.0635, 0.0543, 0.1314, 0.2007, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0636, 0.0790, 0.0928, 0.0803, 0.0590, 0.0634, 0.0649, 0.0753], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:52:35,507 INFO [train.py:904] (5/8) Epoch 18, batch 3850, loss[loss=0.1675, simple_loss=0.2394, pruned_loss=0.04782, over 16901.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2545, pruned_loss=0.04792, over 3291392.20 frames. ], batch size: 116, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:52:53,511 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176414.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:53:42,934 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 2.342e+02 2.782e+02 3.184e+02 4.336e+02, threshold=5.564e+02, percent-clipped=0.0 2023-04-30 16:53:48,961 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9997, 3.0922, 3.2479, 2.0256, 2.8012, 2.2868, 3.5072, 3.4947], device='cuda:5'), covar=tensor([0.0231, 0.0845, 0.0603, 0.1839, 0.0854, 0.0996, 0.0515, 0.0762], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0161, 0.0165, 0.0151, 0.0142, 0.0127, 0.0142, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:53:49,451 INFO [train.py:904] (5/8) Epoch 18, batch 3900, loss[loss=0.1703, simple_loss=0.2472, pruned_loss=0.04671, over 16795.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2549, pruned_loss=0.04856, over 3283707.78 frames. ], batch size: 124, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:54:11,193 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:55:00,910 INFO [train.py:904] (5/8) Epoch 18, batch 3950, loss[loss=0.181, simple_loss=0.2659, pruned_loss=0.04808, over 17064.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2545, pruned_loss=0.04868, over 3280713.12 frames. ], batch size: 55, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:55:35,348 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176527.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:55:50,477 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176537.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:56:05,767 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.227e+02 2.521e+02 3.093e+02 7.043e+02, threshold=5.041e+02, percent-clipped=2.0 2023-04-30 16:56:12,332 INFO [train.py:904] (5/8) Epoch 18, batch 4000, loss[loss=0.1715, simple_loss=0.2545, pruned_loss=0.0442, over 16311.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2551, pruned_loss=0.04914, over 3279783.21 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:56:23,234 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3136, 3.3972, 3.6227, 2.0854, 2.9866, 2.4252, 3.6343, 3.7633], device='cuda:5'), covar=tensor([0.0197, 0.0762, 0.0594, 0.1992, 0.0904, 0.0919, 0.0557, 0.0837], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0161, 0.0165, 0.0151, 0.0143, 0.0127, 0.0142, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:57:00,439 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:18,457 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:23,564 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176601.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:24,380 INFO [train.py:904] (5/8) Epoch 18, batch 4050, loss[loss=0.2041, simple_loss=0.277, pruned_loss=0.06565, over 12449.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2564, pruned_loss=0.04878, over 3279242.47 frames. ], batch size: 247, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:57:39,766 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:58,533 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:13,037 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.8460, 6.1603, 5.8388, 6.0139, 5.6252, 5.4405, 5.5885, 6.2631], device='cuda:5'), covar=tensor([0.1155, 0.0765, 0.1012, 0.0718, 0.0779, 0.0606, 0.0955, 0.0833], device='cuda:5'), in_proj_covar=tensor([0.0650, 0.0803, 0.0651, 0.0599, 0.0503, 0.0512, 0.0668, 0.0622], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:58:30,485 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.366e+02 1.835e+02 2.096e+02 2.511e+02 3.496e+02, threshold=4.192e+02, percent-clipped=0.0 2023-04-30 16:58:33,594 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176649.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:36,787 INFO [train.py:904] (5/8) Epoch 18, batch 4100, loss[loss=0.1761, simple_loss=0.2653, pruned_loss=0.04348, over 16631.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2574, pruned_loss=0.04809, over 3275850.18 frames. ], batch size: 62, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:58:57,268 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9883, 4.9855, 4.7850, 4.1889, 4.9196, 1.8193, 4.6565, 4.4881], device='cuda:5'), covar=tensor([0.0061, 0.0050, 0.0144, 0.0277, 0.0057, 0.2688, 0.0090, 0.0187], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0148, 0.0196, 0.0178, 0.0170, 0.0204, 0.0186, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 16:58:57,294 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:59,708 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7670, 2.9303, 2.5862, 4.5181, 3.5849, 4.0384, 1.7777, 2.8693], device='cuda:5'), covar=tensor([0.1320, 0.0741, 0.1237, 0.0176, 0.0361, 0.0430, 0.1504, 0.0899], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0182, 0.0205, 0.0215, 0.0195, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 16:59:28,843 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:59:51,172 INFO [train.py:904] (5/8) Epoch 18, batch 4150, loss[loss=0.2531, simple_loss=0.3191, pruned_loss=0.09355, over 11648.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2652, pruned_loss=0.05065, over 3254883.70 frames. ], batch size: 247, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:00:27,867 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:01:00,612 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.334e+02 2.630e+02 3.373e+02 5.770e+02, threshold=5.259e+02, percent-clipped=9.0 2023-04-30 17:01:06,258 INFO [train.py:904] (5/8) Epoch 18, batch 4200, loss[loss=0.2122, simple_loss=0.2996, pruned_loss=0.06234, over 16901.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2721, pruned_loss=0.05234, over 3226280.86 frames. ], batch size: 109, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:01:06,875 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9926, 3.1703, 3.3294, 2.0599, 2.8713, 2.1428, 3.3978, 3.4026], device='cuda:5'), covar=tensor([0.0279, 0.0934, 0.0603, 0.1973, 0.0910, 0.1031, 0.0672, 0.1036], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0160, 0.0164, 0.0150, 0.0142, 0.0126, 0.0142, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:02:21,469 INFO [train.py:904] (5/8) Epoch 18, batch 4250, loss[loss=0.1769, simple_loss=0.2733, pruned_loss=0.04026, over 16195.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2753, pruned_loss=0.05189, over 3221755.48 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:02:51,847 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:03:16,624 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5015, 3.5330, 2.0245, 4.0901, 2.7228, 4.0722, 2.3323, 2.8207], device='cuda:5'), covar=tensor([0.0281, 0.0368, 0.1803, 0.0156, 0.0883, 0.0431, 0.1524, 0.0805], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0175, 0.0192, 0.0156, 0.0174, 0.0216, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:03:30,838 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.197e+02 2.694e+02 3.150e+02 7.622e+02, threshold=5.388e+02, percent-clipped=1.0 2023-04-30 17:03:35,717 INFO [train.py:904] (5/8) Epoch 18, batch 4300, loss[loss=0.1995, simple_loss=0.2889, pruned_loss=0.05508, over 11707.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2762, pruned_loss=0.05097, over 3208329.39 frames. ], batch size: 246, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:03:38,887 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:44,718 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176898.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:49,713 INFO [train.py:904] (5/8) Epoch 18, batch 4350, loss[loss=0.1812, simple_loss=0.2753, pruned_loss=0.0436, over 16715.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2793, pruned_loss=0.05201, over 3207002.37 frames. ], batch size: 83, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:04:56,563 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8334, 5.0384, 5.2905, 5.0148, 5.1503, 5.6923, 5.1408, 4.7981], device='cuda:5'), covar=tensor([0.0876, 0.1650, 0.1595, 0.1701, 0.2116, 0.0809, 0.1367, 0.2115], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0572, 0.0624, 0.0478, 0.0642, 0.0654, 0.0497, 0.0640], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 17:05:06,480 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176913.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:05:07,735 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176914.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:05:55,204 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:05:57,200 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 2.251e+02 2.629e+02 3.304e+02 5.605e+02, threshold=5.258e+02, percent-clipped=2.0 2023-04-30 17:06:01,336 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5001, 4.4823, 4.3057, 2.7036, 3.8063, 4.2680, 3.7032, 2.5236], device='cuda:5'), covar=tensor([0.0463, 0.0021, 0.0038, 0.0370, 0.0086, 0.0073, 0.0086, 0.0364], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0078, 0.0079, 0.0132, 0.0094, 0.0104, 0.0091, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 17:06:03,307 INFO [train.py:904] (5/8) Epoch 18, batch 4400, loss[loss=0.1771, simple_loss=0.2639, pruned_loss=0.04513, over 17224.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2814, pruned_loss=0.05309, over 3203088.19 frames. ], batch size: 52, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:06:17,431 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:06:45,591 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:07:16,219 INFO [train.py:904] (5/8) Epoch 18, batch 4450, loss[loss=0.2116, simple_loss=0.303, pruned_loss=0.06013, over 16850.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2847, pruned_loss=0.05441, over 3208011.78 frames. ], batch size: 83, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:07:29,069 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7650, 3.9565, 3.1150, 2.4673, 2.7475, 2.5044, 4.3493, 3.5805], device='cuda:5'), covar=tensor([0.2710, 0.0639, 0.1588, 0.2301, 0.2383, 0.1881, 0.0375, 0.1076], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0266, 0.0301, 0.0304, 0.0294, 0.0248, 0.0288, 0.0329], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 17:07:44,708 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:08:10,472 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2271, 5.5385, 5.2856, 5.3265, 5.0290, 4.7781, 4.9782, 5.6481], device='cuda:5'), covar=tensor([0.1090, 0.0773, 0.0946, 0.0748, 0.0789, 0.0822, 0.1079, 0.0793], device='cuda:5'), in_proj_covar=tensor([0.0630, 0.0777, 0.0634, 0.0580, 0.0489, 0.0498, 0.0646, 0.0603], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 17:08:24,608 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.969e+02 2.287e+02 2.678e+02 4.361e+02, threshold=4.575e+02, percent-clipped=0.0 2023-04-30 17:08:30,160 INFO [train.py:904] (5/8) Epoch 18, batch 4500, loss[loss=0.1938, simple_loss=0.282, pruned_loss=0.0528, over 16468.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2856, pruned_loss=0.05509, over 3218659.69 frames. ], batch size: 75, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:43,613 INFO [train.py:904] (5/8) Epoch 18, batch 4550, loss[loss=0.2209, simple_loss=0.294, pruned_loss=0.07395, over 12152.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2868, pruned_loss=0.05621, over 3219149.33 frames. ], batch size: 247, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:46,500 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 17:09:52,941 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177109.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:09:55,563 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:11,745 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177122.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:22,057 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 17:10:25,199 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-30 17:10:48,537 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 1.819e+02 2.100e+02 2.524e+02 4.531e+02, threshold=4.200e+02, percent-clipped=0.0 2023-04-30 17:10:53,842 INFO [train.py:904] (5/8) Epoch 18, batch 4600, loss[loss=0.1973, simple_loss=0.2889, pruned_loss=0.05282, over 16726.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2878, pruned_loss=0.05667, over 3201953.91 frames. ], batch size: 124, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:11:16,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7850, 3.7339, 2.3926, 4.6755, 3.0476, 4.5331, 2.4441, 3.0974], device='cuda:5'), covar=tensor([0.0266, 0.0364, 0.1565, 0.0106, 0.0787, 0.0366, 0.1470, 0.0731], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0174, 0.0192, 0.0153, 0.0173, 0.0214, 0.0200, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:11:19,983 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:20,138 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:22,926 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1379, 3.5365, 3.3668, 5.2019, 4.2346, 4.5000, 2.2370, 3.2635], device='cuda:5'), covar=tensor([0.1157, 0.0587, 0.0880, 0.0102, 0.0381, 0.0343, 0.1332, 0.0776], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0171, 0.0192, 0.0182, 0.0206, 0.0216, 0.0197, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:11:24,076 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:12:05,786 INFO [train.py:904] (5/8) Epoch 18, batch 4650, loss[loss=0.1915, simple_loss=0.2747, pruned_loss=0.05417, over 17042.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2861, pruned_loss=0.0564, over 3204038.59 frames. ], batch size: 41, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:12:16,103 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177209.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:13:10,934 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 1.973e+02 2.204e+02 2.634e+02 3.745e+02, threshold=4.408e+02, percent-clipped=0.0 2023-04-30 17:13:16,227 INFO [train.py:904] (5/8) Epoch 18, batch 4700, loss[loss=0.1972, simple_loss=0.2772, pruned_loss=0.05861, over 17044.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2839, pruned_loss=0.0553, over 3216894.45 frames. ], batch size: 55, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:13:59,449 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:14:17,177 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8048, 3.7355, 3.8724, 3.9923, 4.0876, 3.6855, 4.0110, 4.1157], device='cuda:5'), covar=tensor([0.1515, 0.1027, 0.1278, 0.0662, 0.0530, 0.1845, 0.0841, 0.0578], device='cuda:5'), in_proj_covar=tensor([0.0603, 0.0744, 0.0875, 0.0763, 0.0560, 0.0601, 0.0613, 0.0709], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 17:14:28,063 INFO [train.py:904] (5/8) Epoch 18, batch 4750, loss[loss=0.167, simple_loss=0.2514, pruned_loss=0.04132, over 17222.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2788, pruned_loss=0.05268, over 3233448.88 frames. ], batch size: 44, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:14:56,840 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177321.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:15:07,310 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:15:27,412 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0810, 2.4292, 2.5909, 1.9041, 2.7692, 2.8098, 2.4544, 2.3537], device='cuda:5'), covar=tensor([0.0702, 0.0223, 0.0193, 0.0929, 0.0093, 0.0216, 0.0423, 0.0419], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0107, 0.0095, 0.0139, 0.0076, 0.0123, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:15:34,637 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.255e+02 1.837e+02 2.097e+02 2.577e+02 4.933e+02, threshold=4.195e+02, percent-clipped=1.0 2023-04-30 17:15:40,564 INFO [train.py:904] (5/8) Epoch 18, batch 4800, loss[loss=0.2263, simple_loss=0.2982, pruned_loss=0.07715, over 12148.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2759, pruned_loss=0.05099, over 3221812.97 frames. ], batch size: 248, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:15:47,115 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-30 17:16:07,489 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177369.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:16:57,033 INFO [train.py:904] (5/8) Epoch 18, batch 4850, loss[loss=0.2014, simple_loss=0.2855, pruned_loss=0.05866, over 12087.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2771, pruned_loss=0.05045, over 3191917.15 frames. ], batch size: 246, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:17:17,189 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 17:17:26,324 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9171, 3.2422, 3.4047, 1.5328, 3.5953, 3.6766, 2.9467, 2.5129], device='cuda:5'), covar=tensor([0.1208, 0.0187, 0.0158, 0.1426, 0.0081, 0.0139, 0.0358, 0.0661], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0107, 0.0095, 0.0139, 0.0076, 0.0122, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:18:05,882 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.952e+02 2.254e+02 2.679e+02 4.935e+02, threshold=4.508e+02, percent-clipped=3.0 2023-04-30 17:18:11,218 INFO [train.py:904] (5/8) Epoch 18, batch 4900, loss[loss=0.1716, simple_loss=0.2632, pruned_loss=0.04001, over 16747.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2764, pruned_loss=0.04948, over 3173990.18 frames. ], batch size: 124, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:18:30,806 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177465.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:18:33,659 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177467.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:19:07,843 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 17:19:25,471 INFO [train.py:904] (5/8) Epoch 18, batch 4950, loss[loss=0.1887, simple_loss=0.2815, pruned_loss=0.04794, over 17134.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.276, pruned_loss=0.04882, over 3188445.63 frames. ], batch size: 47, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:19:34,709 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177509.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:20:14,915 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1530, 4.9886, 5.1947, 5.3801, 5.5811, 4.9103, 5.5542, 5.5289], device='cuda:5'), covar=tensor([0.1663, 0.1256, 0.1671, 0.0763, 0.0561, 0.0778, 0.0500, 0.0567], device='cuda:5'), in_proj_covar=tensor([0.0600, 0.0743, 0.0872, 0.0761, 0.0558, 0.0598, 0.0611, 0.0706], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 17:20:30,520 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.064e+02 2.400e+02 2.988e+02 6.295e+02, threshold=4.800e+02, percent-clipped=2.0 2023-04-30 17:20:35,825 INFO [train.py:904] (5/8) Epoch 18, batch 5000, loss[loss=0.2035, simple_loss=0.2943, pruned_loss=0.0564, over 15406.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2778, pruned_loss=0.04924, over 3193775.86 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:20:37,499 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177553.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:20:41,868 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:21:47,168 INFO [train.py:904] (5/8) Epoch 18, batch 5050, loss[loss=0.1873, simple_loss=0.2791, pruned_loss=0.04772, over 15289.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2782, pruned_loss=0.04917, over 3200521.20 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:21:49,889 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8456, 5.0933, 5.3112, 5.0675, 5.1188, 5.7109, 5.1844, 4.8981], device='cuda:5'), covar=tensor([0.0958, 0.1728, 0.1763, 0.1777, 0.2347, 0.0821, 0.1269, 0.2198], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0562, 0.0615, 0.0474, 0.0633, 0.0645, 0.0485, 0.0635], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 17:22:04,646 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:22:54,214 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.030e+02 2.394e+02 2.756e+02 4.573e+02, threshold=4.789e+02, percent-clipped=0.0 2023-04-30 17:22:58,443 INFO [train.py:904] (5/8) Epoch 18, batch 5100, loss[loss=0.1727, simple_loss=0.25, pruned_loss=0.04767, over 16610.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2759, pruned_loss=0.0482, over 3212100.78 frames. ], batch size: 62, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:23:16,223 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177664.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:23:40,943 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7723, 4.8010, 4.6230, 4.1966, 4.2394, 4.6704, 4.5472, 4.4318], device='cuda:5'), covar=tensor([0.0567, 0.0358, 0.0319, 0.0331, 0.1117, 0.0432, 0.0417, 0.0681], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0395, 0.0329, 0.0319, 0.0341, 0.0371, 0.0224, 0.0390], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 17:24:05,191 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 17:24:10,484 INFO [train.py:904] (5/8) Epoch 18, batch 5150, loss[loss=0.1919, simple_loss=0.2905, pruned_loss=0.04663, over 16784.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2755, pruned_loss=0.04735, over 3194416.68 frames. ], batch size: 83, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:24:38,721 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0353, 5.1238, 5.4912, 5.4489, 5.4396, 5.0855, 5.0230, 4.7539], device='cuda:5'), covar=tensor([0.0264, 0.0384, 0.0271, 0.0296, 0.0448, 0.0316, 0.0941, 0.0414], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0413, 0.0407, 0.0380, 0.0453, 0.0427, 0.0522, 0.0340], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 17:24:46,062 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:04,686 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2342, 5.2844, 5.6914, 5.6586, 5.6770, 5.2910, 5.2039, 4.9583], device='cuda:5'), covar=tensor([0.0335, 0.0535, 0.0319, 0.0389, 0.0516, 0.0375, 0.1048, 0.0433], device='cuda:5'), in_proj_covar=tensor([0.0380, 0.0413, 0.0408, 0.0381, 0.0453, 0.0427, 0.0522, 0.0341], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 17:25:20,508 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 1.888e+02 2.241e+02 2.680e+02 5.710e+02, threshold=4.482e+02, percent-clipped=2.0 2023-04-30 17:25:24,680 INFO [train.py:904] (5/8) Epoch 18, batch 5200, loss[loss=0.1973, simple_loss=0.2693, pruned_loss=0.06264, over 17011.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2744, pruned_loss=0.04708, over 3189820.79 frames. ], batch size: 41, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:25:44,170 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:47,821 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:06,694 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:07,081 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 17:26:22,032 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 17:26:31,603 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177797.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:39,176 INFO [train.py:904] (5/8) Epoch 18, batch 5250, loss[loss=0.1881, simple_loss=0.2721, pruned_loss=0.05206, over 12086.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2723, pruned_loss=0.04662, over 3180910.90 frames. ], batch size: 246, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:26:55,361 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:57,994 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5540, 3.6421, 3.4547, 3.1713, 3.2309, 3.5598, 3.2770, 3.3667], device='cuda:5'), covar=tensor([0.0596, 0.0571, 0.0334, 0.0298, 0.0699, 0.0495, 0.1652, 0.0508], device='cuda:5'), in_proj_covar=tensor([0.0280, 0.0395, 0.0329, 0.0320, 0.0340, 0.0370, 0.0224, 0.0389], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 17:26:58,862 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:27:37,106 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8402, 3.7957, 3.9474, 3.7053, 3.9222, 4.2935, 3.9898, 3.6699], device='cuda:5'), covar=tensor([0.1998, 0.2472, 0.2060, 0.2624, 0.2669, 0.1558, 0.1475, 0.2627], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0561, 0.0614, 0.0473, 0.0632, 0.0643, 0.0485, 0.0633], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 17:27:37,235 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177841.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:27:48,912 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 1.963e+02 2.196e+02 2.536e+02 4.359e+02, threshold=4.392e+02, percent-clipped=1.0 2023-04-30 17:27:50,022 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4783, 3.4715, 2.0721, 3.9175, 2.6513, 3.8846, 2.2800, 2.7996], device='cuda:5'), covar=tensor([0.0283, 0.0356, 0.1728, 0.0164, 0.0833, 0.0484, 0.1545, 0.0728], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0172, 0.0191, 0.0151, 0.0173, 0.0211, 0.0199, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:27:52,860 INFO [train.py:904] (5/8) Epoch 18, batch 5300, loss[loss=0.1967, simple_loss=0.2822, pruned_loss=0.05559, over 12119.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2695, pruned_loss=0.0456, over 3184788.21 frames. ], batch size: 247, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:28:02,138 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177858.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:29:05,791 INFO [train.py:904] (5/8) Epoch 18, batch 5350, loss[loss=0.1723, simple_loss=0.2612, pruned_loss=0.04167, over 17139.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.267, pruned_loss=0.04479, over 3185431.22 frames. ], batch size: 46, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:29:17,498 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:30:15,242 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 1.886e+02 2.232e+02 2.628e+02 4.605e+02, threshold=4.463e+02, percent-clipped=1.0 2023-04-30 17:30:19,772 INFO [train.py:904] (5/8) Epoch 18, batch 5400, loss[loss=0.1958, simple_loss=0.2801, pruned_loss=0.05569, over 16661.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2693, pruned_loss=0.04539, over 3200604.60 frames. ], batch size: 62, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:30:21,441 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6507, 2.4571, 2.4270, 3.5087, 2.3619, 3.7055, 1.5599, 2.7427], device='cuda:5'), covar=tensor([0.1373, 0.0770, 0.1200, 0.0144, 0.0161, 0.0387, 0.1615, 0.0819], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0170, 0.0191, 0.0180, 0.0204, 0.0214, 0.0196, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:31:39,767 INFO [train.py:904] (5/8) Epoch 18, batch 5450, loss[loss=0.2174, simple_loss=0.3017, pruned_loss=0.06654, over 15446.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2723, pruned_loss=0.04695, over 3205124.52 frames. ], batch size: 191, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:31:52,994 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8055, 2.9685, 2.6038, 4.6515, 3.3984, 4.1547, 1.6029, 3.0921], device='cuda:5'), covar=tensor([0.1408, 0.0728, 0.1267, 0.0161, 0.0331, 0.0433, 0.1705, 0.0814], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0169, 0.0191, 0.0179, 0.0203, 0.0213, 0.0195, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:31:55,215 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 17:32:08,843 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178020.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:32:52,432 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.290e+02 2.936e+02 3.716e+02 7.248e+02, threshold=5.872e+02, percent-clipped=9.0 2023-04-30 17:32:56,918 INFO [train.py:904] (5/8) Epoch 18, batch 5500, loss[loss=0.2564, simple_loss=0.3247, pruned_loss=0.09404, over 11953.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2794, pruned_loss=0.05113, over 3182428.28 frames. ], batch size: 247, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:33:23,883 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 17:34:14,387 INFO [train.py:904] (5/8) Epoch 18, batch 5550, loss[loss=0.2677, simple_loss=0.3324, pruned_loss=0.1015, over 10998.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.287, pruned_loss=0.05701, over 3131437.97 frames. ], batch size: 246, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:05,331 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-04-30 17:35:09,862 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:35:31,151 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.108e+02 3.250e+02 3.857e+02 5.113e+02 1.112e+03, threshold=7.715e+02, percent-clipped=10.0 2023-04-30 17:35:34,172 INFO [train.py:904] (5/8) Epoch 18, batch 5600, loss[loss=0.2054, simple_loss=0.2898, pruned_loss=0.06052, over 16519.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2919, pruned_loss=0.06114, over 3110038.96 frames. ], batch size: 62, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:36,608 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:36:18,296 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1768, 3.0704, 3.3074, 1.6877, 3.4900, 3.4895, 2.8237, 2.6141], device='cuda:5'), covar=tensor([0.0815, 0.0244, 0.0180, 0.1206, 0.0076, 0.0202, 0.0411, 0.0475], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0105, 0.0094, 0.0137, 0.0076, 0.0121, 0.0124, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 17:36:58,596 INFO [train.py:904] (5/8) Epoch 18, batch 5650, loss[loss=0.2334, simple_loss=0.3099, pruned_loss=0.07845, over 15419.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2969, pruned_loss=0.06465, over 3094189.77 frames. ], batch size: 190, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:37:09,862 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178209.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:38:16,451 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 3.197e+02 3.818e+02 4.692e+02 7.543e+02, threshold=7.636e+02, percent-clipped=0.0 2023-04-30 17:38:17,803 INFO [train.py:904] (5/8) Epoch 18, batch 5700, loss[loss=0.2129, simple_loss=0.3099, pruned_loss=0.05795, over 16433.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2983, pruned_loss=0.06581, over 3086549.82 frames. ], batch size: 35, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:38:20,143 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8960, 2.7033, 2.6339, 1.9502, 2.5427, 2.6873, 2.5291, 1.9108], device='cuda:5'), covar=tensor([0.0408, 0.0080, 0.0097, 0.0355, 0.0135, 0.0121, 0.0124, 0.0378], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0130, 0.0093, 0.0103, 0.0090, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 17:38:25,293 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:39:39,093 INFO [train.py:904] (5/8) Epoch 18, batch 5750, loss[loss=0.1975, simple_loss=0.2851, pruned_loss=0.05499, over 17022.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3014, pruned_loss=0.068, over 3048193.39 frames. ], batch size: 55, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:40:08,760 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:40:57,701 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 17:40:59,320 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 3.189e+02 3.915e+02 4.745e+02 1.114e+03, threshold=7.831e+02, percent-clipped=3.0 2023-04-30 17:41:00,704 INFO [train.py:904] (5/8) Epoch 18, batch 5800, loss[loss=0.2024, simple_loss=0.2926, pruned_loss=0.05607, over 15322.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.3005, pruned_loss=0.06661, over 3049201.86 frames. ], batch size: 190, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:41:28,156 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:42:19,756 INFO [train.py:904] (5/8) Epoch 18, batch 5850, loss[loss=0.2347, simple_loss=0.3152, pruned_loss=0.07711, over 11337.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2981, pruned_loss=0.06451, over 3073292.06 frames. ], batch size: 246, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:16,272 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:43:39,874 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.654e+02 3.099e+02 3.804e+02 9.220e+02, threshold=6.197e+02, percent-clipped=1.0 2023-04-30 17:43:41,476 INFO [train.py:904] (5/8) Epoch 18, batch 5900, loss[loss=0.1978, simple_loss=0.2869, pruned_loss=0.05431, over 16613.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2977, pruned_loss=0.06435, over 3073413.48 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:42,229 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:43:44,114 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178453.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:32,740 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178484.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:51,628 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3660, 3.8085, 3.9367, 2.6312, 3.6116, 3.9651, 3.6888, 2.1548], device='cuda:5'), covar=tensor([0.0510, 0.0062, 0.0049, 0.0380, 0.0086, 0.0099, 0.0074, 0.0448], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0079, 0.0080, 0.0132, 0.0094, 0.0105, 0.0091, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 17:44:58,685 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:59,560 INFO [train.py:904] (5/8) Epoch 18, batch 5950, loss[loss=0.2047, simple_loss=0.2966, pruned_loss=0.05635, over 16441.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2982, pruned_loss=0.06313, over 3084995.86 frames. ], batch size: 146, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:45:18,675 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:45:21,380 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8226, 3.8024, 2.5012, 4.5580, 3.0592, 4.4446, 2.4942, 3.0673], device='cuda:5'), covar=tensor([0.0260, 0.0375, 0.1484, 0.0197, 0.0728, 0.0563, 0.1417, 0.0755], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0171, 0.0190, 0.0151, 0.0173, 0.0211, 0.0198, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:45:29,137 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8809, 5.3354, 5.4890, 5.3152, 5.3217, 5.8763, 5.3245, 5.1601], device='cuda:5'), covar=tensor([0.1017, 0.1804, 0.1960, 0.1895, 0.2190, 0.0880, 0.1599, 0.2533], device='cuda:5'), in_proj_covar=tensor([0.0390, 0.0561, 0.0616, 0.0471, 0.0631, 0.0642, 0.0488, 0.0636], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 17:46:17,850 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.727e+02 3.457e+02 4.156e+02 1.353e+03, threshold=6.914e+02, percent-clipped=3.0 2023-04-30 17:46:19,095 INFO [train.py:904] (5/8) Epoch 18, batch 6000, loss[loss=0.2177, simple_loss=0.3021, pruned_loss=0.06663, over 16620.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2976, pruned_loss=0.06307, over 3086522.62 frames. ], batch size: 57, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:46:19,095 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 17:46:29,951 INFO [train.py:938] (5/8) Epoch 18, validation: loss=0.1531, simple_loss=0.2661, pruned_loss=0.02007, over 944034.00 frames. 2023-04-30 17:46:29,952 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 17:47:48,048 INFO [train.py:904] (5/8) Epoch 18, batch 6050, loss[loss=0.1715, simple_loss=0.2772, pruned_loss=0.03292, over 16780.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.296, pruned_loss=0.0624, over 3090099.37 frames. ], batch size: 83, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:48:14,344 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178619.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:48:55,315 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7784, 2.3481, 1.8895, 2.0762, 2.7173, 2.3224, 2.5889, 2.8418], device='cuda:5'), covar=tensor([0.0169, 0.0360, 0.0482, 0.0433, 0.0231, 0.0358, 0.0218, 0.0232], device='cuda:5'), in_proj_covar=tensor([0.0191, 0.0224, 0.0217, 0.0216, 0.0226, 0.0224, 0.0226, 0.0220], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 17:49:06,488 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.787e+02 3.191e+02 3.956e+02 7.476e+02, threshold=6.381e+02, percent-clipped=1.0 2023-04-30 17:49:06,510 INFO [train.py:904] (5/8) Epoch 18, batch 6100, loss[loss=0.2251, simple_loss=0.3144, pruned_loss=0.0679, over 16362.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2954, pruned_loss=0.06131, over 3112203.72 frames. ], batch size: 165, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:49:24,100 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7953, 2.4979, 2.4463, 3.2785, 2.3934, 3.6175, 1.5137, 2.7267], device='cuda:5'), covar=tensor([0.1329, 0.0733, 0.1147, 0.0181, 0.0206, 0.0429, 0.1694, 0.0814], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0170, 0.0191, 0.0180, 0.0206, 0.0215, 0.0196, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:49:51,623 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:13,098 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:16,794 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178697.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:50:18,842 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6506, 3.5049, 4.0478, 1.9545, 4.3569, 4.3249, 3.1353, 3.1222], device='cuda:5'), covar=tensor([0.0793, 0.0291, 0.0188, 0.1244, 0.0054, 0.0136, 0.0385, 0.0466], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0106, 0.0095, 0.0139, 0.0076, 0.0122, 0.0125, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 17:50:23,993 INFO [train.py:904] (5/8) Epoch 18, batch 6150, loss[loss=0.2013, simple_loss=0.2903, pruned_loss=0.05618, over 16449.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2934, pruned_loss=0.06066, over 3121086.12 frames. ], batch size: 146, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:34,492 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-30 17:51:39,659 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 2.816e+02 3.285e+02 4.160e+02 8.412e+02, threshold=6.570e+02, percent-clipped=4.0 2023-04-30 17:51:39,682 INFO [train.py:904] (5/8) Epoch 18, batch 6200, loss[loss=0.1831, simple_loss=0.2748, pruned_loss=0.04575, over 17256.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2918, pruned_loss=0.06014, over 3113701.96 frames. ], batch size: 52, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:45,819 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178755.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 17:51:46,922 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4697, 4.4686, 4.3689, 3.6241, 4.4205, 1.7310, 4.1910, 4.0466], device='cuda:5'), covar=tensor([0.0109, 0.0092, 0.0173, 0.0370, 0.0095, 0.2689, 0.0136, 0.0230], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0142, 0.0188, 0.0173, 0.0163, 0.0198, 0.0178, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 17:51:49,721 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178758.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:52:08,985 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:52:12,923 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3957, 3.3445, 3.3910, 3.4840, 3.4962, 3.2903, 3.4907, 3.5464], device='cuda:5'), covar=tensor([0.1254, 0.0941, 0.1085, 0.0653, 0.0741, 0.2366, 0.0978, 0.0760], device='cuda:5'), in_proj_covar=tensor([0.0602, 0.0741, 0.0871, 0.0759, 0.0564, 0.0597, 0.0610, 0.0705], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 17:52:56,578 INFO [train.py:904] (5/8) Epoch 18, batch 6250, loss[loss=0.2464, simple_loss=0.3094, pruned_loss=0.09167, over 11489.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2925, pruned_loss=0.06083, over 3096341.07 frames. ], batch size: 246, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:53:07,999 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:53:42,362 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:54:15,701 INFO [train.py:904] (5/8) Epoch 18, batch 6300, loss[loss=0.1965, simple_loss=0.2831, pruned_loss=0.05498, over 16988.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.292, pruned_loss=0.05984, over 3113889.48 frames. ], batch size: 55, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:54:17,526 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.018e+02 2.791e+02 3.273e+02 3.863e+02 9.337e+02, threshold=6.545e+02, percent-clipped=2.0 2023-04-30 17:55:22,845 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178894.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:55:34,230 INFO [train.py:904] (5/8) Epoch 18, batch 6350, loss[loss=0.1949, simple_loss=0.283, pruned_loss=0.05343, over 16432.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2922, pruned_loss=0.06073, over 3094853.62 frames. ], batch size: 146, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:56:52,051 INFO [train.py:904] (5/8) Epoch 18, batch 6400, loss[loss=0.1955, simple_loss=0.277, pruned_loss=0.05705, over 16160.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.292, pruned_loss=0.06133, over 3093939.71 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:56:53,838 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 3.036e+02 3.548e+02 4.325e+02 9.483e+02, threshold=7.097e+02, percent-clipped=4.0 2023-04-30 17:56:57,550 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178955.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 17:57:28,090 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178975.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:58:07,508 INFO [train.py:904] (5/8) Epoch 18, batch 6450, loss[loss=0.1865, simple_loss=0.2789, pruned_loss=0.04708, over 16838.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2918, pruned_loss=0.06007, over 3107389.64 frames. ], batch size: 96, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:23,101 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4102, 3.4254, 2.0385, 3.7746, 2.6337, 3.7726, 2.2688, 2.7973], device='cuda:5'), covar=tensor([0.0279, 0.0401, 0.1666, 0.0226, 0.0794, 0.0608, 0.1469, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0152, 0.0174, 0.0212, 0.0199, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 17:59:24,081 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179050.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:59:26,167 INFO [train.py:904] (5/8) Epoch 18, batch 6500, loss[loss=0.221, simple_loss=0.2878, pruned_loss=0.07714, over 11579.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2895, pruned_loss=0.05979, over 3089209.91 frames. ], batch size: 248, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:27,324 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 2.918e+02 3.772e+02 4.451e+02 8.133e+02, threshold=7.544e+02, percent-clipped=2.0 2023-04-30 17:59:28,322 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179053.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 17:59:59,168 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 18:00:13,671 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8468, 3.2454, 3.3461, 1.8587, 2.9118, 2.2532, 3.3599, 3.4736], device='cuda:5'), covar=tensor([0.0265, 0.0742, 0.0562, 0.2065, 0.0805, 0.0978, 0.0612, 0.0817], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0159, 0.0164, 0.0149, 0.0141, 0.0127, 0.0141, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 18:00:44,335 INFO [train.py:904] (5/8) Epoch 18, batch 6550, loss[loss=0.2003, simple_loss=0.3022, pruned_loss=0.04921, over 16141.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2926, pruned_loss=0.06091, over 3090771.18 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:00:52,453 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 18:00:54,407 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179108.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:20,187 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179126.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:58,904 INFO [train.py:904] (5/8) Epoch 18, batch 6600, loss[loss=0.2097, simple_loss=0.2983, pruned_loss=0.06061, over 16419.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2948, pruned_loss=0.06126, over 3119791.77 frames. ], batch size: 146, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:02:00,672 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.804e+02 3.237e+02 3.890e+02 6.888e+02, threshold=6.473e+02, percent-clipped=0.0 2023-04-30 18:02:05,165 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179156.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:02:22,932 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-04-30 18:03:17,526 INFO [train.py:904] (5/8) Epoch 18, batch 6650, loss[loss=0.2208, simple_loss=0.295, pruned_loss=0.07332, over 16628.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2947, pruned_loss=0.06206, over 3110975.05 frames. ], batch size: 57, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:05,222 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6591, 2.5668, 1.8700, 2.6987, 2.1116, 2.7560, 2.1028, 2.3840], device='cuda:5'), covar=tensor([0.0325, 0.0367, 0.1229, 0.0229, 0.0633, 0.0452, 0.1193, 0.0602], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0171, 0.0191, 0.0152, 0.0173, 0.0212, 0.0199, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 18:04:17,627 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8707, 3.3411, 3.3264, 1.9524, 2.8710, 2.1866, 3.2924, 3.5138], device='cuda:5'), covar=tensor([0.0302, 0.0737, 0.0611, 0.2032, 0.0868, 0.1017, 0.0720, 0.0957], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0159, 0.0165, 0.0149, 0.0142, 0.0127, 0.0141, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 18:04:30,599 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179250.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 18:04:32,473 INFO [train.py:904] (5/8) Epoch 18, batch 6700, loss[loss=0.2033, simple_loss=0.2907, pruned_loss=0.05789, over 16238.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2939, pruned_loss=0.0626, over 3105665.36 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:34,183 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.690e+02 3.432e+02 4.163e+02 9.246e+02, threshold=6.864e+02, percent-clipped=3.0 2023-04-30 18:05:09,305 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:05:46,460 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1126, 1.5149, 1.9563, 2.1302, 2.2402, 2.3658, 1.7111, 2.2515], device='cuda:5'), covar=tensor([0.0209, 0.0437, 0.0240, 0.0283, 0.0262, 0.0191, 0.0464, 0.0129], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0187, 0.0173, 0.0177, 0.0186, 0.0145, 0.0190, 0.0140], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:05:48,870 INFO [train.py:904] (5/8) Epoch 18, batch 6750, loss[loss=0.188, simple_loss=0.2791, pruned_loss=0.04844, over 16845.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2933, pruned_loss=0.06284, over 3109422.80 frames. ], batch size: 96, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:06:20,235 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:01,079 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:03,108 INFO [train.py:904] (5/8) Epoch 18, batch 6800, loss[loss=0.2007, simple_loss=0.2822, pruned_loss=0.05962, over 16509.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2937, pruned_loss=0.06312, over 3091819.32 frames. ], batch size: 68, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:07:04,932 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.988e+02 2.962e+02 3.412e+02 4.074e+02 1.001e+03, threshold=6.824e+02, percent-clipped=4.0 2023-04-30 18:07:05,956 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179353.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:07:18,280 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:18,604 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 18:07:48,649 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9957, 2.8542, 2.8694, 2.1170, 2.7012, 2.2069, 2.7128, 2.9885], device='cuda:5'), covar=tensor([0.0275, 0.0688, 0.0492, 0.1544, 0.0707, 0.0878, 0.0562, 0.0666], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0160, 0.0165, 0.0150, 0.0142, 0.0128, 0.0142, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 18:08:16,243 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179398.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:20,458 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179401.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:08:21,239 INFO [train.py:904] (5/8) Epoch 18, batch 6850, loss[loss=0.2576, simple_loss=0.3197, pruned_loss=0.09777, over 11859.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2951, pruned_loss=0.06385, over 3081611.32 frames. ], batch size: 248, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:08:23,239 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 18:08:36,475 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 18:08:48,873 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3325, 3.5005, 3.6643, 3.6146, 3.6292, 3.4323, 3.4592, 3.4999], device='cuda:5'), covar=tensor([0.0428, 0.0670, 0.0488, 0.0498, 0.0593, 0.0581, 0.0912, 0.0574], device='cuda:5'), in_proj_covar=tensor([0.0389, 0.0425, 0.0414, 0.0391, 0.0464, 0.0436, 0.0534, 0.0349], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 18:08:50,239 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:56,280 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:09:11,271 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 18:09:24,215 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179444.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:09:34,753 INFO [train.py:904] (5/8) Epoch 18, batch 6900, loss[loss=0.2721, simple_loss=0.3314, pruned_loss=0.1064, over 11601.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2974, pruned_loss=0.0637, over 3077409.02 frames. ], batch size: 248, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:09:38,471 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.677e+02 2.748e+02 3.355e+02 3.841e+02 5.597e+02, threshold=6.710e+02, percent-clipped=0.0 2023-04-30 18:09:58,312 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 18:10:10,838 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:10:53,475 INFO [train.py:904] (5/8) Epoch 18, batch 6950, loss[loss=0.2076, simple_loss=0.2984, pruned_loss=0.05838, over 16652.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2995, pruned_loss=0.06573, over 3055168.27 frames. ], batch size: 57, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:10:58,877 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179505.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:12:07,081 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179550.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:12:09,850 INFO [train.py:904] (5/8) Epoch 18, batch 7000, loss[loss=0.1943, simple_loss=0.2961, pruned_loss=0.04624, over 16656.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2991, pruned_loss=0.06428, over 3084124.43 frames. ], batch size: 134, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:12:12,179 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.791e+02 3.383e+02 4.252e+02 7.699e+02, threshold=6.767e+02, percent-clipped=2.0 2023-04-30 18:12:56,549 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 18:13:16,967 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179598.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:13:21,680 INFO [train.py:904] (5/8) Epoch 18, batch 7050, loss[loss=0.1988, simple_loss=0.2895, pruned_loss=0.05411, over 16773.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2992, pruned_loss=0.06325, over 3105777.21 frames. ], batch size: 124, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:13:36,670 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3850, 2.9275, 2.6633, 2.2916, 2.2804, 2.2836, 2.8812, 2.8891], device='cuda:5'), covar=tensor([0.2226, 0.0763, 0.1551, 0.2337, 0.2120, 0.1939, 0.0475, 0.1231], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0265, 0.0300, 0.0306, 0.0292, 0.0249, 0.0287, 0.0329], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:14:12,909 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179636.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:14:37,726 INFO [train.py:904] (5/8) Epoch 18, batch 7100, loss[loss=0.1964, simple_loss=0.2803, pruned_loss=0.05627, over 16687.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2979, pruned_loss=0.06285, over 3106040.62 frames. ], batch size: 62, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:14:40,304 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.117e+02 2.863e+02 3.448e+02 4.199e+02 1.001e+03, threshold=6.895e+02, percent-clipped=2.0 2023-04-30 18:15:33,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5982, 1.6355, 2.2595, 2.4957, 2.4834, 2.9474, 1.9017, 2.8978], device='cuda:5'), covar=tensor([0.0229, 0.0560, 0.0326, 0.0337, 0.0329, 0.0165, 0.0551, 0.0127], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0189, 0.0175, 0.0178, 0.0187, 0.0145, 0.0191, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:15:46,862 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:15:53,689 INFO [train.py:904] (5/8) Epoch 18, batch 7150, loss[loss=0.2728, simple_loss=0.3267, pruned_loss=0.1095, over 11463.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2952, pruned_loss=0.06216, over 3114697.62 frames. ], batch size: 248, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:15:55,576 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-04-30 18:16:09,068 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5784, 3.7408, 2.8057, 2.2597, 2.5802, 2.3356, 4.0865, 3.3736], device='cuda:5'), covar=tensor([0.2892, 0.0710, 0.1823, 0.2520, 0.2492, 0.2053, 0.0445, 0.1155], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0292, 0.0249, 0.0287, 0.0328], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:16:16,358 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179717.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:16:27,327 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-04-30 18:17:00,434 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3336, 2.9383, 2.6861, 2.2910, 2.2094, 2.2469, 2.8604, 2.8345], device='cuda:5'), covar=tensor([0.2430, 0.0825, 0.1600, 0.2523, 0.2616, 0.2198, 0.0522, 0.1336], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0291, 0.0249, 0.0286, 0.0328], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:17:08,213 INFO [train.py:904] (5/8) Epoch 18, batch 7200, loss[loss=0.1742, simple_loss=0.2691, pruned_loss=0.03963, over 16727.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2931, pruned_loss=0.06099, over 3090850.16 frames. ], batch size: 134, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:17:10,639 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.765e+02 3.366e+02 4.196e+02 7.871e+02, threshold=6.733e+02, percent-clipped=4.0 2023-04-30 18:17:29,501 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 18:17:30,287 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2119, 2.3967, 2.0567, 2.1595, 2.7898, 2.3827, 2.8385, 2.9377], device='cuda:5'), covar=tensor([0.0127, 0.0442, 0.0551, 0.0489, 0.0270, 0.0394, 0.0254, 0.0270], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0219], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:17:54,666 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 18:18:18,017 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:18:24,746 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179800.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:18:27,060 INFO [train.py:904] (5/8) Epoch 18, batch 7250, loss[loss=0.1727, simple_loss=0.2639, pruned_loss=0.04077, over 16838.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2905, pruned_loss=0.05988, over 3072800.21 frames. ], batch size: 102, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:18:29,399 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2287, 3.4560, 3.5934, 3.5514, 3.5611, 3.3692, 3.4283, 3.4372], device='cuda:5'), covar=tensor([0.0409, 0.0612, 0.0465, 0.0454, 0.0505, 0.0547, 0.0770, 0.0563], device='cuda:5'), in_proj_covar=tensor([0.0390, 0.0426, 0.0416, 0.0393, 0.0467, 0.0438, 0.0534, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 18:19:41,460 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5904, 1.7121, 2.2506, 2.4897, 2.5418, 2.9760, 1.9016, 2.9122], device='cuda:5'), covar=tensor([0.0205, 0.0470, 0.0268, 0.0301, 0.0270, 0.0146, 0.0476, 0.0120], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0186, 0.0173, 0.0176, 0.0185, 0.0143, 0.0189, 0.0139], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:19:42,148 INFO [train.py:904] (5/8) Epoch 18, batch 7300, loss[loss=0.1962, simple_loss=0.2977, pruned_loss=0.04742, over 16820.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2905, pruned_loss=0.06017, over 3056855.76 frames. ], batch size: 102, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:19:45,256 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.839e+02 2.793e+02 3.453e+02 4.292e+02 8.148e+02, threshold=6.907e+02, percent-clipped=1.0 2023-04-30 18:19:51,015 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:20:33,695 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7663, 3.8120, 3.9216, 3.7354, 3.8662, 4.2065, 3.8871, 3.6299], device='cuda:5'), covar=tensor([0.2015, 0.2083, 0.2100, 0.2291, 0.2282, 0.1601, 0.1667, 0.2584], device='cuda:5'), in_proj_covar=tensor([0.0390, 0.0559, 0.0618, 0.0470, 0.0628, 0.0644, 0.0486, 0.0631], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:20:58,401 INFO [train.py:904] (5/8) Epoch 18, batch 7350, loss[loss=0.2346, simple_loss=0.3025, pruned_loss=0.0833, over 11316.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2913, pruned_loss=0.0606, over 3057730.55 frames. ], batch size: 246, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:21:05,545 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179906.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:21:32,191 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:22:16,601 INFO [train.py:904] (5/8) Epoch 18, batch 7400, loss[loss=0.232, simple_loss=0.3029, pruned_loss=0.08057, over 11343.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2922, pruned_loss=0.06141, over 3047712.11 frames. ], batch size: 247, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:22:19,966 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.864e+02 3.444e+02 4.186e+02 8.589e+02, threshold=6.889e+02, percent-clipped=1.0 2023-04-30 18:22:30,124 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5927, 2.2356, 1.7855, 2.0007, 2.5683, 2.1955, 2.3640, 2.6925], device='cuda:5'), covar=tensor([0.0167, 0.0385, 0.0539, 0.0475, 0.0246, 0.0402, 0.0229, 0.0236], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:22:41,126 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179967.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:23:09,747 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:23:21,199 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:23:39,786 INFO [train.py:904] (5/8) Epoch 18, batch 7450, loss[loss=0.209, simple_loss=0.3, pruned_loss=0.05905, over 15363.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2937, pruned_loss=0.06238, over 3053598.29 frames. ], batch size: 192, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:24:05,956 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:24:10,831 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0384, 2.4893, 2.5570, 1.8325, 2.7220, 2.7547, 2.4776, 2.4085], device='cuda:5'), covar=tensor([0.0713, 0.0266, 0.0284, 0.1054, 0.0141, 0.0362, 0.0465, 0.0445], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0105, 0.0094, 0.0138, 0.0075, 0.0120, 0.0124, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 18:24:19,824 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:24:37,384 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1076, 1.5369, 1.9218, 2.0821, 2.2193, 2.3295, 1.6806, 2.1982], device='cuda:5'), covar=tensor([0.0211, 0.0455, 0.0267, 0.0304, 0.0273, 0.0187, 0.0491, 0.0136], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0187, 0.0173, 0.0175, 0.0185, 0.0143, 0.0189, 0.0139], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:24:50,314 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4897, 3.4575, 2.7042, 2.1627, 2.3166, 2.2855, 3.5717, 3.1729], device='cuda:5'), covar=tensor([0.2929, 0.0716, 0.1815, 0.2705, 0.2515, 0.2113, 0.0527, 0.1308], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0263, 0.0299, 0.0304, 0.0291, 0.0247, 0.0285, 0.0327], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:25:01,496 INFO [train.py:904] (5/8) Epoch 18, batch 7500, loss[loss=0.1766, simple_loss=0.2652, pruned_loss=0.04403, over 16626.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2939, pruned_loss=0.06215, over 3027113.30 frames. ], batch size: 76, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:25:04,524 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 3.002e+02 3.628e+02 4.895e+02 8.341e+02, threshold=7.256e+02, percent-clipped=3.0 2023-04-30 18:25:22,816 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:25:33,720 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5166, 4.5659, 4.3828, 4.0853, 4.0646, 4.4718, 4.2287, 4.1381], device='cuda:5'), covar=tensor([0.0604, 0.0519, 0.0291, 0.0310, 0.0898, 0.0459, 0.0603, 0.0695], device='cuda:5'), in_proj_covar=tensor([0.0272, 0.0387, 0.0317, 0.0309, 0.0331, 0.0360, 0.0218, 0.0381], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:25:56,689 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180087.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:26:16,508 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180100.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:26:19,022 INFO [train.py:904] (5/8) Epoch 18, batch 7550, loss[loss=0.2017, simple_loss=0.2903, pruned_loss=0.05652, over 16834.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2928, pruned_loss=0.06248, over 3033273.00 frames. ], batch size: 124, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:26:47,602 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-30 18:27:18,255 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7150, 4.9879, 5.1458, 4.9647, 4.9779, 5.5324, 4.9213, 4.7003], device='cuda:5'), covar=tensor([0.1020, 0.1698, 0.1967, 0.1808, 0.2132, 0.0892, 0.1713, 0.2455], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0559, 0.0619, 0.0469, 0.0627, 0.0644, 0.0488, 0.0631], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:27:30,331 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180148.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:27:36,550 INFO [train.py:904] (5/8) Epoch 18, batch 7600, loss[loss=0.2281, simple_loss=0.3066, pruned_loss=0.07478, over 15389.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2917, pruned_loss=0.06234, over 3044683.59 frames. ], batch size: 190, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:27:36,889 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180152.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:27:39,408 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.698e+02 3.315e+02 4.368e+02 9.372e+02, threshold=6.630e+02, percent-clipped=3.0 2023-04-30 18:28:55,655 INFO [train.py:904] (5/8) Epoch 18, batch 7650, loss[loss=0.2091, simple_loss=0.2968, pruned_loss=0.06067, over 16525.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2923, pruned_loss=0.0626, over 3054418.70 frames. ], batch size: 68, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:29:00,844 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:29:05,859 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0263, 2.3519, 2.3596, 2.8044, 1.9350, 3.1837, 1.7883, 2.6740], device='cuda:5'), covar=tensor([0.1231, 0.0688, 0.1053, 0.0229, 0.0135, 0.0372, 0.1567, 0.0741], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0178, 0.0206, 0.0214, 0.0196, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 18:30:13,296 INFO [train.py:904] (5/8) Epoch 18, batch 7700, loss[loss=0.2034, simple_loss=0.2867, pruned_loss=0.06004, over 16849.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2924, pruned_loss=0.06305, over 3049885.82 frames. ], batch size: 96, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:30:18,209 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 3.108e+02 3.616e+02 4.495e+02 6.527e+02, threshold=7.232e+02, percent-clipped=0.0 2023-04-30 18:30:20,088 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4231, 3.5018, 3.8404, 1.7067, 3.9999, 4.0386, 3.0455, 2.8397], device='cuda:5'), covar=tensor([0.0929, 0.0251, 0.0200, 0.1473, 0.0085, 0.0200, 0.0438, 0.0550], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0105, 0.0094, 0.0139, 0.0075, 0.0121, 0.0125, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 18:30:29,314 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180262.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:30:35,640 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:30:57,480 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:31:15,929 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:31:30,814 INFO [train.py:904] (5/8) Epoch 18, batch 7750, loss[loss=0.2106, simple_loss=0.2998, pruned_loss=0.06065, over 16937.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2928, pruned_loss=0.06294, over 3056867.98 frames. ], batch size: 109, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:32:23,819 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180336.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:32:29,344 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180340.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:32:46,616 INFO [train.py:904] (5/8) Epoch 18, batch 7800, loss[loss=0.1843, simple_loss=0.2753, pruned_loss=0.04667, over 16273.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.293, pruned_loss=0.06243, over 3075407.41 frames. ], batch size: 35, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:32:51,025 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.996e+02 3.455e+02 4.260e+02 9.369e+02, threshold=6.911e+02, percent-clipped=1.0 2023-04-30 18:32:55,339 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1499, 4.1879, 4.1055, 3.3429, 4.1228, 1.6909, 3.9076, 3.7934], device='cuda:5'), covar=tensor([0.0146, 0.0123, 0.0196, 0.0363, 0.0120, 0.2857, 0.0176, 0.0262], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0138, 0.0184, 0.0168, 0.0159, 0.0195, 0.0173, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:33:33,624 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:33:55,576 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180397.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:33:56,871 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5024, 3.3481, 3.8338, 1.8733, 4.0013, 3.9997, 3.0004, 2.8546], device='cuda:5'), covar=tensor([0.0804, 0.0271, 0.0160, 0.1166, 0.0060, 0.0151, 0.0412, 0.0476], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0105, 0.0094, 0.0138, 0.0075, 0.0120, 0.0124, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 18:34:02,035 INFO [train.py:904] (5/8) Epoch 18, batch 7850, loss[loss=0.2269, simple_loss=0.2936, pruned_loss=0.08016, over 11677.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2936, pruned_loss=0.06212, over 3070849.25 frames. ], batch size: 247, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:34:10,416 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3372, 5.6434, 5.3346, 5.3861, 5.1041, 5.0557, 4.9770, 5.7260], device='cuda:5'), covar=tensor([0.1167, 0.0737, 0.1009, 0.0891, 0.0790, 0.0648, 0.1182, 0.0938], device='cuda:5'), in_proj_covar=tensor([0.0618, 0.0760, 0.0624, 0.0569, 0.0475, 0.0490, 0.0636, 0.0591], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:34:59,669 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4105, 4.6764, 4.4505, 4.4910, 4.2245, 4.1804, 4.1580, 4.7151], device='cuda:5'), covar=tensor([0.1108, 0.0783, 0.1002, 0.0820, 0.0741, 0.1304, 0.1078, 0.0872], device='cuda:5'), in_proj_covar=tensor([0.0619, 0.0761, 0.0626, 0.0570, 0.0475, 0.0490, 0.0637, 0.0591], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:35:16,219 INFO [train.py:904] (5/8) Epoch 18, batch 7900, loss[loss=0.2067, simple_loss=0.2968, pruned_loss=0.05831, over 16283.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2925, pruned_loss=0.06144, over 3081137.94 frames. ], batch size: 165, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:35:17,220 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:35:20,363 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.825e+02 3.445e+02 4.253e+02 7.547e+02, threshold=6.890e+02, percent-clipped=0.0 2023-04-30 18:36:25,806 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 18:36:32,647 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180500.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:36:34,734 INFO [train.py:904] (5/8) Epoch 18, batch 7950, loss[loss=0.1889, simple_loss=0.2712, pruned_loss=0.05333, over 16673.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2936, pruned_loss=0.0622, over 3073403.48 frames. ], batch size: 57, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:37:35,949 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 18:37:52,815 INFO [train.py:904] (5/8) Epoch 18, batch 8000, loss[loss=0.2474, simple_loss=0.3254, pruned_loss=0.08473, over 15363.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2944, pruned_loss=0.06314, over 3059926.71 frames. ], batch size: 191, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:37:57,089 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.710e+02 3.198e+02 3.720e+02 8.207e+02, threshold=6.397e+02, percent-clipped=2.0 2023-04-30 18:38:07,731 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:08,942 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180562.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:38:21,359 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:36,602 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:10,321 INFO [train.py:904] (5/8) Epoch 18, batch 8050, loss[loss=0.2538, simple_loss=0.3234, pruned_loss=0.09208, over 11636.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2941, pruned_loss=0.06276, over 3067635.76 frames. ], batch size: 248, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:39:11,850 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 18:39:23,127 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180610.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:39:50,350 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:54,607 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:40:09,075 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5466, 2.6686, 2.1014, 2.4030, 3.0231, 2.6735, 3.1954, 3.2547], device='cuda:5'), covar=tensor([0.0099, 0.0367, 0.0527, 0.0396, 0.0251, 0.0354, 0.0214, 0.0229], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:40:26,597 INFO [train.py:904] (5/8) Epoch 18, batch 8100, loss[loss=0.1877, simple_loss=0.2729, pruned_loss=0.05125, over 16962.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2931, pruned_loss=0.06177, over 3072967.77 frames. ], batch size: 41, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:40:32,033 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.097e+02 2.839e+02 3.233e+02 3.954e+02 8.532e+02, threshold=6.466e+02, percent-clipped=3.0 2023-04-30 18:40:42,069 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2034, 4.2139, 4.1254, 3.3448, 4.1421, 1.6432, 3.9216, 3.7908], device='cuda:5'), covar=tensor([0.0124, 0.0095, 0.0181, 0.0339, 0.0106, 0.2836, 0.0144, 0.0249], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0139, 0.0185, 0.0170, 0.0160, 0.0196, 0.0174, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:41:00,884 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4537, 3.2609, 2.6838, 2.1674, 2.2327, 2.2516, 3.5007, 3.0795], device='cuda:5'), covar=tensor([0.3088, 0.0946, 0.1975, 0.2784, 0.2764, 0.2176, 0.0575, 0.1423], device='cuda:5'), in_proj_covar=tensor([0.0323, 0.0265, 0.0301, 0.0306, 0.0294, 0.0250, 0.0288, 0.0328], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:41:12,143 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:26,945 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:41,235 INFO [train.py:904] (5/8) Epoch 18, batch 8150, loss[loss=0.1729, simple_loss=0.2577, pruned_loss=0.04405, over 16546.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2902, pruned_loss=0.06025, over 3098430.65 frames. ], batch size: 68, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:42:03,541 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-04-30 18:42:24,169 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180730.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:25,748 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-04-30 18:42:26,692 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:56,059 INFO [train.py:904] (5/8) Epoch 18, batch 8200, loss[loss=0.1967, simple_loss=0.2913, pruned_loss=0.05102, over 16835.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2877, pruned_loss=0.05936, over 3105409.52 frames. ], batch size: 116, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:43:02,075 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.834e+02 3.334e+02 4.232e+02 1.685e+03, threshold=6.669e+02, percent-clipped=1.0 2023-04-30 18:43:12,721 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6030, 1.7488, 2.2294, 2.5361, 2.5669, 2.8583, 1.9767, 2.8588], device='cuda:5'), covar=tensor([0.0188, 0.0473, 0.0308, 0.0284, 0.0277, 0.0184, 0.0427, 0.0122], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0175, 0.0184, 0.0143, 0.0188, 0.0139], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:44:01,972 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180793.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:44:14,893 INFO [train.py:904] (5/8) Epoch 18, batch 8250, loss[loss=0.1965, simple_loss=0.2951, pruned_loss=0.049, over 16254.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2866, pruned_loss=0.05678, over 3093614.45 frames. ], batch size: 165, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:44:42,419 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 18:44:50,908 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6900, 4.7566, 4.5546, 4.1768, 4.1808, 4.6530, 4.5094, 4.3436], device='cuda:5'), covar=tensor([0.0600, 0.0473, 0.0342, 0.0342, 0.1093, 0.0460, 0.0401, 0.0734], device='cuda:5'), in_proj_covar=tensor([0.0275, 0.0392, 0.0321, 0.0310, 0.0332, 0.0363, 0.0220, 0.0383], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:44:53,333 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180825.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:45:21,702 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 18:45:37,339 INFO [train.py:904] (5/8) Epoch 18, batch 8300, loss[loss=0.1722, simple_loss=0.2678, pruned_loss=0.03832, over 16519.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2841, pruned_loss=0.05417, over 3079941.48 frames. ], batch size: 62, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:45:43,845 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.270e+02 2.749e+02 3.299e+02 7.574e+02, threshold=5.499e+02, percent-clipped=1.0 2023-04-30 18:45:44,502 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8123, 3.7919, 3.8742, 3.6732, 3.8170, 4.2613, 3.9023, 3.5798], device='cuda:5'), covar=tensor([0.1911, 0.2311, 0.2491, 0.2766, 0.2856, 0.1524, 0.1614, 0.2800], device='cuda:5'), in_proj_covar=tensor([0.0386, 0.0556, 0.0620, 0.0468, 0.0624, 0.0643, 0.0488, 0.0630], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:45:52,501 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180861.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:27,105 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-04-30 18:46:32,622 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:34,406 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5355, 4.5026, 4.3432, 3.6516, 4.4171, 1.6845, 4.1431, 4.1549], device='cuda:5'), covar=tensor([0.0080, 0.0082, 0.0161, 0.0308, 0.0093, 0.2809, 0.0132, 0.0205], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0138, 0.0183, 0.0168, 0.0159, 0.0195, 0.0173, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:46:58,363 INFO [train.py:904] (5/8) Epoch 18, batch 8350, loss[loss=0.1902, simple_loss=0.2861, pruned_loss=0.04713, over 16749.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2836, pruned_loss=0.05235, over 3078189.91 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:47:09,895 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:47:10,151 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5694, 3.6595, 2.8383, 2.1093, 2.2866, 2.3103, 3.8394, 3.2404], device='cuda:5'), covar=tensor([0.2764, 0.0555, 0.1655, 0.2970, 0.2798, 0.2189, 0.0400, 0.1207], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0259, 0.0294, 0.0301, 0.0289, 0.0245, 0.0283, 0.0321], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:47:29,627 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-30 18:47:36,087 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180926.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:48:16,300 INFO [train.py:904] (5/8) Epoch 18, batch 8400, loss[loss=0.189, simple_loss=0.2738, pruned_loss=0.0521, over 12323.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2812, pruned_loss=0.05047, over 3056912.17 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:48:22,157 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.220e+02 2.634e+02 3.245e+02 6.969e+02, threshold=5.268e+02, percent-clipped=3.0 2023-04-30 18:49:16,510 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:49:31,280 INFO [train.py:904] (5/8) Epoch 18, batch 8450, loss[loss=0.1748, simple_loss=0.2673, pruned_loss=0.04111, over 16674.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.28, pruned_loss=0.04935, over 3053971.31 frames. ], batch size: 134, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:50:31,763 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181040.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:50:50,508 INFO [train.py:904] (5/8) Epoch 18, batch 8500, loss[loss=0.161, simple_loss=0.2602, pruned_loss=0.03088, over 16728.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.276, pruned_loss=0.0471, over 3043280.03 frames. ], batch size: 83, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:50:58,703 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.304e+02 3.083e+02 3.822e+02 7.945e+02, threshold=6.166e+02, percent-clipped=7.0 2023-04-30 18:51:48,671 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:52:13,302 INFO [train.py:904] (5/8) Epoch 18, batch 8550, loss[loss=0.1817, simple_loss=0.2778, pruned_loss=0.0428, over 15287.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2735, pruned_loss=0.04577, over 3037760.63 frames. ], batch size: 191, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:53:50,716 INFO [train.py:904] (5/8) Epoch 18, batch 8600, loss[loss=0.1673, simple_loss=0.2522, pruned_loss=0.0412, over 12330.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2738, pruned_loss=0.04462, over 3046007.27 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:54:01,227 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.216e+02 2.624e+02 3.286e+02 8.410e+02, threshold=5.248e+02, percent-clipped=1.0 2023-04-30 18:54:50,073 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181181.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:55:29,860 INFO [train.py:904] (5/8) Epoch 18, batch 8650, loss[loss=0.171, simple_loss=0.2728, pruned_loss=0.03458, over 16291.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2725, pruned_loss=0.04338, over 3062709.71 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:56:25,748 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181226.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:57:15,678 INFO [train.py:904] (5/8) Epoch 18, batch 8700, loss[loss=0.1825, simple_loss=0.2761, pruned_loss=0.04442, over 15317.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2691, pruned_loss=0.04197, over 3037086.57 frames. ], batch size: 190, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:57:25,080 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.245e+02 2.629e+02 3.109e+02 5.522e+02, threshold=5.258e+02, percent-clipped=1.0 2023-04-30 18:57:42,656 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3264, 2.0829, 2.1878, 3.9775, 2.0373, 2.4502, 2.1989, 2.2418], device='cuda:5'), covar=tensor([0.1172, 0.4068, 0.3074, 0.0485, 0.4733, 0.2762, 0.3860, 0.3732], device='cuda:5'), in_proj_covar=tensor([0.0377, 0.0420, 0.0347, 0.0311, 0.0420, 0.0480, 0.0389, 0.0489], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:57:43,764 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6752, 4.7528, 4.5636, 4.2183, 4.2205, 4.6526, 4.4683, 4.3810], device='cuda:5'), covar=tensor([0.0602, 0.0535, 0.0325, 0.0290, 0.0897, 0.0487, 0.0443, 0.0619], device='cuda:5'), in_proj_covar=tensor([0.0272, 0.0387, 0.0317, 0.0307, 0.0326, 0.0359, 0.0220, 0.0378], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:57:53,333 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8544, 1.9010, 2.2426, 3.1434, 1.9972, 2.0796, 2.1316, 1.9974], device='cuda:5'), covar=tensor([0.1461, 0.4830, 0.2928, 0.0752, 0.5533, 0.3439, 0.4153, 0.4498], device='cuda:5'), in_proj_covar=tensor([0.0377, 0.0420, 0.0347, 0.0311, 0.0420, 0.0480, 0.0389, 0.0489], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:57:53,625 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 18:57:56,766 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:29,723 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8170, 1.3131, 1.7482, 1.6842, 1.8550, 1.9320, 1.6289, 1.8441], device='cuda:5'), covar=tensor([0.0258, 0.0412, 0.0205, 0.0314, 0.0292, 0.0198, 0.0411, 0.0118], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0183, 0.0169, 0.0171, 0.0181, 0.0139, 0.0184, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 18:58:39,637 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181297.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:48,993 INFO [train.py:904] (5/8) Epoch 18, batch 8750, loss[loss=0.1742, simple_loss=0.268, pruned_loss=0.04023, over 12300.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2694, pruned_loss=0.0419, over 3044317.38 frames. ], batch size: 250, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:59:10,151 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1525, 3.2407, 3.2329, 2.1626, 2.9957, 3.2613, 3.0672, 1.9873], device='cuda:5'), covar=tensor([0.0479, 0.0052, 0.0051, 0.0395, 0.0098, 0.0075, 0.0080, 0.0449], device='cuda:5'), in_proj_covar=tensor([0.0130, 0.0076, 0.0076, 0.0127, 0.0090, 0.0099, 0.0088, 0.0121], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 18:59:29,081 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6701, 5.0067, 4.7702, 4.7840, 4.5219, 4.4854, 4.3915, 5.0381], device='cuda:5'), covar=tensor([0.1068, 0.0794, 0.0886, 0.0749, 0.0723, 0.1102, 0.1140, 0.0779], device='cuda:5'), in_proj_covar=tensor([0.0618, 0.0755, 0.0623, 0.0564, 0.0473, 0.0489, 0.0633, 0.0588], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:00:28,306 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 19:00:41,596 INFO [train.py:904] (5/8) Epoch 18, batch 8800, loss[loss=0.1767, simple_loss=0.2635, pruned_loss=0.04488, over 12625.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2676, pruned_loss=0.04092, over 3033192.94 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:00:51,216 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.344e+02 2.185e+02 2.552e+02 3.082e+02 8.545e+02, threshold=5.104e+02, percent-clipped=2.0 2023-04-30 19:00:54,109 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181358.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:01:09,011 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3476, 4.4078, 4.2278, 3.9409, 3.9607, 4.3393, 4.0543, 4.0778], device='cuda:5'), covar=tensor([0.0492, 0.0361, 0.0260, 0.0254, 0.0723, 0.0341, 0.0641, 0.0514], device='cuda:5'), in_proj_covar=tensor([0.0270, 0.0383, 0.0315, 0.0304, 0.0324, 0.0355, 0.0218, 0.0375], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:01:28,687 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1780, 4.3628, 4.4633, 4.2828, 4.3471, 4.8205, 4.3867, 4.1628], device='cuda:5'), covar=tensor([0.1566, 0.1498, 0.1389, 0.1879, 0.2208, 0.0864, 0.1377, 0.2097], device='cuda:5'), in_proj_covar=tensor([0.0372, 0.0538, 0.0597, 0.0452, 0.0600, 0.0623, 0.0473, 0.0605], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 19:01:57,983 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:02:27,200 INFO [train.py:904] (5/8) Epoch 18, batch 8850, loss[loss=0.1961, simple_loss=0.2932, pruned_loss=0.04954, over 15279.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2703, pruned_loss=0.04054, over 3036888.01 frames. ], batch size: 190, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:03:44,883 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:03:48,458 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 19:03:52,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0386, 2.6510, 2.9427, 2.0255, 2.6593, 2.0355, 2.7301, 2.8255], device='cuda:5'), covar=tensor([0.0295, 0.0996, 0.0507, 0.1881, 0.0822, 0.1068, 0.0637, 0.0948], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0154, 0.0160, 0.0146, 0.0139, 0.0125, 0.0138, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 19:03:57,674 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 19:04:15,178 INFO [train.py:904] (5/8) Epoch 18, batch 8900, loss[loss=0.1857, simple_loss=0.2759, pruned_loss=0.04773, over 16910.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2711, pruned_loss=0.03982, over 3062011.88 frames. ], batch size: 116, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:04:25,753 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.184e+02 2.557e+02 3.261e+02 8.427e+02, threshold=5.113e+02, percent-clipped=4.0 2023-04-30 19:05:23,229 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:06:18,289 INFO [train.py:904] (5/8) Epoch 18, batch 8950, loss[loss=0.1616, simple_loss=0.2621, pruned_loss=0.03053, over 16771.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2707, pruned_loss=0.04004, over 3074789.37 frames. ], batch size: 83, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:07:17,091 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:07:19,871 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181530.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:08:08,295 INFO [train.py:904] (5/8) Epoch 18, batch 9000, loss[loss=0.1519, simple_loss=0.2422, pruned_loss=0.03082, over 16620.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2674, pruned_loss=0.03898, over 3070465.90 frames. ], batch size: 68, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:08:08,296 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 19:08:17,834 INFO [train.py:938] (5/8) Epoch 18, validation: loss=0.1475, simple_loss=0.2516, pruned_loss=0.02169, over 944034.00 frames. 2023-04-30 19:08:17,835 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 19:08:27,944 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 2.270e+02 2.606e+02 3.112e+02 6.247e+02, threshold=5.212e+02, percent-clipped=1.0 2023-04-30 19:09:06,009 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5598, 4.4926, 4.8736, 4.8455, 4.8562, 4.6087, 4.5701, 4.4697], device='cuda:5'), covar=tensor([0.0283, 0.0636, 0.0425, 0.0444, 0.0406, 0.0398, 0.0733, 0.0380], device='cuda:5'), in_proj_covar=tensor([0.0370, 0.0402, 0.0394, 0.0371, 0.0438, 0.0413, 0.0502, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 19:09:40,032 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181591.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:10:01,201 INFO [train.py:904] (5/8) Epoch 18, batch 9050, loss[loss=0.1734, simple_loss=0.2557, pruned_loss=0.04551, over 16708.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2683, pruned_loss=0.03958, over 3078738.89 frames. ], batch size: 134, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:11:46,776 INFO [train.py:904] (5/8) Epoch 18, batch 9100, loss[loss=0.183, simple_loss=0.2731, pruned_loss=0.04648, over 16163.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2685, pruned_loss=0.04056, over 3078152.20 frames. ], batch size: 35, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:11:50,074 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181653.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:11:55,871 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.401e+02 2.871e+02 3.524e+02 6.480e+02, threshold=5.743e+02, percent-clipped=6.0 2023-04-30 19:12:59,083 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-04-30 19:13:43,588 INFO [train.py:904] (5/8) Epoch 18, batch 9150, loss[loss=0.156, simple_loss=0.2483, pruned_loss=0.03183, over 16674.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2689, pruned_loss=0.04013, over 3068982.01 frames. ], batch size: 57, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:14:55,908 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7540, 2.4494, 2.3599, 3.5545, 2.1065, 3.6944, 1.5303, 2.9499], device='cuda:5'), covar=tensor([0.1335, 0.0755, 0.1145, 0.0167, 0.0107, 0.0359, 0.1642, 0.0665], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0166, 0.0188, 0.0173, 0.0198, 0.0209, 0.0193, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 19:15:27,779 INFO [train.py:904] (5/8) Epoch 18, batch 9200, loss[loss=0.1703, simple_loss=0.2639, pruned_loss=0.03839, over 16205.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2642, pruned_loss=0.03893, over 3084998.47 frames. ], batch size: 165, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:15:36,947 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.118e+02 2.449e+02 2.968e+02 5.097e+02, threshold=4.898e+02, percent-clipped=0.0 2023-04-30 19:15:57,938 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-30 19:16:45,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1683, 4.0387, 4.2570, 4.3857, 4.4817, 4.0187, 4.4410, 4.4947], device='cuda:5'), covar=tensor([0.1707, 0.1072, 0.1310, 0.0635, 0.0541, 0.1362, 0.0600, 0.0570], device='cuda:5'), in_proj_covar=tensor([0.0566, 0.0697, 0.0816, 0.0715, 0.0534, 0.0565, 0.0582, 0.0670], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:17:05,948 INFO [train.py:904] (5/8) Epoch 18, batch 9250, loss[loss=0.1485, simple_loss=0.2356, pruned_loss=0.03076, over 12681.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2635, pruned_loss=0.03879, over 3074965.45 frames. ], batch size: 250, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:18:26,502 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6609, 4.0076, 2.9087, 2.2459, 2.4898, 2.4235, 4.2513, 3.4352], device='cuda:5'), covar=tensor([0.2812, 0.0544, 0.1736, 0.2852, 0.2736, 0.2112, 0.0362, 0.1190], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0254, 0.0289, 0.0294, 0.0278, 0.0241, 0.0278, 0.0313], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:18:36,834 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 19:18:57,431 INFO [train.py:904] (5/8) Epoch 18, batch 9300, loss[loss=0.156, simple_loss=0.2562, pruned_loss=0.02783, over 16713.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2621, pruned_loss=0.03815, over 3070126.99 frames. ], batch size: 83, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:19:06,082 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.346e+02 2.677e+02 3.465e+02 6.012e+02, threshold=5.355e+02, percent-clipped=4.0 2023-04-30 19:20:12,877 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:20:30,346 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6241, 2.4614, 2.2922, 3.7351, 2.1214, 3.8115, 1.3879, 2.7291], device='cuda:5'), covar=tensor([0.1628, 0.0894, 0.1375, 0.0218, 0.0160, 0.0402, 0.1954, 0.0872], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0166, 0.0188, 0.0173, 0.0197, 0.0209, 0.0193, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 19:20:41,436 INFO [train.py:904] (5/8) Epoch 18, batch 9350, loss[loss=0.1892, simple_loss=0.2775, pruned_loss=0.05044, over 16390.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2628, pruned_loss=0.03826, over 3087598.21 frames. ], batch size: 146, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:22:24,038 INFO [train.py:904] (5/8) Epoch 18, batch 9400, loss[loss=0.1529, simple_loss=0.2336, pruned_loss=0.03609, over 12548.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2618, pruned_loss=0.03782, over 3069608.52 frames. ], batch size: 250, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:22:28,150 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:22:33,110 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.111e+02 2.455e+02 3.021e+02 7.291e+02, threshold=4.911e+02, percent-clipped=2.0 2023-04-30 19:23:16,406 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:24:05,248 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182001.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:24:06,042 INFO [train.py:904] (5/8) Epoch 18, batch 9450, loss[loss=0.1694, simple_loss=0.259, pruned_loss=0.03991, over 12378.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2637, pruned_loss=0.03795, over 3066864.73 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:19,196 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182039.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:25:41,977 INFO [train.py:904] (5/8) Epoch 18, batch 9500, loss[loss=0.1797, simple_loss=0.2821, pruned_loss=0.0386, over 16249.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.263, pruned_loss=0.03766, over 3071644.54 frames. ], batch size: 165, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:55,088 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 2.182e+02 2.603e+02 3.134e+02 5.554e+02, threshold=5.207e+02, percent-clipped=1.0 2023-04-30 19:26:15,960 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-30 19:27:27,072 INFO [train.py:904] (5/8) Epoch 18, batch 9550, loss[loss=0.1831, simple_loss=0.27, pruned_loss=0.04814, over 12507.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2626, pruned_loss=0.03777, over 3063776.92 frames. ], batch size: 250, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:28:08,142 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-04-30 19:29:06,831 INFO [train.py:904] (5/8) Epoch 18, batch 9600, loss[loss=0.1897, simple_loss=0.2853, pruned_loss=0.04707, over 16670.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2644, pruned_loss=0.03849, over 3072566.19 frames. ], batch size: 134, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:29:15,381 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.236e+02 2.461e+02 2.915e+02 3.491e+02 6.374e+02, threshold=5.830e+02, percent-clipped=5.0 2023-04-30 19:30:14,210 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182186.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:30:22,189 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 19:30:52,038 INFO [train.py:904] (5/8) Epoch 18, batch 9650, loss[loss=0.1752, simple_loss=0.2614, pruned_loss=0.04449, over 12184.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.266, pruned_loss=0.03889, over 3057549.02 frames. ], batch size: 247, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:00,674 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182234.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:32:37,723 INFO [train.py:904] (5/8) Epoch 18, batch 9700, loss[loss=0.1822, simple_loss=0.2756, pruned_loss=0.0444, over 15207.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2646, pruned_loss=0.03847, over 3056194.19 frames. ], batch size: 190, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:45,795 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.275e+02 2.741e+02 3.246e+02 5.302e+02, threshold=5.483e+02, percent-clipped=0.0 2023-04-30 19:32:58,865 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 19:33:37,412 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:33:44,558 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9341, 2.9582, 2.5120, 2.8383, 3.2750, 3.1065, 3.5609, 3.5454], device='cuda:5'), covar=tensor([0.0090, 0.0359, 0.0451, 0.0368, 0.0249, 0.0333, 0.0229, 0.0197], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0221, 0.0215, 0.0215, 0.0222, 0.0220, 0.0218, 0.0212], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:33:58,386 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3750, 3.7701, 3.7741, 2.2357, 3.2685, 2.6055, 3.7301, 3.7663], device='cuda:5'), covar=tensor([0.0202, 0.0651, 0.0567, 0.1902, 0.0715, 0.0892, 0.0643, 0.0878], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0151, 0.0159, 0.0146, 0.0138, 0.0124, 0.0137, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 19:34:17,889 INFO [train.py:904] (5/8) Epoch 18, batch 9750, loss[loss=0.1524, simple_loss=0.2474, pruned_loss=0.02872, over 16781.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2627, pruned_loss=0.03847, over 3048099.54 frames. ], batch size: 83, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:35:19,737 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9722, 2.0609, 2.5509, 2.9353, 2.7621, 3.2777, 2.1614, 3.3770], device='cuda:5'), covar=tensor([0.0176, 0.0423, 0.0294, 0.0250, 0.0274, 0.0180, 0.0459, 0.0127], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0180, 0.0167, 0.0168, 0.0178, 0.0137, 0.0183, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:35:24,333 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:39,145 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:56,062 INFO [train.py:904] (5/8) Epoch 18, batch 9800, loss[loss=0.164, simple_loss=0.267, pruned_loss=0.0305, over 16886.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2633, pruned_loss=0.03745, over 3068749.04 frames. ], batch size: 96, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:36:05,431 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.961e+02 2.421e+02 2.847e+02 6.201e+02, threshold=4.843e+02, percent-clipped=1.0 2023-04-30 19:37:39,079 INFO [train.py:904] (5/8) Epoch 18, batch 9850, loss[loss=0.174, simple_loss=0.2659, pruned_loss=0.0411, over 16516.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2641, pruned_loss=0.03717, over 3061533.52 frames. ], batch size: 147, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:38:34,423 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9942, 2.7518, 2.9102, 2.1426, 2.6857, 2.1340, 2.6920, 2.9010], device='cuda:5'), covar=tensor([0.0256, 0.0752, 0.0467, 0.1659, 0.0727, 0.0922, 0.0607, 0.0738], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0151, 0.0159, 0.0146, 0.0137, 0.0123, 0.0137, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 19:38:34,447 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4333, 3.0418, 2.7189, 2.2867, 2.1986, 2.2342, 2.9564, 2.8915], device='cuda:5'), covar=tensor([0.2396, 0.0651, 0.1489, 0.2564, 0.2614, 0.2048, 0.0412, 0.1332], device='cuda:5'), in_proj_covar=tensor([0.0309, 0.0252, 0.0287, 0.0291, 0.0274, 0.0239, 0.0275, 0.0310], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:39:04,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0259, 2.3013, 2.3292, 2.8513, 1.8783, 3.3166, 1.6897, 2.8404], device='cuda:5'), covar=tensor([0.1207, 0.0615, 0.1085, 0.0139, 0.0076, 0.0344, 0.1529, 0.0626], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0166, 0.0189, 0.0172, 0.0195, 0.0208, 0.0193, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 19:39:29,936 INFO [train.py:904] (5/8) Epoch 18, batch 9900, loss[loss=0.1918, simple_loss=0.2712, pruned_loss=0.05622, over 12353.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2642, pruned_loss=0.03732, over 3034164.75 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:39:40,682 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.082e+02 2.451e+02 2.864e+02 7.377e+02, threshold=4.903e+02, percent-clipped=2.0 2023-04-30 19:40:34,348 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4271, 2.8618, 3.1888, 1.9577, 2.7521, 2.1154, 3.0139, 3.0355], device='cuda:5'), covar=tensor([0.0259, 0.0868, 0.0511, 0.1952, 0.0817, 0.0978, 0.0638, 0.0981], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0150, 0.0159, 0.0146, 0.0137, 0.0123, 0.0136, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 19:40:50,701 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9247, 2.1865, 1.6535, 1.7910, 2.4657, 2.2490, 2.5594, 2.7616], device='cuda:5'), covar=tensor([0.0176, 0.0485, 0.0655, 0.0596, 0.0307, 0.0459, 0.0223, 0.0289], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0223, 0.0216, 0.0217, 0.0224, 0.0222, 0.0219, 0.0213], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:41:28,678 INFO [train.py:904] (5/8) Epoch 18, batch 9950, loss[loss=0.1632, simple_loss=0.2611, pruned_loss=0.03268, over 16714.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2663, pruned_loss=0.03728, over 3057771.23 frames. ], batch size: 134, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:41:56,059 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9711, 5.2978, 5.1270, 5.0916, 4.8246, 4.7790, 4.7413, 5.3913], device='cuda:5'), covar=tensor([0.1120, 0.0828, 0.0870, 0.0751, 0.0760, 0.0866, 0.1093, 0.0899], device='cuda:5'), in_proj_covar=tensor([0.0606, 0.0744, 0.0607, 0.0553, 0.0467, 0.0479, 0.0617, 0.0576], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:41:58,883 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:43:31,290 INFO [train.py:904] (5/8) Epoch 18, batch 10000, loss[loss=0.1696, simple_loss=0.2704, pruned_loss=0.03438, over 16157.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.265, pruned_loss=0.03687, over 3064258.14 frames. ], batch size: 165, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:43:42,215 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.117e+02 2.346e+02 2.932e+02 5.510e+02, threshold=4.691e+02, percent-clipped=3.0 2023-04-30 19:44:17,872 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:44:33,877 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1502, 4.0364, 4.2354, 4.3521, 4.4662, 4.0079, 4.4409, 4.4909], device='cuda:5'), covar=tensor([0.1646, 0.1090, 0.1361, 0.0723, 0.0544, 0.1199, 0.0606, 0.0625], device='cuda:5'), in_proj_covar=tensor([0.0568, 0.0695, 0.0814, 0.0719, 0.0537, 0.0560, 0.0585, 0.0672], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:44:40,704 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-04-30 19:45:14,038 INFO [train.py:904] (5/8) Epoch 18, batch 10050, loss[loss=0.1764, simple_loss=0.2675, pruned_loss=0.04262, over 12218.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.265, pruned_loss=0.03672, over 3060130.51 frames. ], batch size: 246, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:46:14,965 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:20,793 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182637.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:46,729 INFO [train.py:904] (5/8) Epoch 18, batch 10100, loss[loss=0.154, simple_loss=0.243, pruned_loss=0.03248, over 12868.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2648, pruned_loss=0.03673, over 3054157.67 frames. ], batch size: 247, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:46:55,748 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.131e+02 2.554e+02 3.108e+02 6.507e+02, threshold=5.108e+02, percent-clipped=7.0 2023-04-30 19:47:13,553 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5304, 3.5048, 3.5083, 2.8294, 3.2606, 2.0240, 3.1087, 2.9176], device='cuda:5'), covar=tensor([0.0159, 0.0225, 0.0196, 0.0250, 0.0151, 0.2345, 0.0175, 0.0247], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0135, 0.0176, 0.0158, 0.0154, 0.0191, 0.0167, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:47:30,373 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1011, 3.3108, 3.3999, 1.5854, 3.6304, 3.8155, 2.9456, 2.7085], device='cuda:5'), covar=tensor([0.1095, 0.0214, 0.0231, 0.1418, 0.0081, 0.0129, 0.0432, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0102, 0.0089, 0.0135, 0.0072, 0.0114, 0.0120, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 19:47:44,869 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:47:45,226 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 19:47:59,105 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0825, 2.5365, 2.6243, 1.8728, 2.8112, 2.8480, 2.4660, 2.4583], device='cuda:5'), covar=tensor([0.0651, 0.0256, 0.0220, 0.1012, 0.0099, 0.0222, 0.0438, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0102, 0.0089, 0.0135, 0.0073, 0.0115, 0.0121, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 19:48:03,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7235, 2.6761, 1.9167, 2.8544, 2.1775, 2.8303, 2.1140, 2.4150], device='cuda:5'), covar=tensor([0.0284, 0.0369, 0.1192, 0.0231, 0.0689, 0.0460, 0.1149, 0.0584], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0164, 0.0184, 0.0142, 0.0167, 0.0198, 0.0192, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 19:48:32,142 INFO [train.py:904] (5/8) Epoch 19, batch 0, loss[loss=0.2429, simple_loss=0.3014, pruned_loss=0.09224, over 16895.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3014, pruned_loss=0.09224, over 16895.00 frames. ], batch size: 109, lr: 3.61e-03, grad_scale: 8.0 2023-04-30 19:48:32,142 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 19:48:39,772 INFO [train.py:938] (5/8) Epoch 19, validation: loss=0.1468, simple_loss=0.2504, pruned_loss=0.0216, over 944034.00 frames. 2023-04-30 19:48:39,773 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 19:49:37,173 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7884, 6.1558, 5.8201, 5.9303, 5.4329, 5.3618, 5.4691, 6.2740], device='cuda:5'), covar=tensor([0.1069, 0.0808, 0.1290, 0.0832, 0.0886, 0.0706, 0.1183, 0.0813], device='cuda:5'), in_proj_covar=tensor([0.0612, 0.0750, 0.0615, 0.0557, 0.0471, 0.0483, 0.0626, 0.0580], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:49:50,021 INFO [train.py:904] (5/8) Epoch 19, batch 50, loss[loss=0.1932, simple_loss=0.2742, pruned_loss=0.05611, over 16776.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2689, pruned_loss=0.05014, over 756702.90 frames. ], batch size: 83, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:49:59,030 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.571e+02 3.042e+02 3.790e+02 6.505e+02, threshold=6.085e+02, percent-clipped=6.0 2023-04-30 19:50:55,408 INFO [train.py:904] (5/8) Epoch 19, batch 100, loss[loss=0.1849, simple_loss=0.2599, pruned_loss=0.05496, over 16849.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2693, pruned_loss=0.04898, over 1336512.98 frames. ], batch size: 96, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:52:03,141 INFO [train.py:904] (5/8) Epoch 19, batch 150, loss[loss=0.1899, simple_loss=0.2806, pruned_loss=0.04965, over 16641.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2663, pruned_loss=0.04721, over 1777389.93 frames. ], batch size: 76, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:52:10,236 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182856.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:15,181 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.297e+02 2.720e+02 3.223e+02 8.471e+02, threshold=5.441e+02, percent-clipped=1.0 2023-04-30 19:52:18,037 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:28,189 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:13,460 INFO [train.py:904] (5/8) Epoch 19, batch 200, loss[loss=0.1517, simple_loss=0.239, pruned_loss=0.03217, over 16095.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2675, pruned_loss=0.04829, over 2116471.93 frames. ], batch size: 35, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:53:31,205 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8291, 4.5466, 4.8244, 5.0021, 5.2102, 4.5477, 5.1871, 5.2106], device='cuda:5'), covar=tensor([0.1806, 0.1282, 0.1746, 0.0860, 0.0626, 0.0995, 0.0611, 0.0644], device='cuda:5'), in_proj_covar=tensor([0.0583, 0.0719, 0.0839, 0.0737, 0.0549, 0.0576, 0.0599, 0.0688], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 19:53:34,363 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:42,624 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:54:00,410 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:54:21,118 INFO [train.py:904] (5/8) Epoch 19, batch 250, loss[loss=0.1795, simple_loss=0.2613, pruned_loss=0.04889, over 16863.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2659, pruned_loss=0.04846, over 2379286.74 frames. ], batch size: 116, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:54:32,987 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.306e+02 2.752e+02 3.234e+02 1.372e+03, threshold=5.503e+02, percent-clipped=2.0 2023-04-30 19:55:08,836 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:55:22,694 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 19:55:30,551 INFO [train.py:904] (5/8) Epoch 19, batch 300, loss[loss=0.1747, simple_loss=0.2525, pruned_loss=0.04838, over 16216.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2629, pruned_loss=0.0469, over 2592867.09 frames. ], batch size: 165, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:55:31,645 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3533, 3.3398, 3.5814, 2.5947, 3.3917, 3.6769, 3.4322, 1.8643], device='cuda:5'), covar=tensor([0.0490, 0.0173, 0.0051, 0.0354, 0.0103, 0.0094, 0.0091, 0.0536], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0078, 0.0077, 0.0129, 0.0092, 0.0101, 0.0089, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 19:55:41,314 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 19:55:55,077 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9151, 2.9547, 3.2479, 2.0484, 2.7993, 2.1972, 3.5023, 3.3646], device='cuda:5'), covar=tensor([0.0235, 0.0955, 0.0607, 0.1953, 0.0839, 0.1053, 0.0520, 0.1002], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0153, 0.0161, 0.0148, 0.0139, 0.0125, 0.0138, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 19:56:41,138 INFO [train.py:904] (5/8) Epoch 19, batch 350, loss[loss=0.1539, simple_loss=0.2311, pruned_loss=0.03834, over 16977.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2607, pruned_loss=0.04573, over 2739787.96 frames. ], batch size: 41, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:52,342 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.344e+02 2.193e+02 2.670e+02 3.333e+02 1.052e+03, threshold=5.340e+02, percent-clipped=4.0 2023-04-30 19:57:51,266 INFO [train.py:904] (5/8) Epoch 19, batch 400, loss[loss=0.1472, simple_loss=0.2405, pruned_loss=0.02697, over 17047.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2595, pruned_loss=0.04505, over 2865946.62 frames. ], batch size: 50, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:58:39,107 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9035, 4.9118, 5.3837, 5.3385, 5.3611, 5.0435, 4.9791, 4.7850], device='cuda:5'), covar=tensor([0.0367, 0.0529, 0.0354, 0.0432, 0.0477, 0.0399, 0.0896, 0.0439], device='cuda:5'), in_proj_covar=tensor([0.0386, 0.0421, 0.0410, 0.0388, 0.0454, 0.0432, 0.0521, 0.0346], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 19:59:03,158 INFO [train.py:904] (5/8) Epoch 19, batch 450, loss[loss=0.1549, simple_loss=0.2323, pruned_loss=0.03879, over 15563.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2575, pruned_loss=0.04439, over 2965659.67 frames. ], batch size: 190, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:59:14,100 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.162e+02 2.492e+02 3.004e+02 8.634e+02, threshold=4.984e+02, percent-clipped=1.0 2023-04-30 19:59:28,199 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:12,290 INFO [train.py:904] (5/8) Epoch 19, batch 500, loss[loss=0.1527, simple_loss=0.2509, pruned_loss=0.02727, over 17036.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2562, pruned_loss=0.04359, over 3047322.96 frames. ], batch size: 50, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:00:28,373 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:31,370 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 20:00:35,917 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,939 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:47,706 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3499, 3.4461, 3.6785, 2.6097, 3.3434, 3.7413, 3.4714, 2.2057], device='cuda:5'), covar=tensor([0.0538, 0.0123, 0.0055, 0.0368, 0.0112, 0.0093, 0.0094, 0.0456], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0079, 0.0079, 0.0130, 0.0093, 0.0103, 0.0090, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 20:01:03,108 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5105, 2.3070, 2.3802, 4.2520, 2.2462, 2.6672, 2.3718, 2.4547], device='cuda:5'), covar=tensor([0.1143, 0.3650, 0.2935, 0.0506, 0.4182, 0.2649, 0.3421, 0.3790], device='cuda:5'), in_proj_covar=tensor([0.0389, 0.0433, 0.0359, 0.0321, 0.0432, 0.0494, 0.0403, 0.0504], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:01:23,248 INFO [train.py:904] (5/8) Epoch 19, batch 550, loss[loss=0.1686, simple_loss=0.2453, pruned_loss=0.04602, over 16694.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2552, pruned_loss=0.04326, over 3106441.49 frames. ], batch size: 89, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:01:34,920 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.167e+02 2.560e+02 2.885e+02 5.610e+02, threshold=5.120e+02, percent-clipped=1.0 2023-04-30 20:02:32,742 INFO [train.py:904] (5/8) Epoch 19, batch 600, loss[loss=0.1486, simple_loss=0.2333, pruned_loss=0.03192, over 16767.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2541, pruned_loss=0.04287, over 3163946.49 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:19,843 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4929, 4.4648, 4.4516, 3.9169, 4.4346, 1.8619, 4.2319, 4.1238], device='cuda:5'), covar=tensor([0.0133, 0.0109, 0.0183, 0.0323, 0.0116, 0.2556, 0.0160, 0.0226], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0142, 0.0185, 0.0168, 0.0162, 0.0199, 0.0176, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:03:42,747 INFO [train.py:904] (5/8) Epoch 19, batch 650, loss[loss=0.179, simple_loss=0.2522, pruned_loss=0.05286, over 15560.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2531, pruned_loss=0.04283, over 3202527.87 frames. ], batch size: 190, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:54,612 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.241e+02 2.597e+02 3.196e+02 6.464e+02, threshold=5.194e+02, percent-clipped=2.0 2023-04-30 20:04:10,606 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6581, 6.0252, 5.7763, 5.7986, 5.4351, 5.4342, 5.3962, 6.1382], device='cuda:5'), covar=tensor([0.1255, 0.0793, 0.1009, 0.0806, 0.0924, 0.0684, 0.1137, 0.0915], device='cuda:5'), in_proj_covar=tensor([0.0654, 0.0799, 0.0658, 0.0593, 0.0501, 0.0512, 0.0665, 0.0617], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:04:23,196 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 20:04:24,136 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8031, 2.9477, 2.7622, 5.0187, 4.1278, 4.4937, 1.6434, 3.2562], device='cuda:5'), covar=tensor([0.1378, 0.0748, 0.1149, 0.0180, 0.0241, 0.0383, 0.1585, 0.0756], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0180, 0.0199, 0.0213, 0.0195, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:04:53,086 INFO [train.py:904] (5/8) Epoch 19, batch 700, loss[loss=0.1787, simple_loss=0.2536, pruned_loss=0.05188, over 16836.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2535, pruned_loss=0.04297, over 3228967.53 frames. ], batch size: 96, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:05:01,051 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:05:08,495 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 20:05:59,961 INFO [train.py:904] (5/8) Epoch 19, batch 750, loss[loss=0.1571, simple_loss=0.2373, pruned_loss=0.03841, over 16919.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2535, pruned_loss=0.04262, over 3248889.61 frames. ], batch size: 90, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:06:11,900 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.150e+02 2.519e+02 3.057e+02 7.230e+02, threshold=5.039e+02, percent-clipped=2.0 2023-04-30 20:06:25,274 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183469.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:07:10,060 INFO [train.py:904] (5/8) Epoch 19, batch 800, loss[loss=0.1801, simple_loss=0.2578, pruned_loss=0.05123, over 12394.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2529, pruned_loss=0.04248, over 3261608.72 frames. ], batch size: 246, lr: 3.61e-03, grad_scale: 4.0 2023-04-30 20:07:24,211 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:07:32,577 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:01,555 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3713, 3.1821, 3.4836, 1.9072, 3.5930, 3.5838, 2.9955, 2.7443], device='cuda:5'), covar=tensor([0.0781, 0.0238, 0.0187, 0.1089, 0.0096, 0.0206, 0.0363, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0106, 0.0095, 0.0139, 0.0077, 0.0122, 0.0126, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 20:08:19,002 INFO [train.py:904] (5/8) Epoch 19, batch 850, loss[loss=0.1601, simple_loss=0.2396, pruned_loss=0.04026, over 16429.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2528, pruned_loss=0.04197, over 3280253.40 frames. ], batch size: 75, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:08:29,790 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.149e+02 2.547e+02 3.022e+02 7.911e+02, threshold=5.094e+02, percent-clipped=1.0 2023-04-30 20:08:30,867 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183560.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:38,499 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:09:28,469 INFO [train.py:904] (5/8) Epoch 19, batch 900, loss[loss=0.1543, simple_loss=0.2413, pruned_loss=0.03369, over 17212.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2525, pruned_loss=0.04159, over 3289263.83 frames. ], batch size: 44, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:09:44,221 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:10:01,036 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9319, 1.9973, 2.5280, 2.9439, 2.6555, 3.4225, 2.2759, 3.3596], device='cuda:5'), covar=tensor([0.0214, 0.0498, 0.0327, 0.0316, 0.0334, 0.0169, 0.0468, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0190, 0.0178, 0.0179, 0.0189, 0.0147, 0.0193, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:10:35,714 INFO [train.py:904] (5/8) Epoch 19, batch 950, loss[loss=0.1683, simple_loss=0.2478, pruned_loss=0.04444, over 15377.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2528, pruned_loss=0.04172, over 3292748.26 frames. ], batch size: 190, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:10:45,431 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 20:10:45,889 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 1.990e+02 2.288e+02 2.656e+02 5.552e+02, threshold=4.575e+02, percent-clipped=1.0 2023-04-30 20:11:08,284 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183675.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:11:42,842 INFO [train.py:904] (5/8) Epoch 19, batch 1000, loss[loss=0.1536, simple_loss=0.247, pruned_loss=0.03013, over 17210.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2518, pruned_loss=0.0414, over 3293368.14 frames. ], batch size: 46, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:12:50,606 INFO [train.py:904] (5/8) Epoch 19, batch 1050, loss[loss=0.1555, simple_loss=0.2382, pruned_loss=0.03637, over 16829.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2515, pruned_loss=0.04087, over 3304828.67 frames. ], batch size: 42, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:13:01,772 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.214e+02 2.595e+02 3.147e+02 6.979e+02, threshold=5.191e+02, percent-clipped=4.0 2023-04-30 20:13:06,525 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:14:00,542 INFO [train.py:904] (5/8) Epoch 19, batch 1100, loss[loss=0.1448, simple_loss=0.2322, pruned_loss=0.02872, over 17241.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2501, pruned_loss=0.04046, over 3308940.36 frames. ], batch size: 45, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:14:47,551 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3561, 3.0899, 3.3207, 1.9475, 3.4509, 3.4374, 2.8447, 2.6254], device='cuda:5'), covar=tensor([0.0774, 0.0247, 0.0228, 0.1084, 0.0113, 0.0238, 0.0437, 0.0456], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0107, 0.0096, 0.0140, 0.0077, 0.0123, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 20:14:51,593 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0907, 2.1754, 2.6840, 2.9699, 2.7397, 3.5330, 2.4256, 3.4824], device='cuda:5'), covar=tensor([0.0221, 0.0482, 0.0317, 0.0331, 0.0332, 0.0180, 0.0453, 0.0178], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0190, 0.0178, 0.0179, 0.0190, 0.0147, 0.0193, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:14:55,840 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183842.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:15:08,304 INFO [train.py:904] (5/8) Epoch 19, batch 1150, loss[loss=0.1601, simple_loss=0.2433, pruned_loss=0.03845, over 16447.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2503, pruned_loss=0.0402, over 3309165.91 frames. ], batch size: 75, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:15:20,305 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 2.057e+02 2.490e+02 3.144e+02 5.373e+02, threshold=4.980e+02, percent-clipped=1.0 2023-04-30 20:15:26,862 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0287, 1.9918, 2.5666, 2.8790, 2.7445, 3.0611, 2.1015, 3.2030], device='cuda:5'), covar=tensor([0.0197, 0.0483, 0.0341, 0.0257, 0.0326, 0.0240, 0.0511, 0.0166], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0190, 0.0177, 0.0178, 0.0189, 0.0146, 0.0192, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:16:18,544 INFO [train.py:904] (5/8) Epoch 19, batch 1200, loss[loss=0.1441, simple_loss=0.2345, pruned_loss=0.02683, over 16797.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2499, pruned_loss=0.03988, over 3318386.13 frames. ], batch size: 42, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:16:20,039 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183903.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:16:43,964 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6010, 3.6423, 4.1238, 2.3502, 3.3098, 2.6796, 4.1163, 3.8501], device='cuda:5'), covar=tensor([0.0231, 0.0883, 0.0421, 0.1809, 0.0718, 0.0904, 0.0518, 0.1045], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0159, 0.0165, 0.0150, 0.0142, 0.0127, 0.0143, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:17:25,282 INFO [train.py:904] (5/8) Epoch 19, batch 1250, loss[loss=0.1838, simple_loss=0.2556, pruned_loss=0.05597, over 16771.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2504, pruned_loss=0.04071, over 3315864.59 frames. ], batch size: 124, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:17:35,898 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.248e+02 2.515e+02 3.055e+02 4.782e+02, threshold=5.031e+02, percent-clipped=0.0 2023-04-30 20:17:38,220 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4482, 5.4155, 5.2595, 4.7208, 4.8415, 5.3001, 5.2424, 4.8864], device='cuda:5'), covar=tensor([0.0565, 0.0432, 0.0291, 0.0365, 0.1133, 0.0439, 0.0295, 0.0760], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0415, 0.0343, 0.0331, 0.0351, 0.0386, 0.0234, 0.0405], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:17:47,510 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9321, 5.2733, 5.0394, 5.0220, 4.7732, 4.7005, 4.7111, 5.3361], device='cuda:5'), covar=tensor([0.1208, 0.0819, 0.1083, 0.0938, 0.0907, 0.1104, 0.1148, 0.0969], device='cuda:5'), in_proj_covar=tensor([0.0662, 0.0812, 0.0667, 0.0604, 0.0508, 0.0519, 0.0676, 0.0631], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:17:50,870 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183970.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:18:37,503 INFO [train.py:904] (5/8) Epoch 19, batch 1300, loss[loss=0.1879, simple_loss=0.2664, pruned_loss=0.05466, over 16490.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2503, pruned_loss=0.04127, over 3310543.89 frames. ], batch size: 68, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:18:39,268 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8635, 2.9767, 3.2118, 2.0541, 2.7468, 2.2341, 3.3377, 3.3184], device='cuda:5'), covar=tensor([0.0259, 0.0943, 0.0569, 0.1939, 0.0898, 0.1096, 0.0600, 0.0964], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:19:44,272 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:19:46,322 INFO [train.py:904] (5/8) Epoch 19, batch 1350, loss[loss=0.1421, simple_loss=0.229, pruned_loss=0.02759, over 16991.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2496, pruned_loss=0.04091, over 3292825.64 frames. ], batch size: 41, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:19:55,437 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.201e+02 2.541e+02 3.098e+02 6.218e+02, threshold=5.081e+02, percent-clipped=6.0 2023-04-30 20:20:03,862 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:20:55,161 INFO [train.py:904] (5/8) Epoch 19, batch 1400, loss[loss=0.1635, simple_loss=0.2587, pruned_loss=0.0342, over 17117.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2498, pruned_loss=0.04077, over 3299813.46 frames. ], batch size: 48, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:21:09,636 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:10,487 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184112.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:31,585 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 20:22:05,089 INFO [train.py:904] (5/8) Epoch 19, batch 1450, loss[loss=0.1527, simple_loss=0.2422, pruned_loss=0.03163, over 17224.00 frames. ], tot_loss[loss=0.165, simple_loss=0.249, pruned_loss=0.04051, over 3313895.87 frames. ], batch size: 44, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:22:15,262 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 20:22:15,459 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.293e+02 2.595e+02 3.219e+02 5.738e+02, threshold=5.191e+02, percent-clipped=1.0 2023-04-30 20:22:30,865 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-04-30 20:23:08,685 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184198.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:23:13,289 INFO [train.py:904] (5/8) Epoch 19, batch 1500, loss[loss=0.1664, simple_loss=0.2451, pruned_loss=0.04383, over 16529.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2489, pruned_loss=0.04017, over 3318422.32 frames. ], batch size: 75, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:24:11,352 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2452, 2.1183, 2.4214, 4.0376, 2.1731, 2.4157, 2.2187, 2.2889], device='cuda:5'), covar=tensor([0.1549, 0.4010, 0.2793, 0.0638, 0.4435, 0.2973, 0.3868, 0.3627], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0439, 0.0364, 0.0328, 0.0436, 0.0504, 0.0409, 0.0512], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:24:22,400 INFO [train.py:904] (5/8) Epoch 19, batch 1550, loss[loss=0.1586, simple_loss=0.2549, pruned_loss=0.03114, over 17053.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2508, pruned_loss=0.04129, over 3315542.52 frames. ], batch size: 53, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:24:34,828 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.353e+02 2.741e+02 3.128e+02 4.649e+02, threshold=5.482e+02, percent-clipped=0.0 2023-04-30 20:24:47,543 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:25:22,498 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9519, 4.3808, 4.4300, 3.2294, 3.7158, 4.3323, 3.8966, 2.5356], device='cuda:5'), covar=tensor([0.0456, 0.0065, 0.0043, 0.0342, 0.0124, 0.0091, 0.0077, 0.0432], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0082, 0.0080, 0.0133, 0.0096, 0.0105, 0.0092, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 20:25:31,174 INFO [train.py:904] (5/8) Epoch 19, batch 1600, loss[loss=0.1578, simple_loss=0.2519, pruned_loss=0.03188, over 17147.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2524, pruned_loss=0.04176, over 3322154.10 frames. ], batch size: 48, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:25:53,347 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:23,712 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0165, 2.0849, 2.6530, 2.9854, 2.8199, 3.5576, 2.2714, 3.4355], device='cuda:5'), covar=tensor([0.0231, 0.0511, 0.0330, 0.0330, 0.0325, 0.0166, 0.0494, 0.0191], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0192, 0.0180, 0.0181, 0.0191, 0.0148, 0.0194, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:26:39,334 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7567, 3.8481, 4.2320, 2.5158, 3.4183, 2.9450, 4.2325, 4.0957], device='cuda:5'), covar=tensor([0.0227, 0.0884, 0.0466, 0.1803, 0.0730, 0.0816, 0.0516, 0.0965], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0159, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:26:39,960 INFO [train.py:904] (5/8) Epoch 19, batch 1650, loss[loss=0.1476, simple_loss=0.2451, pruned_loss=0.02505, over 15961.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2539, pruned_loss=0.04195, over 3308601.16 frames. ], batch size: 35, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:26:50,282 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:52,313 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.220e+02 2.674e+02 3.187e+02 5.456e+02, threshold=5.348e+02, percent-clipped=0.0 2023-04-30 20:27:49,670 INFO [train.py:904] (5/8) Epoch 19, batch 1700, loss[loss=0.1763, simple_loss=0.2531, pruned_loss=0.04979, over 16879.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2553, pruned_loss=0.04258, over 3304698.18 frames. ], batch size: 116, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:27:55,029 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:14,023 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:58,465 INFO [train.py:904] (5/8) Epoch 19, batch 1750, loss[loss=0.2411, simple_loss=0.3178, pruned_loss=0.08223, over 12546.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2566, pruned_loss=0.04267, over 3307785.40 frames. ], batch size: 246, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:29:10,990 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.342e+02 2.772e+02 3.206e+02 7.183e+02, threshold=5.544e+02, percent-clipped=2.0 2023-04-30 20:29:40,719 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-30 20:29:45,783 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184486.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:30:02,750 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:30:07,544 INFO [train.py:904] (5/8) Epoch 19, batch 1800, loss[loss=0.1661, simple_loss=0.2582, pruned_loss=0.03699, over 17074.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2575, pruned_loss=0.04317, over 3310674.69 frames. ], batch size: 55, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:31:07,261 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184546.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:09,236 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184547.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:16,694 INFO [train.py:904] (5/8) Epoch 19, batch 1850, loss[loss=0.2271, simple_loss=0.305, pruned_loss=0.07457, over 11927.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.259, pruned_loss=0.04337, over 3315460.39 frames. ], batch size: 246, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:31:20,777 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 20:31:29,856 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.319e+02 2.167e+02 2.563e+02 3.002e+02 5.511e+02, threshold=5.126e+02, percent-clipped=0.0 2023-04-30 20:32:02,619 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 20:32:26,194 INFO [train.py:904] (5/8) Epoch 19, batch 1900, loss[loss=0.1661, simple_loss=0.2382, pruned_loss=0.04698, over 16908.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2588, pruned_loss=0.04353, over 3317353.53 frames. ], batch size: 90, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:33:35,928 INFO [train.py:904] (5/8) Epoch 19, batch 1950, loss[loss=0.1702, simple_loss=0.2523, pruned_loss=0.04401, over 16776.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2575, pruned_loss=0.04258, over 3317580.67 frames. ], batch size: 83, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:33:48,871 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.129e+02 2.542e+02 2.951e+02 4.781e+02, threshold=5.085e+02, percent-clipped=0.0 2023-04-30 20:34:00,690 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2383, 5.2122, 5.6825, 5.6667, 5.6811, 5.3178, 5.2605, 5.1345], device='cuda:5'), covar=tensor([0.0305, 0.0523, 0.0321, 0.0394, 0.0437, 0.0372, 0.0939, 0.0396], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0443, 0.0431, 0.0405, 0.0478, 0.0455, 0.0547, 0.0361], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 20:34:45,276 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0088, 2.1360, 2.6612, 2.9655, 2.7767, 3.5475, 2.3246, 3.5031], device='cuda:5'), covar=tensor([0.0241, 0.0486, 0.0318, 0.0320, 0.0316, 0.0173, 0.0477, 0.0142], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0192, 0.0179, 0.0182, 0.0191, 0.0149, 0.0195, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:34:47,053 INFO [train.py:904] (5/8) Epoch 19, batch 2000, loss[loss=0.1751, simple_loss=0.2446, pruned_loss=0.05284, over 16753.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2571, pruned_loss=0.04235, over 3312447.35 frames. ], batch size: 102, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:34:47,404 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3833, 3.3322, 3.4203, 3.5062, 3.5678, 3.2733, 3.4514, 3.6193], device='cuda:5'), covar=tensor([0.1208, 0.0985, 0.1043, 0.0670, 0.0610, 0.2507, 0.1576, 0.0712], device='cuda:5'), in_proj_covar=tensor([0.0642, 0.0793, 0.0928, 0.0807, 0.0601, 0.0632, 0.0650, 0.0754], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:34:47,429 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5512, 3.6234, 3.3246, 2.9670, 3.2078, 3.4479, 3.3251, 3.3103], device='cuda:5'), covar=tensor([0.0689, 0.0627, 0.0310, 0.0253, 0.0565, 0.0443, 0.1140, 0.0476], device='cuda:5'), in_proj_covar=tensor([0.0297, 0.0422, 0.0347, 0.0337, 0.0357, 0.0393, 0.0237, 0.0414], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:34:53,174 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184706.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:06,430 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184715.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:26,031 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2002, 4.2590, 4.6175, 4.6208, 4.6336, 4.2630, 4.3264, 4.1728], device='cuda:5'), covar=tensor([0.0386, 0.0756, 0.0449, 0.0463, 0.0490, 0.0494, 0.0899, 0.0707], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0441, 0.0428, 0.0404, 0.0476, 0.0454, 0.0545, 0.0359], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 20:35:48,055 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 20:35:56,928 INFO [train.py:904] (5/8) Epoch 19, batch 2050, loss[loss=0.1874, simple_loss=0.2626, pruned_loss=0.05607, over 16898.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2574, pruned_loss=0.04304, over 3305288.88 frames. ], batch size: 109, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:35:59,934 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184754.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:36:10,018 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.173e+02 2.577e+02 2.933e+02 5.913e+02, threshold=5.153e+02, percent-clipped=1.0 2023-04-30 20:37:10,296 INFO [train.py:904] (5/8) Epoch 19, batch 2100, loss[loss=0.1696, simple_loss=0.2539, pruned_loss=0.04265, over 16896.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2587, pruned_loss=0.04353, over 3302582.52 frames. ], batch size: 102, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:37:25,613 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1601, 3.9631, 4.2730, 2.1510, 4.3882, 4.6397, 3.4816, 3.5440], device='cuda:5'), covar=tensor([0.0704, 0.0277, 0.0275, 0.1232, 0.0132, 0.0170, 0.0407, 0.0401], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0108, 0.0097, 0.0141, 0.0079, 0.0125, 0.0128, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:38:05,011 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184842.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:38:18,978 INFO [train.py:904] (5/8) Epoch 19, batch 2150, loss[loss=0.2024, simple_loss=0.2735, pruned_loss=0.06564, over 16884.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2596, pruned_loss=0.04397, over 3298585.18 frames. ], batch size: 116, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:38:30,744 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.154e+02 2.741e+02 3.179e+02 7.758e+02, threshold=5.482e+02, percent-clipped=2.0 2023-04-30 20:38:38,235 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184866.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:38:54,788 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 20:39:03,568 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 20:39:27,111 INFO [train.py:904] (5/8) Epoch 19, batch 2200, loss[loss=0.1475, simple_loss=0.2269, pruned_loss=0.03403, over 16798.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2593, pruned_loss=0.04353, over 3309582.84 frames. ], batch size: 39, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:01,229 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:17,879 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:35,954 INFO [train.py:904] (5/8) Epoch 19, batch 2250, loss[loss=0.1861, simple_loss=0.2602, pruned_loss=0.05599, over 16738.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2606, pruned_loss=0.04463, over 3307426.64 frames. ], batch size: 134, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:48,430 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.330e+02 2.636e+02 3.046e+02 4.750e+02, threshold=5.271e+02, percent-clipped=0.0 2023-04-30 20:41:45,032 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185000.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:41:46,805 INFO [train.py:904] (5/8) Epoch 19, batch 2300, loss[loss=0.1637, simple_loss=0.257, pruned_loss=0.03516, over 17097.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2597, pruned_loss=0.044, over 3310158.71 frames. ], batch size: 48, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:42:05,871 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:42:09,453 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3133, 5.2877, 5.0047, 4.5503, 5.0905, 1.9361, 4.8567, 4.9474], device='cuda:5'), covar=tensor([0.0077, 0.0068, 0.0198, 0.0353, 0.0109, 0.2689, 0.0132, 0.0200], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0152, 0.0196, 0.0179, 0.0173, 0.0208, 0.0188, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:42:56,748 INFO [train.py:904] (5/8) Epoch 19, batch 2350, loss[loss=0.1561, simple_loss=0.2446, pruned_loss=0.03378, over 16795.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2594, pruned_loss=0.04368, over 3320913.31 frames. ], batch size: 39, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:43:08,970 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:43:09,640 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.245e+02 2.599e+02 3.315e+02 5.469e+02, threshold=5.198e+02, percent-clipped=1.0 2023-04-30 20:43:12,100 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:44:00,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0151, 4.1736, 2.8057, 4.8621, 3.2343, 4.7713, 3.0043, 3.5324], device='cuda:5'), covar=tensor([0.0274, 0.0325, 0.1443, 0.0286, 0.0811, 0.0467, 0.1308, 0.0655], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0178, 0.0196, 0.0163, 0.0177, 0.0218, 0.0203, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:44:06,547 INFO [train.py:904] (5/8) Epoch 19, batch 2400, loss[loss=0.1871, simple_loss=0.289, pruned_loss=0.04265, over 16744.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2613, pruned_loss=0.04404, over 3313366.15 frames. ], batch size: 57, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:44:33,434 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185121.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 20:44:43,165 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:45:02,139 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:45:15,143 INFO [train.py:904] (5/8) Epoch 19, batch 2450, loss[loss=0.1916, simple_loss=0.2778, pruned_loss=0.05271, over 15440.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2622, pruned_loss=0.04343, over 3319506.73 frames. ], batch size: 190, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:45:16,600 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9960, 5.3549, 5.5108, 5.2551, 5.2796, 5.9163, 5.3835, 5.0896], device='cuda:5'), covar=tensor([0.1077, 0.2058, 0.2473, 0.1931, 0.2854, 0.0991, 0.1540, 0.2348], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0593, 0.0655, 0.0495, 0.0665, 0.0695, 0.0514, 0.0665], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 20:45:27,052 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 2.259e+02 2.760e+02 3.178e+02 5.977e+02, threshold=5.520e+02, percent-clipped=3.0 2023-04-30 20:46:04,368 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0402, 3.1408, 3.3061, 2.1325, 3.0917, 3.4173, 3.0658, 1.8449], device='cuda:5'), covar=tensor([0.0518, 0.0146, 0.0066, 0.0410, 0.0126, 0.0086, 0.0102, 0.0488], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0082, 0.0082, 0.0134, 0.0096, 0.0106, 0.0093, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 20:46:08,234 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:08,400 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:23,978 INFO [train.py:904] (5/8) Epoch 19, batch 2500, loss[loss=0.1925, simple_loss=0.2873, pruned_loss=0.04883, over 16692.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2623, pruned_loss=0.04391, over 3325015.55 frames. ], batch size: 62, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:46:51,991 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185222.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:59,999 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7434, 2.9481, 3.1351, 2.0843, 2.7681, 2.1724, 3.2581, 3.3128], device='cuda:5'), covar=tensor([0.0240, 0.0904, 0.0576, 0.1778, 0.0839, 0.0956, 0.0568, 0.0826], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0161, 0.0165, 0.0151, 0.0142, 0.0127, 0.0142, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:47:16,352 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9631, 4.3357, 4.2780, 3.1315, 3.6842, 4.2487, 3.9152, 2.0933], device='cuda:5'), covar=tensor([0.0494, 0.0078, 0.0067, 0.0399, 0.0168, 0.0130, 0.0113, 0.0670], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0134, 0.0096, 0.0106, 0.0093, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 20:47:32,255 INFO [train.py:904] (5/8) Epoch 19, batch 2550, loss[loss=0.1713, simple_loss=0.2683, pruned_loss=0.03714, over 16675.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2625, pruned_loss=0.04411, over 3319986.94 frames. ], batch size: 62, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:47:44,040 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.145e+02 2.534e+02 2.988e+02 8.366e+02, threshold=5.067e+02, percent-clipped=1.0 2023-04-30 20:48:17,929 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7230, 6.0316, 5.7679, 5.9206, 5.4922, 5.4378, 5.5414, 6.2249], device='cuda:5'), covar=tensor([0.1290, 0.0923, 0.1293, 0.0866, 0.0919, 0.0695, 0.1092, 0.1005], device='cuda:5'), in_proj_covar=tensor([0.0669, 0.0823, 0.0679, 0.0610, 0.0515, 0.0523, 0.0684, 0.0637], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:48:20,240 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9292, 2.0826, 2.2246, 3.4455, 2.0644, 2.3212, 2.2172, 2.2101], device='cuda:5'), covar=tensor([0.1487, 0.3604, 0.2798, 0.0730, 0.3972, 0.2657, 0.3538, 0.3324], device='cuda:5'), in_proj_covar=tensor([0.0396, 0.0439, 0.0362, 0.0327, 0.0434, 0.0505, 0.0408, 0.0512], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:48:31,910 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:48:36,383 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 20:48:40,161 INFO [train.py:904] (5/8) Epoch 19, batch 2600, loss[loss=0.1428, simple_loss=0.229, pruned_loss=0.02837, over 17216.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2616, pruned_loss=0.04362, over 3325392.46 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:49:06,395 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4761, 3.5781, 2.1791, 3.7432, 2.7057, 3.6592, 2.1970, 2.8311], device='cuda:5'), covar=tensor([0.0243, 0.0392, 0.1499, 0.0308, 0.0787, 0.0855, 0.1535, 0.0703], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0178, 0.0196, 0.0164, 0.0177, 0.0219, 0.0203, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:49:15,758 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 20:49:50,078 INFO [train.py:904] (5/8) Epoch 19, batch 2650, loss[loss=0.1812, simple_loss=0.2717, pruned_loss=0.04533, over 17069.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2614, pruned_loss=0.04312, over 3327043.18 frames. ], batch size: 53, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:50:03,485 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.113e+02 2.462e+02 3.036e+02 8.000e+02, threshold=4.924e+02, percent-clipped=5.0 2023-04-30 20:50:59,019 INFO [train.py:904] (5/8) Epoch 19, batch 2700, loss[loss=0.1675, simple_loss=0.2596, pruned_loss=0.03768, over 17243.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2619, pruned_loss=0.04289, over 3322865.03 frames. ], batch size: 45, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:51:04,744 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4990, 3.5513, 2.1294, 3.7442, 2.7949, 3.7089, 2.3027, 2.8431], device='cuda:5'), covar=tensor([0.0258, 0.0393, 0.1598, 0.0310, 0.0696, 0.0826, 0.1361, 0.0748], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0178, 0.0195, 0.0163, 0.0176, 0.0219, 0.0203, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:51:19,155 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185416.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 20:52:08,652 INFO [train.py:904] (5/8) Epoch 19, batch 2750, loss[loss=0.1853, simple_loss=0.2575, pruned_loss=0.05658, over 16799.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2626, pruned_loss=0.04317, over 3330893.32 frames. ], batch size: 124, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:52:20,529 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.496e+02 2.075e+02 2.462e+02 3.061e+02 7.702e+02, threshold=4.923e+02, percent-clipped=3.0 2023-04-30 20:52:30,906 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:52:37,909 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 20:52:54,431 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185485.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:52:57,887 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4844, 4.3000, 4.5085, 4.6914, 4.8236, 4.3379, 4.7042, 4.7891], device='cuda:5'), covar=tensor([0.1659, 0.1288, 0.1414, 0.0732, 0.0595, 0.1038, 0.1745, 0.0744], device='cuda:5'), in_proj_covar=tensor([0.0646, 0.0800, 0.0938, 0.0815, 0.0606, 0.0640, 0.0654, 0.0764], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:53:18,344 INFO [train.py:904] (5/8) Epoch 19, batch 2800, loss[loss=0.1835, simple_loss=0.2666, pruned_loss=0.0502, over 15495.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2619, pruned_loss=0.04302, over 3339794.11 frames. ], batch size: 191, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:53:46,712 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185522.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:53:56,364 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:54:03,955 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5073, 2.3222, 2.2890, 4.3250, 2.2676, 2.7611, 2.4313, 2.3918], device='cuda:5'), covar=tensor([0.1197, 0.3658, 0.3023, 0.0490, 0.4246, 0.2514, 0.3350, 0.3787], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0440, 0.0362, 0.0328, 0.0435, 0.0506, 0.0409, 0.0514], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 20:54:28,309 INFO [train.py:904] (5/8) Epoch 19, batch 2850, loss[loss=0.1684, simple_loss=0.2444, pruned_loss=0.04624, over 16741.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2608, pruned_loss=0.04275, over 3337065.49 frames. ], batch size: 102, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:54:41,479 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.166e+02 2.620e+02 3.152e+02 5.145e+02, threshold=5.240e+02, percent-clipped=1.0 2023-04-30 20:54:53,469 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:28,858 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185595.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:39,059 INFO [train.py:904] (5/8) Epoch 19, batch 2900, loss[loss=0.157, simple_loss=0.2561, pruned_loss=0.02896, over 17129.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2596, pruned_loss=0.04252, over 3345847.08 frames. ], batch size: 48, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:56:36,427 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:56:36,674 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3515, 3.7507, 3.9195, 2.2809, 3.0399, 2.6133, 3.8067, 3.9549], device='cuda:5'), covar=tensor([0.0286, 0.0784, 0.0505, 0.1874, 0.0874, 0.0956, 0.0624, 0.0933], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0160, 0.0164, 0.0150, 0.0142, 0.0126, 0.0141, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:56:47,757 INFO [train.py:904] (5/8) Epoch 19, batch 2950, loss[loss=0.1528, simple_loss=0.244, pruned_loss=0.03082, over 17207.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2591, pruned_loss=0.04322, over 3341172.70 frames. ], batch size: 45, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:56:56,294 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7556, 3.9720, 2.5939, 4.5939, 3.1244, 4.5525, 2.7110, 3.1479], device='cuda:5'), covar=tensor([0.0342, 0.0369, 0.1554, 0.0312, 0.0832, 0.0546, 0.1423, 0.0807], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0179, 0.0195, 0.0164, 0.0177, 0.0220, 0.0203, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 20:57:00,939 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.308e+02 2.753e+02 3.431e+02 9.141e+02, threshold=5.506e+02, percent-clipped=4.0 2023-04-30 20:57:43,699 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7345, 4.7964, 4.9649, 4.7701, 4.8410, 5.4415, 4.9507, 4.5958], device='cuda:5'), covar=tensor([0.1404, 0.2144, 0.2271, 0.2278, 0.2590, 0.1035, 0.1665, 0.2749], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0591, 0.0653, 0.0494, 0.0662, 0.0689, 0.0508, 0.0659], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 20:57:58,858 INFO [train.py:904] (5/8) Epoch 19, batch 3000, loss[loss=0.2062, simple_loss=0.2794, pruned_loss=0.06654, over 16767.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2591, pruned_loss=0.04359, over 3332583.15 frames. ], batch size: 124, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:57:58,858 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 20:58:07,644 INFO [train.py:938] (5/8) Epoch 19, validation: loss=0.1362, simple_loss=0.2416, pruned_loss=0.01538, over 944034.00 frames. 2023-04-30 20:58:07,645 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 20:58:28,251 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185716.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 20:59:16,288 INFO [train.py:904] (5/8) Epoch 19, batch 3050, loss[loss=0.1863, simple_loss=0.2692, pruned_loss=0.0517, over 16407.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2586, pruned_loss=0.04353, over 3329661.25 frames. ], batch size: 146, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:59:28,126 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185759.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:59:29,998 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 2.137e+02 2.455e+02 2.756e+02 4.328e+02, threshold=4.910e+02, percent-clipped=0.0 2023-04-30 20:59:35,051 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:03,988 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:23,331 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185799.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:00:23,347 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5379, 2.4506, 1.8846, 2.0224, 2.7719, 2.5653, 3.2434, 3.1355], device='cuda:5'), covar=tensor([0.0186, 0.0539, 0.0720, 0.0670, 0.0369, 0.0462, 0.0295, 0.0314], device='cuda:5'), in_proj_covar=tensor([0.0205, 0.0235, 0.0225, 0.0227, 0.0236, 0.0235, 0.0240, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:00:26,859 INFO [train.py:904] (5/8) Epoch 19, batch 3100, loss[loss=0.208, simple_loss=0.2756, pruned_loss=0.07015, over 12113.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2586, pruned_loss=0.04371, over 3321649.35 frames. ], batch size: 248, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:00:27,497 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-04-30 21:00:51,403 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185820.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:00:57,020 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185824.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:01:09,060 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185833.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:01:31,363 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1468, 3.2340, 3.4402, 2.2219, 3.2403, 3.5759, 3.3018, 2.0703], device='cuda:5'), covar=tensor([0.0495, 0.0115, 0.0063, 0.0405, 0.0111, 0.0088, 0.0095, 0.0445], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0082, 0.0082, 0.0134, 0.0096, 0.0107, 0.0093, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:01:34,505 INFO [train.py:904] (5/8) Epoch 19, batch 3150, loss[loss=0.1696, simple_loss=0.2538, pruned_loss=0.04273, over 16818.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.258, pruned_loss=0.04356, over 3324391.78 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:01:46,094 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185860.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:01:47,113 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.254e+02 2.659e+02 3.271e+02 6.332e+02, threshold=5.317e+02, percent-clipped=3.0 2023-04-30 21:02:43,074 INFO [train.py:904] (5/8) Epoch 19, batch 3200, loss[loss=0.1841, simple_loss=0.2607, pruned_loss=0.05375, over 16742.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2561, pruned_loss=0.04251, over 3329245.81 frames. ], batch size: 124, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:03:52,863 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1611, 2.3600, 2.8133, 3.0476, 3.0449, 3.6672, 2.4869, 3.6128], device='cuda:5'), covar=tensor([0.0225, 0.0387, 0.0273, 0.0299, 0.0272, 0.0143, 0.0436, 0.0145], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0192, 0.0179, 0.0182, 0.0192, 0.0151, 0.0195, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:03:53,536 INFO [train.py:904] (5/8) Epoch 19, batch 3250, loss[loss=0.1481, simple_loss=0.2331, pruned_loss=0.03153, over 16806.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2568, pruned_loss=0.0425, over 3331176.13 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:03:57,974 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1135, 5.1840, 5.6185, 5.5955, 5.6056, 5.2211, 5.1722, 5.0158], device='cuda:5'), covar=tensor([0.0352, 0.0472, 0.0341, 0.0446, 0.0550, 0.0374, 0.0975, 0.0421], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0449, 0.0435, 0.0408, 0.0483, 0.0460, 0.0556, 0.0365], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 21:04:06,307 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.264e+02 2.586e+02 3.072e+02 8.632e+02, threshold=5.173e+02, percent-clipped=3.0 2023-04-30 21:04:09,302 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1983, 3.3958, 3.5969, 2.1454, 3.0607, 2.3848, 3.6226, 3.7311], device='cuda:5'), covar=tensor([0.0251, 0.0916, 0.0584, 0.1921, 0.0844, 0.1022, 0.0558, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0161, 0.0165, 0.0151, 0.0143, 0.0127, 0.0143, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:05:07,477 INFO [train.py:904] (5/8) Epoch 19, batch 3300, loss[loss=0.1562, simple_loss=0.2519, pruned_loss=0.03028, over 17208.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2579, pruned_loss=0.04265, over 3323379.97 frames. ], batch size: 44, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:05:46,164 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3380, 2.2046, 2.3769, 4.0976, 2.2260, 2.6153, 2.2629, 2.4180], device='cuda:5'), covar=tensor([0.1262, 0.3485, 0.2699, 0.0541, 0.3915, 0.2473, 0.3697, 0.2996], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0439, 0.0362, 0.0328, 0.0434, 0.0506, 0.0409, 0.0513], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:05:58,605 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3666, 5.3133, 5.1215, 4.5401, 5.1347, 2.1789, 4.9281, 5.0557], device='cuda:5'), covar=tensor([0.0076, 0.0083, 0.0180, 0.0415, 0.0105, 0.2415, 0.0129, 0.0187], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0150, 0.0195, 0.0179, 0.0173, 0.0206, 0.0188, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:06:15,819 INFO [train.py:904] (5/8) Epoch 19, batch 3350, loss[loss=0.1625, simple_loss=0.251, pruned_loss=0.03698, over 16411.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2593, pruned_loss=0.04341, over 3320369.49 frames. ], batch size: 68, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:06:28,002 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.170e+02 2.521e+02 2.996e+02 5.638e+02, threshold=5.041e+02, percent-clipped=4.0 2023-04-30 21:06:48,427 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 21:06:55,237 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7515, 3.8571, 3.0029, 2.2767, 2.5884, 2.4701, 3.8518, 3.4264], device='cuda:5'), covar=tensor([0.2630, 0.0601, 0.1671, 0.2915, 0.2544, 0.1911, 0.0583, 0.1324], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0268, 0.0301, 0.0307, 0.0297, 0.0253, 0.0290, 0.0334], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:07:23,704 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186100.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:07:25,693 INFO [train.py:904] (5/8) Epoch 19, batch 3400, loss[loss=0.1631, simple_loss=0.2484, pruned_loss=0.03889, over 16855.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2589, pruned_loss=0.04265, over 3326280.22 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:07:44,686 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186115.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:07:57,590 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186124.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:34,728 INFO [train.py:904] (5/8) Epoch 19, batch 3450, loss[loss=0.1791, simple_loss=0.2505, pruned_loss=0.05387, over 16882.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2579, pruned_loss=0.04221, over 3325345.35 frames. ], batch size: 109, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:08:35,067 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7111, 5.0355, 5.4229, 5.4048, 5.4424, 5.0296, 4.7738, 4.7733], device='cuda:5'), covar=tensor([0.0607, 0.0668, 0.0568, 0.0685, 0.0700, 0.0594, 0.1571, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0448, 0.0433, 0.0406, 0.0480, 0.0458, 0.0552, 0.0364], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 21:08:37,979 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-30 21:08:39,237 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186155.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:08:40,535 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4096, 5.7883, 5.5073, 5.6176, 5.2044, 5.2709, 5.1962, 5.9232], device='cuda:5'), covar=tensor([0.1395, 0.0843, 0.1103, 0.0842, 0.0989, 0.0719, 0.1228, 0.0866], device='cuda:5'), in_proj_covar=tensor([0.0676, 0.0828, 0.0683, 0.0616, 0.0520, 0.0527, 0.0691, 0.0639], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:08:44,156 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186158.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:47,842 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.301e+02 2.793e+02 3.380e+02 7.495e+02, threshold=5.586e+02, percent-clipped=5.0 2023-04-30 21:08:48,329 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:09:03,002 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:09:12,799 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3463, 3.6321, 3.9060, 2.1338, 3.0879, 2.4724, 3.9008, 3.7968], device='cuda:5'), covar=tensor([0.0270, 0.0878, 0.0454, 0.1967, 0.0844, 0.0941, 0.0598, 0.0990], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0162, 0.0166, 0.0152, 0.0144, 0.0128, 0.0143, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:09:45,141 INFO [train.py:904] (5/8) Epoch 19, batch 3500, loss[loss=0.1935, simple_loss=0.2659, pruned_loss=0.06059, over 16742.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2563, pruned_loss=0.04106, over 3336587.48 frames. ], batch size: 134, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:10:08,981 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:10:22,043 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9140, 1.9382, 2.4556, 2.8457, 2.6811, 3.2766, 2.2287, 3.3361], device='cuda:5'), covar=tensor([0.0267, 0.0487, 0.0315, 0.0314, 0.0334, 0.0209, 0.0470, 0.0168], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0192, 0.0179, 0.0182, 0.0192, 0.0152, 0.0195, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:10:55,188 INFO [train.py:904] (5/8) Epoch 19, batch 3550, loss[loss=0.1414, simple_loss=0.2311, pruned_loss=0.02581, over 16852.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2552, pruned_loss=0.04094, over 3328566.12 frames. ], batch size: 42, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:11:01,641 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4443, 2.4675, 2.1052, 2.2578, 2.8195, 2.6111, 3.1420, 3.1768], device='cuda:5'), covar=tensor([0.0223, 0.0450, 0.0557, 0.0455, 0.0302, 0.0415, 0.0284, 0.0257], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0237, 0.0226, 0.0227, 0.0238, 0.0235, 0.0243, 0.0231], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:11:05,304 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186259.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:11:06,981 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 2.073e+02 2.370e+02 2.856e+02 5.323e+02, threshold=4.740e+02, percent-clipped=0.0 2023-04-30 21:11:29,944 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8923, 2.0429, 2.4827, 2.8328, 2.6867, 3.3529, 2.2601, 3.2896], device='cuda:5'), covar=tensor([0.0252, 0.0475, 0.0315, 0.0311, 0.0315, 0.0176, 0.0436, 0.0174], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0191, 0.0179, 0.0182, 0.0192, 0.0152, 0.0195, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:11:30,020 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8623, 4.2736, 3.0336, 2.3084, 2.7349, 2.4982, 4.6161, 3.6341], device='cuda:5'), covar=tensor([0.2700, 0.0544, 0.1723, 0.2763, 0.2800, 0.2029, 0.0354, 0.1198], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0268, 0.0301, 0.0306, 0.0297, 0.0253, 0.0290, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:11:42,009 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0504, 4.4991, 4.5217, 3.3170, 3.8689, 4.4652, 4.0913, 2.8140], device='cuda:5'), covar=tensor([0.0426, 0.0052, 0.0037, 0.0326, 0.0118, 0.0082, 0.0076, 0.0403], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0133, 0.0095, 0.0106, 0.0093, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:11:54,483 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 21:12:03,403 INFO [train.py:904] (5/8) Epoch 19, batch 3600, loss[loss=0.1582, simple_loss=0.2364, pruned_loss=0.03998, over 16475.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2544, pruned_loss=0.04099, over 3317221.81 frames. ], batch size: 75, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:12:28,383 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:13:13,936 INFO [train.py:904] (5/8) Epoch 19, batch 3650, loss[loss=0.1788, simple_loss=0.2436, pruned_loss=0.05705, over 16854.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2535, pruned_loss=0.04162, over 3312756.84 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:13:27,664 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.244e+02 2.696e+02 3.128e+02 5.239e+02, threshold=5.393e+02, percent-clipped=3.0 2023-04-30 21:14:16,298 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4853, 3.6199, 3.8884, 2.6564, 3.6094, 3.9367, 3.7068, 2.3292], device='cuda:5'), covar=tensor([0.0486, 0.0178, 0.0051, 0.0367, 0.0095, 0.0089, 0.0088, 0.0430], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0132, 0.0095, 0.0106, 0.0092, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:14:28,527 INFO [train.py:904] (5/8) Epoch 19, batch 3700, loss[loss=0.1605, simple_loss=0.232, pruned_loss=0.04449, over 16874.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2519, pruned_loss=0.04292, over 3284080.81 frames. ], batch size: 90, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:14:48,803 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186415.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:14:57,161 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:15:41,469 INFO [train.py:904] (5/8) Epoch 19, batch 3750, loss[loss=0.1673, simple_loss=0.2415, pruned_loss=0.04655, over 16785.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2526, pruned_loss=0.04464, over 3267076.55 frames. ], batch size: 102, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:15:46,735 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186455.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:15:47,878 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186456.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:15:55,024 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.254e+02 2.571e+02 3.173e+02 4.680e+02, threshold=5.142e+02, percent-clipped=0.0 2023-04-30 21:15:57,663 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186463.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:00,597 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1089, 4.3840, 3.3518, 2.9027, 3.2933, 2.9964, 4.6943, 4.1006], device='cuda:5'), covar=tensor([0.2499, 0.0587, 0.1617, 0.2090, 0.2119, 0.1649, 0.0415, 0.0903], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0267, 0.0301, 0.0306, 0.0298, 0.0253, 0.0290, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:16:15,868 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186475.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:18,765 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6321, 4.6982, 5.0103, 4.9948, 5.0368, 4.6875, 4.6794, 4.4958], device='cuda:5'), covar=tensor([0.0397, 0.0711, 0.0416, 0.0413, 0.0520, 0.0447, 0.1005, 0.0655], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0448, 0.0433, 0.0406, 0.0478, 0.0457, 0.0552, 0.0365], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 21:16:23,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9602, 5.3808, 5.5623, 5.2911, 5.3469, 5.9184, 5.4210, 5.0929], device='cuda:5'), covar=tensor([0.0969, 0.1592, 0.1555, 0.1680, 0.2165, 0.0791, 0.1310, 0.2252], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0587, 0.0648, 0.0491, 0.0655, 0.0685, 0.0508, 0.0657], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:16:26,548 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:27,619 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5330, 4.4548, 4.4574, 4.1956, 4.2094, 4.4852, 4.2867, 4.3119], device='cuda:5'), covar=tensor([0.0647, 0.0781, 0.0319, 0.0293, 0.0844, 0.0519, 0.0592, 0.0593], device='cuda:5'), in_proj_covar=tensor([0.0301, 0.0429, 0.0352, 0.0343, 0.0362, 0.0398, 0.0240, 0.0417], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:16:53,854 INFO [train.py:904] (5/8) Epoch 19, batch 3800, loss[loss=0.1796, simple_loss=0.2738, pruned_loss=0.0427, over 17216.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2538, pruned_loss=0.04575, over 3261006.12 frames. ], batch size: 40, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:16:55,984 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186503.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:17:11,673 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:15,566 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186517.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:43,768 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186536.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:47,996 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186539.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:05,819 INFO [train.py:904] (5/8) Epoch 19, batch 3850, loss[loss=0.2037, simple_loss=0.2748, pruned_loss=0.06636, over 16797.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.254, pruned_loss=0.0465, over 3269569.87 frames. ], batch size: 116, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:18:22,161 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.100e+02 2.475e+02 3.097e+02 5.903e+02, threshold=4.949e+02, percent-clipped=1.0 2023-04-30 21:18:28,977 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:40,365 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 21:18:45,034 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:56,187 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186586.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:16,926 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186600.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:18,740 INFO [train.py:904] (5/8) Epoch 19, batch 3900, loss[loss=0.1799, simple_loss=0.2566, pruned_loss=0.05163, over 16857.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.254, pruned_loss=0.04745, over 3256918.92 frames. ], batch size: 116, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:19:27,536 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-30 21:19:37,490 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:39,315 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6830, 4.6258, 4.6175, 4.3423, 4.2952, 4.6638, 4.4647, 4.4122], device='cuda:5'), covar=tensor([0.0616, 0.0720, 0.0313, 0.0289, 0.0976, 0.0494, 0.0461, 0.0627], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0425, 0.0349, 0.0340, 0.0360, 0.0396, 0.0239, 0.0414], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:19:55,113 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:21,627 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186647.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:28,331 INFO [train.py:904] (5/8) Epoch 19, batch 3950, loss[loss=0.1675, simple_loss=0.2473, pruned_loss=0.04383, over 16896.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2537, pruned_loss=0.04791, over 3255563.16 frames. ], batch size: 96, lr: 3.57e-03, grad_scale: 4.0 2023-04-30 21:20:42,976 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.601e+02 2.399e+02 2.822e+02 3.355e+02 8.994e+02, threshold=5.644e+02, percent-clipped=5.0 2023-04-30 21:21:11,730 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1120, 2.0587, 2.2023, 3.6274, 2.1195, 2.3609, 2.2303, 2.2469], device='cuda:5'), covar=tensor([0.1347, 0.3672, 0.2786, 0.0652, 0.3769, 0.2608, 0.3460, 0.3209], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0441, 0.0363, 0.0328, 0.0434, 0.0509, 0.0410, 0.0516], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:21:20,570 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4417, 4.4488, 4.4284, 3.9102, 4.4295, 1.8206, 4.1895, 4.0405], device='cuda:5'), covar=tensor([0.0108, 0.0095, 0.0177, 0.0294, 0.0093, 0.2658, 0.0155, 0.0229], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0150, 0.0196, 0.0179, 0.0173, 0.0206, 0.0188, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:21:39,187 INFO [train.py:904] (5/8) Epoch 19, batch 4000, loss[loss=0.1794, simple_loss=0.2619, pruned_loss=0.0485, over 16953.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2543, pruned_loss=0.04832, over 3260934.76 frames. ], batch size: 41, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:21:46,263 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8779, 4.8632, 4.7030, 4.0732, 4.8240, 1.9111, 4.5535, 4.3767], device='cuda:5'), covar=tensor([0.0095, 0.0081, 0.0178, 0.0308, 0.0082, 0.2675, 0.0129, 0.0225], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0150, 0.0196, 0.0179, 0.0173, 0.0205, 0.0188, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:21:49,120 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186709.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:22:49,924 INFO [train.py:904] (5/8) Epoch 19, batch 4050, loss[loss=0.1892, simple_loss=0.2691, pruned_loss=0.05467, over 12100.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2547, pruned_loss=0.04717, over 3254882.47 frames. ], batch size: 246, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:22:57,311 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:06,013 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.873e+02 2.216e+02 2.589e+02 4.495e+02, threshold=4.432e+02, percent-clipped=0.0 2023-04-30 21:23:16,266 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:27,136 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:41,754 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 21:24:03,628 INFO [train.py:904] (5/8) Epoch 19, batch 4100, loss[loss=0.2186, simple_loss=0.301, pruned_loss=0.06807, over 16733.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2568, pruned_loss=0.04695, over 3261485.76 frames. ], batch size: 124, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:24:07,125 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:21,979 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:42,407 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3216, 4.1855, 4.0225, 2.6983, 3.6888, 4.1122, 3.6444, 2.2273], device='cuda:5'), covar=tensor([0.0517, 0.0030, 0.0043, 0.0382, 0.0085, 0.0095, 0.0084, 0.0450], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0080, 0.0080, 0.0132, 0.0095, 0.0105, 0.0092, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:24:46,678 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:59,853 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186840.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:17,896 INFO [train.py:904] (5/8) Epoch 19, batch 4150, loss[loss=0.1826, simple_loss=0.277, pruned_loss=0.04403, over 16792.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2641, pruned_loss=0.04959, over 3230150.46 frames. ], batch size: 83, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:25:34,163 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:34,925 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.347e+02 2.016e+02 2.433e+02 2.890e+02 5.187e+02, threshold=4.866e+02, percent-clipped=4.0 2023-04-30 21:25:49,783 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186873.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:23,130 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:31,995 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186901.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:26:32,685 INFO [train.py:904] (5/8) Epoch 19, batch 4200, loss[loss=0.2286, simple_loss=0.3258, pruned_loss=0.0657, over 16670.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2711, pruned_loss=0.05101, over 3222573.62 frames. ], batch size: 62, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:26:47,072 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:53,418 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:59,339 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0437, 2.0758, 2.6267, 3.0029, 2.8930, 3.5154, 2.1257, 3.5127], device='cuda:5'), covar=tensor([0.0190, 0.0452, 0.0304, 0.0274, 0.0295, 0.0142, 0.0477, 0.0114], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0189, 0.0177, 0.0180, 0.0191, 0.0150, 0.0193, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:27:06,372 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:34,740 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:49,664 INFO [train.py:904] (5/8) Epoch 19, batch 4250, loss[loss=0.1775, simple_loss=0.2703, pruned_loss=0.04239, over 17253.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2744, pruned_loss=0.05101, over 3205689.83 frames. ], batch size: 52, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:28:05,622 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.275e+02 2.568e+02 3.131e+02 7.157e+02, threshold=5.135e+02, percent-clipped=7.0 2023-04-30 21:28:06,530 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:28:19,207 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:02,886 INFO [train.py:904] (5/8) Epoch 19, batch 4300, loss[loss=0.1929, simple_loss=0.2923, pruned_loss=0.0467, over 16182.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.275, pruned_loss=0.04976, over 3209920.51 frames. ], batch size: 165, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:29:37,800 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187025.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:56,590 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:16,639 INFO [train.py:904] (5/8) Epoch 19, batch 4350, loss[loss=0.201, simple_loss=0.2864, pruned_loss=0.05779, over 17214.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2784, pruned_loss=0.05093, over 3198531.54 frames. ], batch size: 46, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:30:32,933 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.200e+02 2.562e+02 2.922e+02 5.186e+02, threshold=5.123e+02, percent-clipped=1.0 2023-04-30 21:30:35,929 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:54,491 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:07,433 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187086.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:31:26,286 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187099.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:31,230 INFO [train.py:904] (5/8) Epoch 19, batch 4400, loss[loss=0.2034, simple_loss=0.2882, pruned_loss=0.05934, over 16677.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2804, pruned_loss=0.05215, over 3194804.70 frames. ], batch size: 57, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:31:46,434 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1889, 5.1570, 4.9393, 4.3335, 5.1052, 1.8813, 4.8126, 4.5372], device='cuda:5'), covar=tensor([0.0045, 0.0039, 0.0139, 0.0259, 0.0050, 0.2704, 0.0084, 0.0229], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0147, 0.0193, 0.0176, 0.0170, 0.0202, 0.0184, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:32:05,653 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:13,561 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187131.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:43,363 INFO [train.py:904] (5/8) Epoch 19, batch 4450, loss[loss=0.2053, simple_loss=0.2976, pruned_loss=0.05653, over 16450.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2833, pruned_loss=0.05275, over 3207399.92 frames. ], batch size: 146, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:32:47,751 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5712, 4.5745, 4.8901, 4.8474, 4.9118, 4.5495, 4.5468, 4.3713], device='cuda:5'), covar=tensor([0.0265, 0.0455, 0.0291, 0.0337, 0.0332, 0.0366, 0.0862, 0.0445], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0432, 0.0419, 0.0392, 0.0461, 0.0443, 0.0534, 0.0349], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 21:33:00,623 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.070e+02 2.320e+02 2.761e+02 4.804e+02, threshold=4.641e+02, percent-clipped=0.0 2023-04-30 21:33:15,043 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187173.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:24,882 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:31,779 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2684, 3.5015, 3.7565, 2.0958, 3.1860, 2.3384, 3.5502, 3.6728], device='cuda:5'), covar=tensor([0.0189, 0.0706, 0.0447, 0.2028, 0.0725, 0.0943, 0.0578, 0.0914], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0150, 0.0142, 0.0127, 0.0142, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:33:38,906 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2023-04-30 21:33:45,080 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9286, 3.9490, 2.5421, 4.9571, 3.2495, 4.8395, 2.8744, 3.2157], device='cuda:5'), covar=tensor([0.0267, 0.0370, 0.1607, 0.0111, 0.0744, 0.0325, 0.1279, 0.0728], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0176, 0.0194, 0.0158, 0.0176, 0.0216, 0.0199, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:33:48,141 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187195.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:50,025 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187196.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:33:53,890 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6739, 2.8666, 2.8681, 4.6454, 3.6946, 4.2009, 1.8628, 3.1202], device='cuda:5'), covar=tensor([0.1402, 0.0783, 0.1081, 0.0133, 0.0294, 0.0330, 0.1534, 0.0790], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0183, 0.0204, 0.0214, 0.0196, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:33:57,518 INFO [train.py:904] (5/8) Epoch 19, batch 4500, loss[loss=0.1906, simple_loss=0.2805, pruned_loss=0.05032, over 16848.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.284, pruned_loss=0.05378, over 3201918.64 frames. ], batch size: 42, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:34:24,638 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:27,224 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:54,308 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:55,419 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187243.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:08,491 INFO [train.py:904] (5/8) Epoch 19, batch 4550, loss[loss=0.2049, simple_loss=0.2928, pruned_loss=0.05847, over 16694.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2853, pruned_loss=0.05479, over 3217355.18 frames. ], batch size: 124, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:35:24,436 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 1.917e+02 2.180e+02 2.712e+02 4.446e+02, threshold=4.359e+02, percent-clipped=0.0 2023-04-30 21:35:27,320 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3267, 2.3438, 2.9425, 3.1738, 3.0490, 3.7242, 2.4107, 3.7846], device='cuda:5'), covar=tensor([0.0174, 0.0436, 0.0257, 0.0264, 0.0252, 0.0120, 0.0454, 0.0095], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0190, 0.0178, 0.0181, 0.0192, 0.0150, 0.0193, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:35:30,248 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:36,696 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187271.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:36:05,137 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:36:20,312 INFO [train.py:904] (5/8) Epoch 19, batch 4600, loss[loss=0.1744, simple_loss=0.2638, pruned_loss=0.04251, over 16764.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2857, pruned_loss=0.05494, over 3222293.07 frames. ], batch size: 62, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:04,648 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1462, 1.5437, 1.9112, 2.1112, 2.1801, 2.3624, 1.6879, 2.2986], device='cuda:5'), covar=tensor([0.0209, 0.0462, 0.0273, 0.0297, 0.0280, 0.0189, 0.0487, 0.0136], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0190, 0.0178, 0.0181, 0.0192, 0.0150, 0.0193, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:37:13,265 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6423, 2.4556, 2.4338, 3.3543, 2.4559, 3.6426, 1.4809, 2.6626], device='cuda:5'), covar=tensor([0.1356, 0.0805, 0.1157, 0.0154, 0.0158, 0.0354, 0.1666, 0.0820], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0173, 0.0193, 0.0183, 0.0205, 0.0215, 0.0197, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:37:33,051 INFO [train.py:904] (5/8) Epoch 19, batch 4650, loss[loss=0.1885, simple_loss=0.2788, pruned_loss=0.04911, over 16398.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2851, pruned_loss=0.05514, over 3214954.80 frames. ], batch size: 146, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:49,312 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.930e+02 2.188e+02 2.546e+02 5.637e+02, threshold=4.377e+02, percent-clipped=1.0 2023-04-30 21:37:52,342 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187365.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:01,620 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:15,619 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187381.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:38:23,759 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7878, 2.7314, 2.6458, 4.6011, 3.4088, 4.1291, 1.5897, 3.1015], device='cuda:5'), covar=tensor([0.1312, 0.0800, 0.1137, 0.0133, 0.0284, 0.0384, 0.1631, 0.0729], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0173, 0.0192, 0.0183, 0.0205, 0.0215, 0.0197, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:38:33,238 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:44,857 INFO [train.py:904] (5/8) Epoch 19, batch 4700, loss[loss=0.1755, simple_loss=0.2656, pruned_loss=0.04275, over 16715.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2832, pruned_loss=0.05459, over 3215535.97 frames. ], batch size: 83, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:39:01,768 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187413.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:30,050 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:55,450 INFO [train.py:904] (5/8) Epoch 19, batch 4750, loss[loss=0.1723, simple_loss=0.2603, pruned_loss=0.04219, over 16285.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2791, pruned_loss=0.05236, over 3211855.93 frames. ], batch size: 165, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:40:11,094 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 1.903e+02 2.161e+02 2.490e+02 3.592e+02, threshold=4.322e+02, percent-clipped=0.0 2023-04-30 21:40:57,694 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187496.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:05,674 INFO [train.py:904] (5/8) Epoch 19, batch 4800, loss[loss=0.178, simple_loss=0.2707, pruned_loss=0.04262, over 16516.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2753, pruned_loss=0.05054, over 3199962.15 frames. ], batch size: 75, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:41:21,427 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:30,322 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:34,857 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9812, 2.3596, 2.2609, 2.9088, 1.9423, 3.1881, 1.7761, 2.7204], device='cuda:5'), covar=tensor([0.1213, 0.0660, 0.1153, 0.0166, 0.0126, 0.0377, 0.1486, 0.0715], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0171, 0.0191, 0.0181, 0.0203, 0.0213, 0.0195, 0.0188], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:41:37,686 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187523.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:08,454 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187544.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:19,535 INFO [train.py:904] (5/8) Epoch 19, batch 4850, loss[loss=0.1929, simple_loss=0.2864, pruned_loss=0.0497, over 16761.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2756, pruned_loss=0.04978, over 3174577.27 frames. ], batch size: 83, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:42:36,216 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 1.857e+02 2.110e+02 2.587e+02 7.641e+02, threshold=4.221e+02, percent-clipped=2.0 2023-04-30 21:42:41,918 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:51,098 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:02,060 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:07,574 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:32,182 INFO [train.py:904] (5/8) Epoch 19, batch 4900, loss[loss=0.1647, simple_loss=0.2517, pruned_loss=0.03886, over 17131.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2744, pruned_loss=0.04813, over 3169723.86 frames. ], batch size: 49, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:43:51,080 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:44:04,767 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6117, 1.7838, 1.5764, 1.5070, 1.9268, 1.6166, 1.5499, 1.9096], device='cuda:5'), covar=tensor([0.0175, 0.0330, 0.0421, 0.0376, 0.0231, 0.0283, 0.0206, 0.0244], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0227, 0.0219, 0.0220, 0.0228, 0.0227, 0.0230, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:44:43,113 INFO [train.py:904] (5/8) Epoch 19, batch 4950, loss[loss=0.1814, simple_loss=0.2735, pruned_loss=0.04465, over 16487.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.274, pruned_loss=0.04768, over 3172231.35 frames. ], batch size: 75, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:44:58,343 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 2.032e+02 2.342e+02 2.691e+02 5.269e+02, threshold=4.685e+02, percent-clipped=1.0 2023-04-30 21:45:20,704 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7481, 3.7373, 2.7874, 2.4014, 2.7131, 2.5952, 4.1980, 3.5507], device='cuda:5'), covar=tensor([0.2611, 0.0774, 0.1931, 0.2420, 0.2388, 0.1782, 0.0481, 0.1087], device='cuda:5'), in_proj_covar=tensor([0.0323, 0.0266, 0.0302, 0.0307, 0.0297, 0.0251, 0.0290, 0.0331], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:45:24,754 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187681.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:45:43,953 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:45:47,592 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5537, 4.4161, 4.6299, 4.7693, 4.8984, 4.4511, 4.8915, 4.9262], device='cuda:5'), covar=tensor([0.1606, 0.1151, 0.1443, 0.0671, 0.0470, 0.0913, 0.0502, 0.0535], device='cuda:5'), in_proj_covar=tensor([0.0605, 0.0746, 0.0878, 0.0766, 0.0570, 0.0598, 0.0613, 0.0715], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:45:55,112 INFO [train.py:904] (5/8) Epoch 19, batch 5000, loss[loss=0.1961, simple_loss=0.2791, pruned_loss=0.05658, over 17063.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2756, pruned_loss=0.04752, over 3190867.75 frames. ], batch size: 53, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:46:32,202 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187728.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:33,447 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:52,693 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:47:07,086 INFO [train.py:904] (5/8) Epoch 19, batch 5050, loss[loss=0.1758, simple_loss=0.2652, pruned_loss=0.04321, over 17173.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2763, pruned_loss=0.04769, over 3193384.87 frames. ], batch size: 46, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:47:21,707 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.150e+02 2.460e+02 2.787e+02 5.516e+02, threshold=4.919e+02, percent-clipped=2.0 2023-04-30 21:47:35,592 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 21:48:01,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5295, 3.6374, 2.7375, 2.1582, 2.3732, 2.2937, 3.8630, 3.3627], device='cuda:5'), covar=tensor([0.2994, 0.0689, 0.1890, 0.2830, 0.2585, 0.2051, 0.0506, 0.1113], device='cuda:5'), in_proj_covar=tensor([0.0323, 0.0266, 0.0302, 0.0307, 0.0296, 0.0251, 0.0290, 0.0331], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:48:17,857 INFO [train.py:904] (5/8) Epoch 19, batch 5100, loss[loss=0.2268, simple_loss=0.2949, pruned_loss=0.0794, over 12128.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2747, pruned_loss=0.04696, over 3197976.29 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:48:39,813 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7040, 2.4895, 2.5456, 3.8660, 2.7172, 3.8594, 1.4820, 2.7956], device='cuda:5'), covar=tensor([0.1360, 0.0816, 0.1179, 0.0131, 0.0139, 0.0367, 0.1655, 0.0826], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0174, 0.0194, 0.0184, 0.0205, 0.0216, 0.0198, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:49:30,703 INFO [train.py:904] (5/8) Epoch 19, batch 5150, loss[loss=0.1776, simple_loss=0.2743, pruned_loss=0.04043, over 16833.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2746, pruned_loss=0.04642, over 3203506.37 frames. ], batch size: 116, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:49:31,397 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-30 21:49:47,461 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.389e+02 2.013e+02 2.291e+02 2.768e+02 7.535e+02, threshold=4.582e+02, percent-clipped=1.0 2023-04-30 21:49:54,412 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187868.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:04,813 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187875.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:11,346 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187879.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:41,940 INFO [train.py:904] (5/8) Epoch 19, batch 5200, loss[loss=0.1643, simple_loss=0.2499, pruned_loss=0.03937, over 16275.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2735, pruned_loss=0.04615, over 3183704.86 frames. ], batch size: 35, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:50:53,275 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6325, 3.6977, 2.2650, 4.1260, 2.8272, 4.1134, 2.4304, 2.9437], device='cuda:5'), covar=tensor([0.0227, 0.0285, 0.1462, 0.0152, 0.0843, 0.0382, 0.1396, 0.0737], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0173, 0.0190, 0.0153, 0.0172, 0.0211, 0.0197, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 21:50:53,842 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 21:51:02,532 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-30 21:51:53,325 INFO [train.py:904] (5/8) Epoch 19, batch 5250, loss[loss=0.1517, simple_loss=0.2508, pruned_loss=0.0263, over 16909.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.271, pruned_loss=0.04591, over 3172221.95 frames. ], batch size: 102, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:52:08,312 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.924e+02 2.321e+02 2.643e+02 4.137e+02, threshold=4.643e+02, percent-clipped=0.0 2023-04-30 21:52:34,712 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6040, 4.6391, 4.4493, 4.1461, 4.1173, 4.5285, 4.3487, 4.2568], device='cuda:5'), covar=tensor([0.0589, 0.0469, 0.0310, 0.0297, 0.1026, 0.0460, 0.0473, 0.0654], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0403, 0.0333, 0.0323, 0.0343, 0.0377, 0.0227, 0.0394], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:52:34,753 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1120, 3.8387, 3.7994, 2.3801, 3.3546, 3.7846, 3.4540, 2.0486], device='cuda:5'), covar=tensor([0.0564, 0.0051, 0.0042, 0.0408, 0.0110, 0.0099, 0.0104, 0.0489], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0077, 0.0078, 0.0129, 0.0093, 0.0103, 0.0090, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 21:53:08,608 INFO [train.py:904] (5/8) Epoch 19, batch 5300, loss[loss=0.157, simple_loss=0.2411, pruned_loss=0.03645, over 16918.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2671, pruned_loss=0.04443, over 3179987.56 frames. ], batch size: 116, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:53:46,820 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:53:54,842 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7608, 3.7208, 3.8676, 3.9561, 4.0520, 3.6912, 4.0235, 4.0940], device='cuda:5'), covar=tensor([0.1567, 0.1114, 0.1309, 0.0675, 0.0545, 0.1541, 0.0693, 0.0633], device='cuda:5'), in_proj_covar=tensor([0.0619, 0.0762, 0.0897, 0.0783, 0.0583, 0.0610, 0.0625, 0.0730], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:54:21,265 INFO [train.py:904] (5/8) Epoch 19, batch 5350, loss[loss=0.1858, simple_loss=0.2684, pruned_loss=0.05156, over 12012.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2656, pruned_loss=0.04388, over 3161130.26 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:54:38,016 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 1.961e+02 2.159e+02 2.589e+02 4.705e+02, threshold=4.317e+02, percent-clipped=1.0 2023-04-30 21:54:57,087 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:55:35,756 INFO [train.py:904] (5/8) Epoch 19, batch 5400, loss[loss=0.1751, simple_loss=0.2683, pruned_loss=0.04092, over 17016.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2682, pruned_loss=0.04441, over 3162365.78 frames. ], batch size: 41, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:56:49,747 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3530, 1.7661, 2.1285, 2.2935, 2.3697, 2.6447, 1.8149, 2.6073], device='cuda:5'), covar=tensor([0.0210, 0.0446, 0.0269, 0.0329, 0.0283, 0.0174, 0.0437, 0.0118], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0189, 0.0177, 0.0181, 0.0190, 0.0149, 0.0192, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:56:54,054 INFO [train.py:904] (5/8) Epoch 19, batch 5450, loss[loss=0.1894, simple_loss=0.2714, pruned_loss=0.05371, over 17114.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2714, pruned_loss=0.04601, over 3154427.26 frames. ], batch size: 49, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:56:59,336 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-04-30 21:57:11,922 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 1.998e+02 2.677e+02 3.507e+02 6.066e+02, threshold=5.353e+02, percent-clipped=8.0 2023-04-30 21:57:19,737 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188168.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:32,213 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188175.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:38,478 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:00,510 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2478, 2.2539, 2.1314, 3.8929, 2.1229, 2.6398, 2.3344, 2.4055], device='cuda:5'), covar=tensor([0.1247, 0.3521, 0.2923, 0.0529, 0.4220, 0.2361, 0.3329, 0.3321], device='cuda:5'), in_proj_covar=tensor([0.0390, 0.0431, 0.0357, 0.0319, 0.0427, 0.0498, 0.0402, 0.0503], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:58:14,505 INFO [train.py:904] (5/8) Epoch 19, batch 5500, loss[loss=0.2307, simple_loss=0.3172, pruned_loss=0.07212, over 16704.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2795, pruned_loss=0.05086, over 3131250.74 frames. ], batch size: 76, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:58:38,578 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:41,722 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5375, 3.5815, 3.3207, 3.0544, 3.2171, 3.4763, 3.3280, 3.2939], device='cuda:5'), covar=tensor([0.0507, 0.0566, 0.0261, 0.0237, 0.0507, 0.0433, 0.1302, 0.0439], device='cuda:5'), in_proj_covar=tensor([0.0283, 0.0407, 0.0334, 0.0325, 0.0345, 0.0380, 0.0229, 0.0398], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-04-30 21:58:44,477 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 21:58:48,994 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:55,304 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:59:35,427 INFO [train.py:904] (5/8) Epoch 19, batch 5550, loss[loss=0.3103, simple_loss=0.3639, pruned_loss=0.1284, over 10998.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2863, pruned_loss=0.05599, over 3106887.28 frames. ], batch size: 247, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:59:53,553 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.484e+02 3.263e+02 3.931e+02 4.573e+02 1.048e+03, threshold=7.861e+02, percent-clipped=6.0 2023-04-30 22:00:26,111 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:00:55,100 INFO [train.py:904] (5/8) Epoch 19, batch 5600, loss[loss=0.281, simple_loss=0.3518, pruned_loss=0.1051, over 15230.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2913, pruned_loss=0.05994, over 3096418.44 frames. ], batch size: 190, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:02:07,229 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:02:11,408 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-04-30 22:02:19,482 INFO [train.py:904] (5/8) Epoch 19, batch 5650, loss[loss=0.1915, simple_loss=0.2809, pruned_loss=0.05102, over 16374.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2958, pruned_loss=0.06358, over 3081003.02 frames. ], batch size: 68, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:02:36,992 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.386e+02 3.211e+02 3.758e+02 5.159e+02 1.051e+03, threshold=7.516e+02, percent-clipped=3.0 2023-04-30 22:03:36,662 INFO [train.py:904] (5/8) Epoch 19, batch 5700, loss[loss=0.1947, simple_loss=0.2849, pruned_loss=0.05221, over 16417.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2973, pruned_loss=0.06467, over 3070864.01 frames. ], batch size: 68, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:03:40,665 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7670, 1.3728, 1.7021, 1.6585, 1.7973, 1.8916, 1.6138, 1.8053], device='cuda:5'), covar=tensor([0.0237, 0.0380, 0.0201, 0.0273, 0.0253, 0.0163, 0.0364, 0.0132], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0188, 0.0176, 0.0179, 0.0190, 0.0148, 0.0191, 0.0143], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:03:54,699 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4992, 3.5907, 2.8114, 2.1562, 2.3849, 2.2851, 3.8346, 3.2344], device='cuda:5'), covar=tensor([0.2927, 0.0628, 0.1705, 0.2599, 0.2613, 0.2061, 0.0428, 0.1218], device='cuda:5'), in_proj_covar=tensor([0.0323, 0.0265, 0.0299, 0.0306, 0.0293, 0.0250, 0.0289, 0.0328], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 22:04:41,933 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4251, 3.4189, 3.3828, 2.3297, 3.2825, 2.0992, 2.9610, 2.5429], device='cuda:5'), covar=tensor([0.0192, 0.0152, 0.0232, 0.0420, 0.0137, 0.2775, 0.0174, 0.0341], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0143, 0.0189, 0.0172, 0.0165, 0.0198, 0.0179, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:04:55,841 INFO [train.py:904] (5/8) Epoch 19, batch 5750, loss[loss=0.2066, simple_loss=0.2962, pruned_loss=0.05845, over 17020.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2997, pruned_loss=0.0659, over 3064467.72 frames. ], batch size: 53, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:05:12,412 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 22:05:12,679 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.963e+02 2.939e+02 3.429e+02 4.149e+02 7.338e+02, threshold=6.857e+02, percent-clipped=0.0 2023-04-30 22:05:45,230 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:06:19,319 INFO [train.py:904] (5/8) Epoch 19, batch 5800, loss[loss=0.2348, simple_loss=0.3089, pruned_loss=0.08039, over 11506.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2996, pruned_loss=0.06551, over 3033134.15 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:06:26,974 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 22:07:24,096 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:07:38,960 INFO [train.py:904] (5/8) Epoch 19, batch 5850, loss[loss=0.2199, simple_loss=0.3104, pruned_loss=0.06474, over 16375.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2967, pruned_loss=0.06303, over 3055878.98 frames. ], batch size: 146, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:51,553 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9135, 2.7509, 2.8172, 2.1452, 2.6412, 2.1321, 2.7303, 2.9726], device='cuda:5'), covar=tensor([0.0276, 0.0738, 0.0554, 0.1682, 0.0751, 0.0942, 0.0527, 0.0680], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0160, 0.0166, 0.0151, 0.0144, 0.0128, 0.0143, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:07:57,315 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.681e+02 3.202e+02 3.753e+02 8.228e+02, threshold=6.404e+02, percent-clipped=1.0 2023-04-30 22:08:28,163 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 22:08:47,981 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 22:08:59,612 INFO [train.py:904] (5/8) Epoch 19, batch 5900, loss[loss=0.2437, simple_loss=0.3079, pruned_loss=0.0898, over 11498.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2966, pruned_loss=0.06311, over 3062336.37 frames. ], batch size: 246, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:09:38,533 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1774, 3.6100, 3.6528, 2.5340, 3.4215, 3.6816, 3.5139, 1.9575], device='cuda:5'), covar=tensor([0.0517, 0.0070, 0.0060, 0.0373, 0.0096, 0.0121, 0.0083, 0.0472], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0079, 0.0080, 0.0132, 0.0095, 0.0106, 0.0092, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 22:10:01,977 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:10:21,849 INFO [train.py:904] (5/8) Epoch 19, batch 5950, loss[loss=0.1948, simple_loss=0.2874, pruned_loss=0.05112, over 16457.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2964, pruned_loss=0.06133, over 3089051.71 frames. ], batch size: 146, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:10:26,017 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 22:10:40,537 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.478e+02 2.950e+02 3.832e+02 6.444e+02, threshold=5.900e+02, percent-clipped=1.0 2023-04-30 22:11:41,788 INFO [train.py:904] (5/8) Epoch 19, batch 6000, loss[loss=0.1793, simple_loss=0.272, pruned_loss=0.04325, over 16479.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.295, pruned_loss=0.06061, over 3093627.62 frames. ], batch size: 68, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:11:41,788 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 22:11:52,555 INFO [train.py:938] (5/8) Epoch 19, validation: loss=0.1523, simple_loss=0.2653, pruned_loss=0.01968, over 944034.00 frames. 2023-04-30 22:11:52,556 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 22:12:32,488 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8371, 1.4213, 1.7514, 1.6809, 1.7636, 1.9446, 1.6023, 1.8650], device='cuda:5'), covar=tensor([0.0247, 0.0343, 0.0207, 0.0271, 0.0263, 0.0181, 0.0384, 0.0121], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0187, 0.0175, 0.0177, 0.0188, 0.0147, 0.0190, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:12:32,735 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 22:12:33,710 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3113, 5.1815, 5.3390, 5.5173, 5.7037, 4.9589, 5.6537, 5.6848], device='cuda:5'), covar=tensor([0.1903, 0.1098, 0.1456, 0.0649, 0.0494, 0.0821, 0.0514, 0.0537], device='cuda:5'), in_proj_covar=tensor([0.0609, 0.0748, 0.0877, 0.0769, 0.0573, 0.0602, 0.0618, 0.0720], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:12:59,821 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 22:13:13,864 INFO [train.py:904] (5/8) Epoch 19, batch 6050, loss[loss=0.222, simple_loss=0.2902, pruned_loss=0.0769, over 11774.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2937, pruned_loss=0.06047, over 3087474.00 frames. ], batch size: 250, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:13:24,326 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5345, 3.6266, 2.8102, 2.2042, 2.4303, 2.3714, 3.7933, 3.3199], device='cuda:5'), covar=tensor([0.2986, 0.0673, 0.1721, 0.2830, 0.2485, 0.2016, 0.0505, 0.1289], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0304, 0.0291, 0.0249, 0.0288, 0.0327], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 22:13:33,108 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.719e+02 3.422e+02 3.983e+02 7.831e+02, threshold=6.844e+02, percent-clipped=4.0 2023-04-30 22:14:05,853 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1408, 2.0821, 1.6840, 1.7963, 2.3276, 1.9820, 2.0688, 2.3786], device='cuda:5'), covar=tensor([0.0184, 0.0333, 0.0496, 0.0439, 0.0241, 0.0359, 0.0195, 0.0245], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0229, 0.0222, 0.0222, 0.0231, 0.0229, 0.0231, 0.0224], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:14:32,950 INFO [train.py:904] (5/8) Epoch 19, batch 6100, loss[loss=0.234, simple_loss=0.3045, pruned_loss=0.08178, over 11661.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2936, pruned_loss=0.05925, over 3116777.45 frames. ], batch size: 248, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:14:49,372 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-30 22:15:18,271 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8398, 3.7283, 3.8773, 3.9843, 4.0684, 3.6901, 4.0122, 4.0825], device='cuda:5'), covar=tensor([0.1496, 0.1121, 0.1313, 0.0653, 0.0580, 0.1610, 0.0873, 0.0694], device='cuda:5'), in_proj_covar=tensor([0.0611, 0.0749, 0.0882, 0.0771, 0.0576, 0.0604, 0.0621, 0.0723], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:15:33,845 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:15:57,518 INFO [train.py:904] (5/8) Epoch 19, batch 6150, loss[loss=0.2028, simple_loss=0.2924, pruned_loss=0.05659, over 16750.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.292, pruned_loss=0.05917, over 3117700.61 frames. ], batch size: 124, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:16:16,886 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.888e+02 3.402e+02 4.182e+02 1.052e+03, threshold=6.804e+02, percent-clipped=3.0 2023-04-30 22:17:16,340 INFO [train.py:904] (5/8) Epoch 19, batch 6200, loss[loss=0.2022, simple_loss=0.288, pruned_loss=0.05826, over 16217.00 frames. ], tot_loss[loss=0.205, simple_loss=0.291, pruned_loss=0.05956, over 3108381.49 frames. ], batch size: 165, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:17:47,265 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 22:17:58,791 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:01,885 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188931.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:13,175 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4425, 2.9348, 3.0412, 1.9665, 2.7504, 2.1269, 3.0753, 3.1756], device='cuda:5'), covar=tensor([0.0280, 0.0798, 0.0569, 0.1956, 0.0830, 0.0981, 0.0643, 0.0847], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0143, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:18:15,047 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:34,124 INFO [train.py:904] (5/8) Epoch 19, batch 6250, loss[loss=0.2171, simple_loss=0.292, pruned_loss=0.07114, over 11924.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2909, pruned_loss=0.05963, over 3105432.47 frames. ], batch size: 248, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:18:52,890 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.827e+02 3.410e+02 4.287e+02 1.283e+03, threshold=6.821e+02, percent-clipped=4.0 2023-04-30 22:19:28,072 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:30,838 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5551, 4.6079, 4.9485, 4.9193, 4.9303, 4.6277, 4.5766, 4.4480], device='cuda:5'), covar=tensor([0.0352, 0.0580, 0.0386, 0.0414, 0.0510, 0.0417, 0.1014, 0.0548], device='cuda:5'), in_proj_covar=tensor([0.0396, 0.0434, 0.0422, 0.0396, 0.0469, 0.0443, 0.0537, 0.0355], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 22:19:32,365 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:36,008 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:51,026 INFO [train.py:904] (5/8) Epoch 19, batch 6300, loss[loss=0.1959, simple_loss=0.2825, pruned_loss=0.05466, over 16482.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2905, pruned_loss=0.05883, over 3119101.23 frames. ], batch size: 68, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:20:35,795 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189030.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:20:54,560 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2877, 4.3357, 4.6434, 4.6266, 4.6487, 4.3556, 4.3719, 4.2581], device='cuda:5'), covar=tensor([0.0385, 0.0714, 0.0481, 0.0536, 0.0548, 0.0541, 0.0902, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0395, 0.0433, 0.0422, 0.0395, 0.0468, 0.0444, 0.0535, 0.0354], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 22:21:09,048 INFO [train.py:904] (5/8) Epoch 19, batch 6350, loss[loss=0.2447, simple_loss=0.3079, pruned_loss=0.09073, over 11832.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2915, pruned_loss=0.06, over 3105778.95 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:21:27,082 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.691e+02 3.246e+02 4.132e+02 1.028e+03, threshold=6.492e+02, percent-clipped=3.0 2023-04-30 22:21:28,379 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:21:43,618 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1048, 4.2012, 3.9987, 3.8052, 3.5579, 4.1386, 3.8290, 3.7834], device='cuda:5'), covar=tensor([0.0674, 0.0618, 0.0399, 0.0352, 0.1056, 0.0539, 0.1022, 0.0756], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0405, 0.0330, 0.0320, 0.0340, 0.0373, 0.0227, 0.0394], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:21:51,874 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:11,168 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:26,890 INFO [train.py:904] (5/8) Epoch 19, batch 6400, loss[loss=0.1944, simple_loss=0.2904, pruned_loss=0.04916, over 16882.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2923, pruned_loss=0.06116, over 3092924.14 frames. ], batch size: 96, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:23:02,730 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:21,496 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:25,611 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189140.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:33,403 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-30 22:23:36,534 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1727, 5.4907, 5.2713, 5.3090, 4.9375, 4.8804, 4.9174, 5.6236], device='cuda:5'), covar=tensor([0.1286, 0.0943, 0.1076, 0.0965, 0.0891, 0.0851, 0.1235, 0.0964], device='cuda:5'), in_proj_covar=tensor([0.0636, 0.0783, 0.0642, 0.0583, 0.0489, 0.0500, 0.0650, 0.0602], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:23:43,284 INFO [train.py:904] (5/8) Epoch 19, batch 6450, loss[loss=0.1793, simple_loss=0.2549, pruned_loss=0.05182, over 11303.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.292, pruned_loss=0.06046, over 3080679.12 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:23:48,350 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7212, 3.0772, 2.7154, 5.1235, 3.8502, 4.2722, 1.9847, 2.9988], device='cuda:5'), covar=tensor([0.1523, 0.0806, 0.1329, 0.0207, 0.0487, 0.0509, 0.1570, 0.0971], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0171, 0.0193, 0.0182, 0.0204, 0.0213, 0.0196, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:24:01,002 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.834e+02 2.902e+02 3.401e+02 4.159e+02 9.749e+02, threshold=6.803e+02, percent-clipped=3.0 2023-04-30 22:24:26,555 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0724, 5.1495, 5.5010, 5.4921, 5.5333, 5.1463, 5.1320, 4.8729], device='cuda:5'), covar=tensor([0.0318, 0.0500, 0.0389, 0.0395, 0.0416, 0.0362, 0.0967, 0.0437], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0430, 0.0420, 0.0394, 0.0466, 0.0441, 0.0531, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 22:24:34,447 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:24:48,432 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-04-30 22:25:00,521 INFO [train.py:904] (5/8) Epoch 19, batch 6500, loss[loss=0.2084, simple_loss=0.2916, pruned_loss=0.0626, over 15214.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.29, pruned_loss=0.05946, over 3104965.43 frames. ], batch size: 190, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:25:54,431 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7489, 1.8451, 1.6078, 1.4840, 1.9563, 1.6039, 1.6727, 1.9258], device='cuda:5'), covar=tensor([0.0177, 0.0255, 0.0342, 0.0313, 0.0186, 0.0225, 0.0177, 0.0186], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0224, 0.0218, 0.0218, 0.0225, 0.0224, 0.0226, 0.0220], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:26:03,210 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 22:26:14,676 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:26:16,140 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 22:26:19,689 INFO [train.py:904] (5/8) Epoch 19, batch 6550, loss[loss=0.2095, simple_loss=0.3079, pruned_loss=0.05555, over 16566.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.293, pruned_loss=0.06113, over 3070170.30 frames. ], batch size: 75, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:26:30,427 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5439, 3.3205, 3.6791, 1.9720, 3.7822, 3.8573, 3.0107, 2.8791], device='cuda:5'), covar=tensor([0.0738, 0.0263, 0.0210, 0.1191, 0.0079, 0.0216, 0.0387, 0.0475], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0107, 0.0097, 0.0139, 0.0078, 0.0122, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:26:36,419 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6375, 2.6479, 2.4440, 3.6969, 2.5578, 3.8362, 1.4226, 2.8258], device='cuda:5'), covar=tensor([0.1445, 0.0728, 0.1187, 0.0179, 0.0204, 0.0440, 0.1729, 0.0794], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0172, 0.0193, 0.0183, 0.0205, 0.0214, 0.0197, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:26:37,018 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 2.752e+02 3.137e+02 4.130e+02 8.401e+02, threshold=6.273e+02, percent-clipped=2.0 2023-04-30 22:27:04,513 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9177, 4.1713, 3.9799, 4.0411, 3.7197, 3.8218, 3.8390, 4.1746], device='cuda:5'), covar=tensor([0.1161, 0.0937, 0.1121, 0.0863, 0.0846, 0.1581, 0.1056, 0.0999], device='cuda:5'), in_proj_covar=tensor([0.0640, 0.0785, 0.0647, 0.0586, 0.0492, 0.0503, 0.0654, 0.0605], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:27:09,603 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:27:12,563 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:27:34,968 INFO [train.py:904] (5/8) Epoch 19, batch 6600, loss[loss=0.2419, simple_loss=0.3122, pruned_loss=0.08577, over 11798.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2952, pruned_loss=0.06173, over 3063395.08 frames. ], batch size: 248, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:27:37,323 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5377, 2.6841, 2.2393, 2.5466, 3.0424, 2.7050, 3.1465, 3.1959], device='cuda:5'), covar=tensor([0.0098, 0.0405, 0.0475, 0.0398, 0.0235, 0.0363, 0.0226, 0.0230], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0224, 0.0218, 0.0218, 0.0226, 0.0224, 0.0226, 0.0221], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:27:45,610 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:44,317 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:50,955 INFO [train.py:904] (5/8) Epoch 19, batch 6650, loss[loss=0.2243, simple_loss=0.3005, pruned_loss=0.07405, over 15317.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2953, pruned_loss=0.06265, over 3054503.76 frames. ], batch size: 191, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:29:08,223 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.870e+02 3.524e+02 4.277e+02 9.308e+02, threshold=7.047e+02, percent-clipped=5.0 2023-04-30 22:29:39,787 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 22:29:42,461 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189386.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:06,164 INFO [train.py:904] (5/8) Epoch 19, batch 6700, loss[loss=0.1973, simple_loss=0.2818, pruned_loss=0.05643, over 16689.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2944, pruned_loss=0.0628, over 3052446.78 frames. ], batch size: 57, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:30:15,440 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:33,779 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:43,048 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6499, 2.6007, 1.8737, 2.7051, 2.1370, 2.7663, 2.1624, 2.3802], device='cuda:5'), covar=tensor([0.0260, 0.0323, 0.1222, 0.0218, 0.0630, 0.0460, 0.1062, 0.0553], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0176, 0.0194, 0.0156, 0.0175, 0.0214, 0.0201, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:30:56,720 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189435.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:31:21,926 INFO [train.py:904] (5/8) Epoch 19, batch 6750, loss[loss=0.1761, simple_loss=0.262, pruned_loss=0.04512, over 16896.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2933, pruned_loss=0.06316, over 3047392.53 frames. ], batch size: 109, lr: 3.55e-03, grad_scale: 4.0 2023-04-30 22:31:42,236 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.911e+02 3.292e+02 3.890e+02 7.012e+02, threshold=6.585e+02, percent-clipped=0.0 2023-04-30 22:32:38,331 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:32:39,013 INFO [train.py:904] (5/8) Epoch 19, batch 6800, loss[loss=0.2049, simple_loss=0.2858, pruned_loss=0.062, over 16649.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2935, pruned_loss=0.06299, over 3066730.62 frames. ], batch size: 57, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:32:46,092 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7243, 3.8201, 2.3722, 4.4733, 2.9035, 4.3489, 2.3964, 2.9766], device='cuda:5'), covar=tensor([0.0260, 0.0342, 0.1694, 0.0167, 0.0858, 0.0491, 0.1548, 0.0782], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0176, 0.0194, 0.0156, 0.0175, 0.0215, 0.0201, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:33:19,152 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.46 vs. limit=2.0 2023-04-30 22:33:58,201 INFO [train.py:904] (5/8) Epoch 19, batch 6850, loss[loss=0.2655, simple_loss=0.3276, pruned_loss=0.1017, over 11676.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2948, pruned_loss=0.06324, over 3062088.45 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:34:13,641 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189562.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:15,277 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1045, 4.0090, 4.1880, 4.3176, 4.4612, 4.0491, 4.3744, 4.4714], device='cuda:5'), covar=tensor([0.1941, 0.1276, 0.1525, 0.0819, 0.0677, 0.1419, 0.1085, 0.0720], device='cuda:5'), in_proj_covar=tensor([0.0609, 0.0750, 0.0881, 0.0770, 0.0578, 0.0606, 0.0622, 0.0723], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:34:17,031 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.732e+02 3.346e+02 4.467e+02 6.913e+02, threshold=6.693e+02, percent-clipped=4.0 2023-04-30 22:34:47,788 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:50,446 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:35:13,002 INFO [train.py:904] (5/8) Epoch 19, batch 6900, loss[loss=0.209, simple_loss=0.3029, pruned_loss=0.05752, over 16732.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.297, pruned_loss=0.0626, over 3084296.97 frames. ], batch size: 89, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:35:17,365 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:35:43,587 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7601, 4.0855, 4.1444, 2.5343, 3.3627, 2.8476, 4.1852, 4.3202], device='cuda:5'), covar=tensor([0.0222, 0.0692, 0.0486, 0.1861, 0.0780, 0.0890, 0.0521, 0.0817], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0161, 0.0167, 0.0152, 0.0144, 0.0129, 0.0144, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:36:00,960 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189633.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:04,229 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189635.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:12,831 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189640.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:29,248 INFO [train.py:904] (5/8) Epoch 19, batch 6950, loss[loss=0.2154, simple_loss=0.2953, pruned_loss=0.0678, over 15308.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2986, pruned_loss=0.06388, over 3095949.58 frames. ], batch size: 191, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:36:48,883 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.017e+02 3.156e+02 3.874e+02 4.949e+02 9.528e+02, threshold=7.748e+02, percent-clipped=6.0 2023-04-30 22:37:21,393 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:43,869 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:44,556 INFO [train.py:904] (5/8) Epoch 19, batch 7000, loss[loss=0.2201, simple_loss=0.3126, pruned_loss=0.06384, over 15423.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2984, pruned_loss=0.06323, over 3091843.99 frames. ], batch size: 191, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:37:47,746 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189703.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:49,114 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:13,518 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:25,597 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189728.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:34,807 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:36,188 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:44,679 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4258, 4.5765, 4.3221, 4.0849, 3.8300, 4.4716, 4.2574, 4.0572], device='cuda:5'), covar=tensor([0.0716, 0.0595, 0.0372, 0.0336, 0.1086, 0.0517, 0.0627, 0.0741], device='cuda:5'), in_proj_covar=tensor([0.0277, 0.0400, 0.0325, 0.0317, 0.0336, 0.0369, 0.0225, 0.0390], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:39:01,655 INFO [train.py:904] (5/8) Epoch 19, batch 7050, loss[loss=0.182, simple_loss=0.2809, pruned_loss=0.04158, over 16874.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2985, pruned_loss=0.06219, over 3108483.99 frames. ], batch size: 96, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:39:22,080 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.118e+02 2.966e+02 3.403e+02 4.255e+02 9.294e+02, threshold=6.806e+02, percent-clipped=3.0 2023-04-30 22:39:23,267 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:27,447 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189768.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:42,318 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 22:39:50,863 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:56,931 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8713, 3.1578, 2.8643, 5.2501, 4.1978, 4.4205, 1.9230, 3.1691], device='cuda:5'), covar=tensor([0.1378, 0.0723, 0.1222, 0.0166, 0.0418, 0.0396, 0.1570, 0.0946], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0173, 0.0194, 0.0184, 0.0207, 0.0215, 0.0198, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:40:01,261 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189789.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:19,980 INFO [train.py:904] (5/8) Epoch 19, batch 7100, loss[loss=0.2102, simple_loss=0.2918, pruned_loss=0.06426, over 16345.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2967, pruned_loss=0.0619, over 3101565.60 frames. ], batch size: 165, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:41:38,326 INFO [train.py:904] (5/8) Epoch 19, batch 7150, loss[loss=0.2531, simple_loss=0.3212, pruned_loss=0.09252, over 11223.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2945, pruned_loss=0.06108, over 3110775.24 frames. ], batch size: 246, lr: 3.54e-03, grad_scale: 4.0 2023-04-30 22:41:45,579 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:41:58,950 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.755e+02 3.301e+02 3.879e+02 9.632e+02, threshold=6.601e+02, percent-clipped=3.0 2023-04-30 22:42:01,563 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 22:42:51,652 INFO [train.py:904] (5/8) Epoch 19, batch 7200, loss[loss=0.1758, simple_loss=0.2645, pruned_loss=0.04349, over 11801.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2922, pruned_loss=0.05971, over 3089877.19 frames. ], batch size: 247, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:42:55,782 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189904.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:44:08,857 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0092, 1.9927, 2.5598, 2.9212, 2.7414, 3.3351, 2.1423, 3.2232], device='cuda:5'), covar=tensor([0.0168, 0.0502, 0.0284, 0.0264, 0.0280, 0.0136, 0.0521, 0.0128], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0189, 0.0175, 0.0178, 0.0191, 0.0148, 0.0191, 0.0142], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:44:12,503 INFO [train.py:904] (5/8) Epoch 19, batch 7250, loss[loss=0.1983, simple_loss=0.2812, pruned_loss=0.0577, over 15428.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2894, pruned_loss=0.05859, over 3080991.92 frames. ], batch size: 190, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:44:12,831 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189952.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:44:34,220 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.477e+02 2.873e+02 3.623e+02 8.553e+02, threshold=5.746e+02, percent-clipped=4.0 2023-04-30 22:45:12,371 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-30 22:45:19,912 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189996.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:45:32,095 INFO [train.py:904] (5/8) Epoch 19, batch 7300, loss[loss=0.2823, simple_loss=0.3351, pruned_loss=0.1148, over 11403.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2885, pruned_loss=0.05833, over 3076340.09 frames. ], batch size: 247, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:45:33,995 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190003.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:46:48,531 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190051.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:46:49,312 INFO [train.py:904] (5/8) Epoch 19, batch 7350, loss[loss=0.1844, simple_loss=0.279, pruned_loss=0.04494, over 16854.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2895, pruned_loss=0.05939, over 3053015.86 frames. ], batch size: 102, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:46:58,338 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2810, 3.2018, 3.2390, 3.4205, 3.4140, 3.2133, 3.4095, 3.4802], device='cuda:5'), covar=tensor([0.1225, 0.1088, 0.1385, 0.0738, 0.0844, 0.2787, 0.1198, 0.0961], device='cuda:5'), in_proj_covar=tensor([0.0598, 0.0736, 0.0868, 0.0756, 0.0568, 0.0596, 0.0613, 0.0708], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:47:01,719 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:47:10,691 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.253e+02 3.016e+02 3.386e+02 4.070e+02 1.285e+03, threshold=6.773e+02, percent-clipped=7.0 2023-04-30 22:47:41,034 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:48:08,865 INFO [train.py:904] (5/8) Epoch 19, batch 7400, loss[loss=0.22, simple_loss=0.3072, pruned_loss=0.06647, over 16686.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2901, pruned_loss=0.05976, over 3065412.42 frames. ], batch size: 134, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:49:09,834 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190139.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 22:49:11,233 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2918, 4.1822, 4.3232, 4.4582, 4.6267, 4.2164, 4.5774, 4.6401], device='cuda:5'), covar=tensor([0.1995, 0.1289, 0.1691, 0.0848, 0.0643, 0.1198, 0.0816, 0.0702], device='cuda:5'), in_proj_covar=tensor([0.0602, 0.0740, 0.0872, 0.0761, 0.0571, 0.0599, 0.0617, 0.0713], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:49:30,149 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190151.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:49:30,834 INFO [train.py:904] (5/8) Epoch 19, batch 7450, loss[loss=0.2289, simple_loss=0.3202, pruned_loss=0.06882, over 16314.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2926, pruned_loss=0.06174, over 3052566.46 frames. ], batch size: 165, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:49:41,070 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190157.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:49:55,755 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 2.938e+02 3.493e+02 4.428e+02 1.014e+03, threshold=6.986e+02, percent-clipped=6.0 2023-04-30 22:50:51,737 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190200.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 22:50:54,226 INFO [train.py:904] (5/8) Epoch 19, batch 7500, loss[loss=0.1968, simple_loss=0.275, pruned_loss=0.05928, over 16852.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2922, pruned_loss=0.06059, over 3066949.18 frames. ], batch size: 116, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:50:55,481 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3300, 2.5302, 1.8969, 2.3194, 2.9810, 2.5994, 3.0006, 3.1463], device='cuda:5'), covar=tensor([0.0134, 0.0422, 0.0618, 0.0483, 0.0243, 0.0408, 0.0236, 0.0249], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0226, 0.0218, 0.0218, 0.0227, 0.0225, 0.0227, 0.0220], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:50:59,305 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:51:10,823 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:52:13,622 INFO [train.py:904] (5/8) Epoch 19, batch 7550, loss[loss=0.1947, simple_loss=0.2795, pruned_loss=0.055, over 16766.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2913, pruned_loss=0.06072, over 3061437.07 frames. ], batch size: 124, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:52:16,592 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7174, 3.7338, 2.4223, 4.3444, 2.8898, 4.2680, 2.4533, 3.0088], device='cuda:5'), covar=tensor([0.0278, 0.0431, 0.1562, 0.0146, 0.0752, 0.0602, 0.1536, 0.0752], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0174, 0.0193, 0.0155, 0.0174, 0.0212, 0.0200, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-04-30 22:52:34,942 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.804e+02 2.643e+02 3.256e+02 4.054e+02 9.532e+02, threshold=6.511e+02, percent-clipped=2.0 2023-04-30 22:53:22,545 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190296.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:53:31,297 INFO [train.py:904] (5/8) Epoch 19, batch 7600, loss[loss=0.1965, simple_loss=0.289, pruned_loss=0.052, over 17184.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2909, pruned_loss=0.06102, over 3069162.82 frames. ], batch size: 46, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:53:53,574 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190316.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:54:03,510 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 22:54:37,075 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:54:46,516 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2545, 3.4630, 3.5887, 3.5556, 3.5778, 3.3999, 3.4411, 3.4838], device='cuda:5'), covar=tensor([0.0424, 0.0627, 0.0458, 0.0474, 0.0509, 0.0541, 0.0794, 0.0506], device='cuda:5'), in_proj_covar=tensor([0.0388, 0.0426, 0.0414, 0.0388, 0.0462, 0.0435, 0.0527, 0.0349], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 22:54:48,959 INFO [train.py:904] (5/8) Epoch 19, batch 7650, loss[loss=0.2011, simple_loss=0.2948, pruned_loss=0.05375, over 16865.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2916, pruned_loss=0.06183, over 3054001.31 frames. ], batch size: 102, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:55:02,261 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:55:10,294 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.023e+02 2.803e+02 3.556e+02 4.194e+02 6.713e+02, threshold=7.113e+02, percent-clipped=1.0 2023-04-30 22:55:27,571 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190377.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:55:38,599 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:02,482 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:06,248 INFO [train.py:904] (5/8) Epoch 19, batch 7700, loss[loss=0.2377, simple_loss=0.3072, pruned_loss=0.08409, over 11728.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2912, pruned_loss=0.06165, over 3066622.83 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:56:15,129 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:50,414 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:55,004 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3370, 3.3291, 3.3966, 3.4750, 3.4842, 3.2620, 3.4653, 3.5457], device='cuda:5'), covar=tensor([0.1297, 0.0897, 0.1039, 0.0620, 0.0661, 0.2158, 0.1056, 0.0821], device='cuda:5'), in_proj_covar=tensor([0.0599, 0.0739, 0.0867, 0.0756, 0.0571, 0.0595, 0.0615, 0.0709], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 22:57:21,416 INFO [train.py:904] (5/8) Epoch 19, batch 7750, loss[loss=0.1747, simple_loss=0.2721, pruned_loss=0.03862, over 16883.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2908, pruned_loss=0.06118, over 3058035.23 frames. ], batch size: 102, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:57:35,162 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:57:43,814 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.725e+02 3.462e+02 4.075e+02 8.569e+02, threshold=6.924e+02, percent-clipped=2.0 2023-04-30 22:58:30,672 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190495.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 22:58:39,276 INFO [train.py:904] (5/8) Epoch 19, batch 7800, loss[loss=0.2163, simple_loss=0.3043, pruned_loss=0.06418, over 15369.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2921, pruned_loss=0.06184, over 3060916.19 frames. ], batch size: 191, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:58:47,781 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190507.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:59:06,268 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 22:59:19,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 22:59:57,145 INFO [train.py:904] (5/8) Epoch 19, batch 7850, loss[loss=0.2025, simple_loss=0.2872, pruned_loss=0.05888, over 16495.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.293, pruned_loss=0.06192, over 3061049.62 frames. ], batch size: 75, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:00:17,837 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.838e+02 3.506e+02 4.307e+02 8.316e+02, threshold=7.012e+02, percent-clipped=1.0 2023-04-30 23:00:32,873 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 23:00:56,965 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190591.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:01:10,439 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 23:01:12,387 INFO [train.py:904] (5/8) Epoch 19, batch 7900, loss[loss=0.2355, simple_loss=0.3295, pruned_loss=0.07078, over 15412.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2921, pruned_loss=0.06128, over 3083935.89 frames. ], batch size: 190, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:01:23,373 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190609.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:02:32,492 INFO [train.py:904] (5/8) Epoch 19, batch 7950, loss[loss=0.2088, simple_loss=0.2876, pruned_loss=0.06505, over 16537.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2926, pruned_loss=0.06176, over 3085019.84 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:02:33,057 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190652.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:02:54,097 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.686e+02 2.695e+02 3.262e+02 3.966e+02 7.542e+02, threshold=6.523e+02, percent-clipped=2.0 2023-04-30 23:03:01,441 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190670.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:03:04,352 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190672.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:03:08,661 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190675.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:03:49,826 INFO [train.py:904] (5/8) Epoch 19, batch 8000, loss[loss=0.2178, simple_loss=0.3126, pruned_loss=0.06155, over 16768.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2933, pruned_loss=0.06219, over 3086543.51 frames. ], batch size: 83, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:04:44,478 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190736.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:07,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7751, 1.8564, 2.4293, 2.7625, 2.6956, 3.1552, 2.1214, 3.2507], device='cuda:5'), covar=tensor([0.0240, 0.0516, 0.0312, 0.0314, 0.0306, 0.0183, 0.0474, 0.0111], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0188, 0.0173, 0.0177, 0.0188, 0.0147, 0.0189, 0.0141], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:05:07,645 INFO [train.py:904] (5/8) Epoch 19, batch 8050, loss[loss=0.2074, simple_loss=0.2883, pruned_loss=0.06325, over 16447.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2927, pruned_loss=0.06153, over 3098346.00 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:05:12,154 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190755.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:28,127 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.906e+02 2.693e+02 3.232e+02 4.056e+02 6.686e+02, threshold=6.463e+02, percent-clipped=1.0 2023-04-30 23:05:47,031 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1891, 4.0592, 4.2824, 4.4009, 4.5272, 4.0918, 4.4455, 4.5278], device='cuda:5'), covar=tensor([0.1789, 0.1243, 0.1463, 0.0708, 0.0616, 0.1299, 0.0860, 0.0678], device='cuda:5'), in_proj_covar=tensor([0.0601, 0.0743, 0.0870, 0.0760, 0.0574, 0.0596, 0.0620, 0.0713], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:06:12,402 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190795.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:06:22,319 INFO [train.py:904] (5/8) Epoch 19, batch 8100, loss[loss=0.1963, simple_loss=0.279, pruned_loss=0.05675, over 17004.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2921, pruned_loss=0.0607, over 3100611.63 frames. ], batch size: 55, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:06:31,256 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:07:25,607 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190843.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:07:39,353 INFO [train.py:904] (5/8) Epoch 19, batch 8150, loss[loss=0.1823, simple_loss=0.2704, pruned_loss=0.04708, over 16448.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2892, pruned_loss=0.05963, over 3096251.10 frames. ], batch size: 146, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:07:44,593 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:07:59,388 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5307, 3.7421, 2.5366, 2.2200, 2.3172, 2.1612, 3.8298, 3.1316], device='cuda:5'), covar=tensor([0.3280, 0.0763, 0.2294, 0.2766, 0.2929, 0.2390, 0.0638, 0.1495], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0264, 0.0298, 0.0304, 0.0293, 0.0250, 0.0288, 0.0326], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 23:08:01,319 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.839e+02 2.760e+02 3.332e+02 4.142e+02 6.995e+02, threshold=6.664e+02, percent-clipped=2.0 2023-04-30 23:08:57,396 INFO [train.py:904] (5/8) Epoch 19, batch 8200, loss[loss=0.1729, simple_loss=0.2634, pruned_loss=0.04115, over 16704.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2866, pruned_loss=0.05893, over 3104876.62 frames. ], batch size: 89, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:12,366 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190947.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 23:10:19,465 INFO [train.py:904] (5/8) Epoch 19, batch 8250, loss[loss=0.1927, simple_loss=0.2898, pruned_loss=0.04783, over 16179.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2856, pruned_loss=0.05663, over 3095195.21 frames. ], batch size: 165, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:39,911 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190965.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 23:10:41,156 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.606e+02 3.069e+02 3.589e+02 7.796e+02, threshold=6.137e+02, percent-clipped=2.0 2023-04-30 23:10:51,897 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:11:39,011 INFO [train.py:904] (5/8) Epoch 19, batch 8300, loss[loss=0.1893, simple_loss=0.2842, pruned_loss=0.0472, over 15353.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2831, pruned_loss=0.05397, over 3068791.41 frames. ], batch size: 190, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:12:10,265 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191020.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:12:27,714 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191031.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:00,830 INFO [train.py:904] (5/8) Epoch 19, batch 8350, loss[loss=0.1753, simple_loss=0.276, pruned_loss=0.03726, over 16865.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2824, pruned_loss=0.05237, over 3068844.91 frames. ], batch size: 96, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:13:04,385 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4092, 3.0321, 2.7145, 2.1754, 2.1742, 2.2065, 3.0381, 2.8980], device='cuda:5'), covar=tensor([0.2525, 0.0844, 0.1648, 0.2985, 0.2714, 0.2411, 0.0529, 0.1298], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0260, 0.0294, 0.0299, 0.0288, 0.0247, 0.0283, 0.0322], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 23:13:08,016 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191055.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:24,346 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.343e+02 2.792e+02 3.340e+02 8.341e+02, threshold=5.585e+02, percent-clipped=1.0 2023-04-30 23:13:28,654 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191068.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:14:22,705 INFO [train.py:904] (5/8) Epoch 19, batch 8400, loss[loss=0.1682, simple_loss=0.2705, pruned_loss=0.03294, over 16744.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.28, pruned_loss=0.05033, over 3071011.82 frames. ], batch size: 89, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:14:24,518 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:06,979 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:11,096 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191132.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:42,695 INFO [train.py:904] (5/8) Epoch 19, batch 8450, loss[loss=0.1734, simple_loss=0.2719, pruned_loss=0.03742, over 16707.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2785, pruned_loss=0.04875, over 3076299.71 frames. ], batch size: 134, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:16:06,336 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.247e+02 2.599e+02 3.228e+02 6.628e+02, threshold=5.197e+02, percent-clipped=1.0 2023-04-30 23:16:36,952 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:16:50,126 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:17:04,707 INFO [train.py:904] (5/8) Epoch 19, batch 8500, loss[loss=0.1777, simple_loss=0.2649, pruned_loss=0.04519, over 15360.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.275, pruned_loss=0.04659, over 3053969.61 frames. ], batch size: 191, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:17:18,663 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5792, 3.0023, 3.3714, 2.0386, 2.9091, 2.1429, 3.2189, 3.2299], device='cuda:5'), covar=tensor([0.0292, 0.0852, 0.0482, 0.2002, 0.0779, 0.1036, 0.0642, 0.0854], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0156, 0.0162, 0.0147, 0.0141, 0.0125, 0.0140, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 23:18:22,866 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191246.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:18:24,987 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191247.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 23:18:28,170 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8632, 2.7272, 2.8775, 2.1524, 2.6915, 2.2080, 2.6817, 2.9576], device='cuda:5'), covar=tensor([0.0362, 0.0956, 0.0505, 0.1806, 0.0801, 0.0941, 0.0635, 0.0721], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0156, 0.0162, 0.0147, 0.0141, 0.0125, 0.0140, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-04-30 23:18:32,476 INFO [train.py:904] (5/8) Epoch 19, batch 8550, loss[loss=0.1748, simple_loss=0.2631, pruned_loss=0.04324, over 16596.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2726, pruned_loss=0.04514, over 3043881.64 frames. ], batch size: 57, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:18:57,581 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191265.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:19:00,617 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.304e+02 2.724e+02 3.153e+02 7.683e+02, threshold=5.448e+02, percent-clipped=1.0 2023-04-30 23:19:58,310 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8948, 5.1891, 5.3801, 5.1847, 5.1988, 5.7522, 5.2657, 5.0577], device='cuda:5'), covar=tensor([0.0995, 0.2069, 0.2298, 0.2034, 0.2547, 0.0942, 0.1675, 0.2234], device='cuda:5'), in_proj_covar=tensor([0.0385, 0.0556, 0.0617, 0.0462, 0.0617, 0.0645, 0.0484, 0.0623], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 23:20:00,165 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191295.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:20:13,447 INFO [train.py:904] (5/8) Epoch 19, batch 8600, loss[loss=0.1706, simple_loss=0.2713, pruned_loss=0.03493, over 16816.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2724, pruned_loss=0.044, over 3038601.13 frames. ], batch size: 102, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:20:37,126 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:12,148 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:52,939 INFO [train.py:904] (5/8) Epoch 19, batch 8650, loss[loss=0.1746, simple_loss=0.2721, pruned_loss=0.03858, over 15426.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2708, pruned_loss=0.04254, over 3050554.74 frames. ], batch size: 192, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:22:25,736 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191365.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:22:29,947 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.144e+02 2.571e+02 3.162e+02 5.065e+02, threshold=5.143e+02, percent-clipped=0.0 2023-04-30 23:22:55,751 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:23:18,081 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 23:23:39,030 INFO [train.py:904] (5/8) Epoch 19, batch 8700, loss[loss=0.1746, simple_loss=0.2675, pruned_loss=0.04083, over 16147.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2681, pruned_loss=0.04145, over 3042075.15 frames. ], batch size: 165, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:24:20,731 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:24:23,709 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:24:42,946 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9659, 4.2706, 4.0883, 4.1175, 3.7797, 3.8505, 3.9184, 4.2502], device='cuda:5'), covar=tensor([0.1188, 0.0921, 0.0990, 0.0802, 0.0800, 0.1861, 0.1107, 0.1014], device='cuda:5'), in_proj_covar=tensor([0.0624, 0.0759, 0.0627, 0.0568, 0.0477, 0.0490, 0.0636, 0.0585], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:25:14,458 INFO [train.py:904] (5/8) Epoch 19, batch 8750, loss[loss=0.1528, simple_loss=0.2466, pruned_loss=0.02957, over 12156.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2673, pruned_loss=0.04072, over 3029108.76 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:25:28,146 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7510, 1.1977, 1.6765, 1.6115, 1.7471, 1.8100, 1.5303, 1.7736], device='cuda:5'), covar=tensor([0.0266, 0.0503, 0.0235, 0.0342, 0.0343, 0.0221, 0.0467, 0.0138], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0185, 0.0171, 0.0174, 0.0186, 0.0144, 0.0187, 0.0139], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:25:55,739 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3403, 3.0288, 2.7021, 2.1923, 2.1594, 2.2662, 3.0169, 2.8165], device='cuda:5'), covar=tensor([0.2630, 0.0730, 0.1563, 0.2585, 0.2579, 0.2195, 0.0447, 0.1472], device='cuda:5'), in_proj_covar=tensor([0.0316, 0.0257, 0.0292, 0.0297, 0.0283, 0.0245, 0.0279, 0.0319], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 23:25:58,342 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 2.109e+02 2.725e+02 3.287e+02 5.129e+02, threshold=5.449e+02, percent-clipped=0.0 2023-04-30 23:26:23,424 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 23:26:40,138 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191488.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:26:48,623 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 23:27:07,645 INFO [train.py:904] (5/8) Epoch 19, batch 8800, loss[loss=0.1745, simple_loss=0.271, pruned_loss=0.03896, over 16709.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2665, pruned_loss=0.03958, over 3062196.95 frames. ], batch size: 134, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:27:54,363 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8393, 3.8666, 3.9801, 3.8227, 3.8836, 4.3272, 3.9927, 3.8140], device='cuda:5'), covar=tensor([0.1977, 0.2013, 0.1866, 0.2252, 0.2630, 0.1476, 0.1511, 0.2266], device='cuda:5'), in_proj_covar=tensor([0.0382, 0.0550, 0.0610, 0.0458, 0.0608, 0.0637, 0.0478, 0.0613], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 23:28:18,329 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 23:28:29,217 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191541.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:28:39,199 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4093, 3.3249, 2.5196, 2.0746, 2.0923, 2.2312, 3.5250, 2.9670], device='cuda:5'), covar=tensor([0.2995, 0.0764, 0.1924, 0.2806, 0.2803, 0.2147, 0.0489, 0.1494], device='cuda:5'), in_proj_covar=tensor([0.0316, 0.0257, 0.0292, 0.0296, 0.0283, 0.0244, 0.0279, 0.0318], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 23:28:51,746 INFO [train.py:904] (5/8) Epoch 19, batch 8850, loss[loss=0.1738, simple_loss=0.2782, pruned_loss=0.03475, over 16262.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2695, pruned_loss=0.03907, over 3072654.28 frames. ], batch size: 166, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:29:28,628 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.105e+02 2.539e+02 3.345e+02 5.950e+02, threshold=5.078e+02, percent-clipped=2.0 2023-04-30 23:30:39,783 INFO [train.py:904] (5/8) Epoch 19, batch 8900, loss[loss=0.1622, simple_loss=0.2601, pruned_loss=0.03213, over 16692.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2699, pruned_loss=0.03871, over 3084300.69 frames. ], batch size: 89, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:32:29,941 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9984, 1.9913, 2.4763, 2.9346, 2.7185, 3.2653, 2.1847, 3.2786], device='cuda:5'), covar=tensor([0.0178, 0.0515, 0.0360, 0.0254, 0.0318, 0.0167, 0.0513, 0.0142], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0185, 0.0170, 0.0173, 0.0186, 0.0144, 0.0188, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:32:45,280 INFO [train.py:904] (5/8) Epoch 19, batch 8950, loss[loss=0.1582, simple_loss=0.2618, pruned_loss=0.02728, over 16385.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2691, pruned_loss=0.03866, over 3106768.27 frames. ], batch size: 146, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:33:21,178 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.162e+02 2.456e+02 2.877e+02 5.499e+02, threshold=4.911e+02, percent-clipped=2.0 2023-04-30 23:34:34,398 INFO [train.py:904] (5/8) Epoch 19, batch 9000, loss[loss=0.1698, simple_loss=0.2604, pruned_loss=0.03958, over 11915.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.266, pruned_loss=0.03771, over 3091916.40 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:34:34,398 INFO [train.py:929] (5/8) Computing validation loss 2023-04-30 23:34:44,207 INFO [train.py:938] (5/8) Epoch 19, validation: loss=0.1462, simple_loss=0.2506, pruned_loss=0.02087, over 944034.00 frames. 2023-04-30 23:34:44,208 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-04-30 23:35:06,454 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4216, 4.3656, 4.2956, 3.6977, 4.3110, 1.6448, 4.0720, 4.0956], device='cuda:5'), covar=tensor([0.0083, 0.0074, 0.0154, 0.0271, 0.0100, 0.2667, 0.0129, 0.0222], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0141, 0.0184, 0.0165, 0.0161, 0.0195, 0.0174, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:35:26,270 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:35:31,577 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:35:53,718 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3233, 3.7621, 3.8128, 2.6767, 3.4270, 3.8290, 3.5662, 2.0870], device='cuda:5'), covar=tensor([0.0477, 0.0038, 0.0039, 0.0318, 0.0079, 0.0062, 0.0066, 0.0454], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0077, 0.0078, 0.0131, 0.0092, 0.0103, 0.0089, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-04-30 23:36:06,005 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5325, 3.5272, 2.7953, 2.1730, 2.2417, 2.3195, 3.6676, 3.1698], device='cuda:5'), covar=tensor([0.2791, 0.0608, 0.1687, 0.2708, 0.2604, 0.2050, 0.0407, 0.1317], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0258, 0.0293, 0.0297, 0.0283, 0.0246, 0.0281, 0.0319], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-04-30 23:36:27,799 INFO [train.py:904] (5/8) Epoch 19, batch 9050, loss[loss=0.1934, simple_loss=0.2775, pruned_loss=0.05458, over 12790.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2668, pruned_loss=0.03848, over 3072538.13 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:36:59,117 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5373, 4.4961, 4.3978, 3.8138, 4.4428, 1.7671, 4.2064, 4.1925], device='cuda:5'), covar=tensor([0.0134, 0.0136, 0.0214, 0.0287, 0.0148, 0.2671, 0.0170, 0.0243], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0141, 0.0185, 0.0166, 0.0162, 0.0197, 0.0175, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:36:59,500 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 23:37:04,260 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.085e+02 2.454e+02 3.075e+02 7.905e+02, threshold=4.907e+02, percent-clipped=4.0 2023-04-30 23:37:09,314 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:37:14,189 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6300, 3.6718, 3.4383, 3.1426, 3.2138, 3.5821, 3.3354, 3.3785], device='cuda:5'), covar=tensor([0.0504, 0.0608, 0.0296, 0.0252, 0.0522, 0.0489, 0.1289, 0.0533], device='cuda:5'), in_proj_covar=tensor([0.0270, 0.0390, 0.0316, 0.0308, 0.0323, 0.0359, 0.0219, 0.0377], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:37:17,333 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5214, 5.9163, 5.6767, 5.6922, 5.2859, 5.3376, 5.2642, 6.0021], device='cuda:5'), covar=tensor([0.1159, 0.0793, 0.0900, 0.0760, 0.0762, 0.0618, 0.1194, 0.0769], device='cuda:5'), in_proj_covar=tensor([0.0622, 0.0759, 0.0624, 0.0566, 0.0477, 0.0488, 0.0635, 0.0584], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:37:39,985 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:38:10,483 INFO [train.py:904] (5/8) Epoch 19, batch 9100, loss[loss=0.1717, simple_loss=0.2582, pruned_loss=0.04256, over 12236.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2662, pruned_loss=0.03907, over 3065831.36 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:39:32,435 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191836.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:39:43,948 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191841.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:40:08,201 INFO [train.py:904] (5/8) Epoch 19, batch 9150, loss[loss=0.1748, simple_loss=0.268, pruned_loss=0.04081, over 16400.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2665, pruned_loss=0.03903, over 3056126.00 frames. ], batch size: 146, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:40:46,346 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.243e+02 2.682e+02 3.704e+02 5.734e+02, threshold=5.364e+02, percent-clipped=7.0 2023-04-30 23:41:25,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0306, 4.0845, 4.4162, 4.4047, 4.4039, 4.1597, 4.1466, 4.1550], device='cuda:5'), covar=tensor([0.0340, 0.0646, 0.0489, 0.0425, 0.0445, 0.0481, 0.0857, 0.0440], device='cuda:5'), in_proj_covar=tensor([0.0382, 0.0415, 0.0408, 0.0382, 0.0449, 0.0426, 0.0517, 0.0344], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 23:41:30,832 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:41:52,741 INFO [train.py:904] (5/8) Epoch 19, batch 9200, loss[loss=0.1663, simple_loss=0.259, pruned_loss=0.03678, over 16697.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2622, pruned_loss=0.03778, over 3060061.42 frames. ], batch size: 62, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:42:22,654 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8377, 3.7123, 4.0779, 2.1094, 4.2361, 4.3196, 3.1726, 3.2178], device='cuda:5'), covar=tensor([0.0677, 0.0240, 0.0182, 0.1086, 0.0067, 0.0102, 0.0356, 0.0418], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0102, 0.0091, 0.0134, 0.0074, 0.0115, 0.0121, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-04-30 23:42:39,152 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 23:43:29,335 INFO [train.py:904] (5/8) Epoch 19, batch 9250, loss[loss=0.1547, simple_loss=0.2533, pruned_loss=0.02805, over 16885.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2622, pruned_loss=0.03779, over 3064444.39 frames. ], batch size: 96, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:44:05,910 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.331e+02 2.686e+02 3.317e+02 7.389e+02, threshold=5.371e+02, percent-clipped=4.0 2023-04-30 23:44:42,386 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6582, 5.6675, 5.5138, 5.0259, 5.1631, 5.5576, 5.4998, 5.2137], device='cuda:5'), covar=tensor([0.0542, 0.0529, 0.0220, 0.0268, 0.0873, 0.0541, 0.0212, 0.0610], device='cuda:5'), in_proj_covar=tensor([0.0270, 0.0388, 0.0315, 0.0306, 0.0322, 0.0359, 0.0219, 0.0375], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:45:23,377 INFO [train.py:904] (5/8) Epoch 19, batch 9300, loss[loss=0.1604, simple_loss=0.2518, pruned_loss=0.03445, over 16911.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2606, pruned_loss=0.03752, over 3041504.78 frames. ], batch size: 125, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:46:09,700 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:47:09,794 INFO [train.py:904] (5/8) Epoch 19, batch 9350, loss[loss=0.1668, simple_loss=0.2617, pruned_loss=0.03591, over 15205.00 frames. ], tot_loss[loss=0.168, simple_loss=0.261, pruned_loss=0.0375, over 3058432.55 frames. ], batch size: 190, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:47:28,508 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7867, 3.8581, 4.1253, 4.1231, 4.1160, 3.9082, 3.9002, 3.9622], device='cuda:5'), covar=tensor([0.0578, 0.1413, 0.0781, 0.0768, 0.0759, 0.0937, 0.1133, 0.0626], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0410, 0.0404, 0.0379, 0.0446, 0.0422, 0.0511, 0.0341], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 23:47:46,416 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:47:47,133 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 1.981e+02 2.500e+02 3.039e+02 5.486e+02, threshold=4.999e+02, percent-clipped=1.0 2023-04-30 23:48:22,366 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1205, 2.0940, 2.0963, 3.7373, 2.0443, 2.4153, 2.2248, 2.1997], device='cuda:5'), covar=tensor([0.1288, 0.3667, 0.3034, 0.0541, 0.4157, 0.2652, 0.3465, 0.3478], device='cuda:5'), in_proj_covar=tensor([0.0380, 0.0423, 0.0351, 0.0311, 0.0422, 0.0485, 0.0395, 0.0492], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:48:37,354 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6888, 2.6359, 1.8953, 2.8643, 2.0875, 2.8431, 2.1283, 2.3949], device='cuda:5'), covar=tensor([0.0337, 0.0365, 0.1342, 0.0279, 0.0768, 0.0470, 0.1287, 0.0666], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0168, 0.0187, 0.0149, 0.0169, 0.0203, 0.0194, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-04-30 23:48:49,319 INFO [train.py:904] (5/8) Epoch 19, batch 9400, loss[loss=0.153, simple_loss=0.2401, pruned_loss=0.033, over 12515.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2614, pruned_loss=0.03761, over 3061340.50 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:48:50,412 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:29,962 INFO [train.py:904] (5/8) Epoch 19, batch 9450, loss[loss=0.1663, simple_loss=0.2634, pruned_loss=0.03458, over 15385.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2635, pruned_loss=0.03789, over 3070333.66 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:50:35,889 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:43,840 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9592, 3.9804, 4.3248, 4.3161, 4.3081, 4.0750, 4.0697, 4.0856], device='cuda:5'), covar=tensor([0.0500, 0.1136, 0.0679, 0.0817, 0.0759, 0.0800, 0.1006, 0.0510], device='cuda:5'), in_proj_covar=tensor([0.0378, 0.0410, 0.0403, 0.0378, 0.0445, 0.0421, 0.0509, 0.0340], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-04-30 23:50:51,063 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192163.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:51:05,243 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.168e+02 2.516e+02 3.148e+02 5.649e+02, threshold=5.031e+02, percent-clipped=1.0 2023-04-30 23:51:36,559 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 2023-04-30 23:52:10,474 INFO [train.py:904] (5/8) Epoch 19, batch 9500, loss[loss=0.1666, simple_loss=0.2539, pruned_loss=0.03967, over 17036.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2624, pruned_loss=0.03736, over 3069645.92 frames. ], batch size: 55, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:52:39,602 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:45,105 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:47,826 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-30 23:52:49,330 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:53:27,513 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4499, 2.0279, 1.8469, 1.7205, 2.2732, 1.9547, 1.9143, 2.3732], device='cuda:5'), covar=tensor([0.0172, 0.0402, 0.0446, 0.0469, 0.0259, 0.0370, 0.0187, 0.0259], device='cuda:5'), in_proj_covar=tensor([0.0189, 0.0223, 0.0215, 0.0216, 0.0225, 0.0222, 0.0221, 0.0214], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:53:55,160 INFO [train.py:904] (5/8) Epoch 19, batch 9550, loss[loss=0.1665, simple_loss=0.2606, pruned_loss=0.03618, over 16709.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2625, pruned_loss=0.03736, over 3083516.94 frames. ], batch size: 76, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:54:34,520 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 2.006e+02 2.351e+02 2.778e+02 5.702e+02, threshold=4.701e+02, percent-clipped=1.0 2023-04-30 23:54:55,669 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:01,200 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:11,051 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4757, 2.0140, 1.7780, 1.7267, 2.2678, 1.9540, 1.9432, 2.3937], device='cuda:5'), covar=tensor([0.0152, 0.0413, 0.0492, 0.0478, 0.0260, 0.0373, 0.0190, 0.0250], device='cuda:5'), in_proj_covar=tensor([0.0189, 0.0224, 0.0215, 0.0217, 0.0225, 0.0222, 0.0221, 0.0215], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-04-30 23:55:38,440 INFO [train.py:904] (5/8) Epoch 19, batch 9600, loss[loss=0.1674, simple_loss=0.2574, pruned_loss=0.03865, over 12465.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2634, pruned_loss=0.03815, over 3049234.24 frames. ], batch size: 247, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:56:53,378 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7238, 4.7272, 4.5692, 4.1647, 4.2441, 4.6523, 4.5063, 4.3381], device='cuda:5'), covar=tensor([0.0590, 0.0689, 0.0323, 0.0332, 0.0945, 0.0582, 0.0385, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0267, 0.0382, 0.0311, 0.0302, 0.0318, 0.0355, 0.0216, 0.0371], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-04-30 23:57:27,649 INFO [train.py:904] (5/8) Epoch 19, batch 9650, loss[loss=0.1798, simple_loss=0.276, pruned_loss=0.04176, over 16346.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2649, pruned_loss=0.03811, over 3064093.53 frames. ], batch size: 146, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:58:09,765 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.165e+02 2.448e+02 2.941e+02 5.888e+02, threshold=4.896e+02, percent-clipped=1.0 2023-04-30 23:58:54,548 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:09,401 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:15,180 INFO [train.py:904] (5/8) Epoch 19, batch 9700, loss[loss=0.1588, simple_loss=0.2515, pruned_loss=0.03308, over 12573.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2639, pruned_loss=0.03788, over 3060498.72 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:00:14,543 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2887, 2.1417, 2.1538, 3.8440, 2.0509, 2.4002, 2.2572, 2.2568], device='cuda:5'), covar=tensor([0.1140, 0.3798, 0.3081, 0.0503, 0.4721, 0.2771, 0.3854, 0.3566], device='cuda:5'), in_proj_covar=tensor([0.0380, 0.0423, 0.0350, 0.0309, 0.0422, 0.0484, 0.0395, 0.0492], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:00:19,293 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192433.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:00:57,137 INFO [train.py:904] (5/8) Epoch 19, batch 9750, loss[loss=0.155, simple_loss=0.2593, pruned_loss=0.02537, over 16873.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2637, pruned_loss=0.03833, over 3063927.82 frames. ], batch size: 102, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:00:59,789 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:09,810 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:12,232 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5292, 4.0297, 3.9581, 2.7981, 3.5573, 3.9638, 3.7831, 2.2385], device='cuda:5'), covar=tensor([0.0487, 0.0034, 0.0040, 0.0360, 0.0091, 0.0080, 0.0061, 0.0501], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0077, 0.0077, 0.0131, 0.0092, 0.0103, 0.0089, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:01:14,086 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:31,659 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.106e+02 2.517e+02 3.092e+02 5.794e+02, threshold=5.033e+02, percent-clipped=3.0 2023-05-01 00:02:23,658 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:02:37,070 INFO [train.py:904] (5/8) Epoch 19, batch 9800, loss[loss=0.1604, simple_loss=0.2639, pruned_loss=0.02846, over 16785.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2634, pruned_loss=0.03726, over 3056385.16 frames. ], batch size: 76, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:02:44,399 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-01 00:02:48,181 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3874, 3.0225, 2.6556, 2.2319, 2.1103, 2.2108, 2.9923, 2.7975], device='cuda:5'), covar=tensor([0.2578, 0.0686, 0.1626, 0.2734, 0.2699, 0.2227, 0.0444, 0.1385], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0255, 0.0289, 0.0295, 0.0277, 0.0243, 0.0277, 0.0315], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:02:55,416 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192511.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:04:09,417 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4104, 2.9059, 3.0988, 1.7429, 2.7002, 1.7494, 3.1162, 3.0033], device='cuda:5'), covar=tensor([0.0289, 0.0807, 0.0643, 0.2330, 0.0945, 0.1237, 0.0669, 0.0880], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0150, 0.0160, 0.0146, 0.0139, 0.0124, 0.0138, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:04:23,520 INFO [train.py:904] (5/8) Epoch 19, batch 9850, loss[loss=0.1699, simple_loss=0.2634, pruned_loss=0.03817, over 15425.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2638, pruned_loss=0.03676, over 3061483.37 frames. ], batch size: 190, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:05:00,570 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 1.961e+02 2.506e+02 2.957e+02 4.630e+02, threshold=5.011e+02, percent-clipped=0.0 2023-05-01 00:05:11,440 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192575.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:05:15,472 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:06:14,640 INFO [train.py:904] (5/8) Epoch 19, batch 9900, loss[loss=0.191, simple_loss=0.2921, pruned_loss=0.04494, over 16884.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.264, pruned_loss=0.03664, over 3049837.96 frames. ], batch size: 116, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:06:47,325 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:07:25,638 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8270, 3.7595, 3.9454, 3.7484, 3.8675, 4.2866, 3.9751, 3.6891], device='cuda:5'), covar=tensor([0.2152, 0.2477, 0.2374, 0.2372, 0.2879, 0.1641, 0.1584, 0.2459], device='cuda:5'), in_proj_covar=tensor([0.0373, 0.0541, 0.0598, 0.0449, 0.0600, 0.0628, 0.0469, 0.0600], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:08:13,296 INFO [train.py:904] (5/8) Epoch 19, batch 9950, loss[loss=0.1529, simple_loss=0.2578, pruned_loss=0.02397, over 16887.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2662, pruned_loss=0.03697, over 3049416.05 frames. ], batch size: 102, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:08:44,353 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9228, 4.2426, 4.0693, 4.1128, 3.7765, 3.8352, 3.8360, 4.2364], device='cuda:5'), covar=tensor([0.1321, 0.1024, 0.1062, 0.0760, 0.0826, 0.1608, 0.0992, 0.1012], device='cuda:5'), in_proj_covar=tensor([0.0615, 0.0757, 0.0614, 0.0560, 0.0473, 0.0482, 0.0628, 0.0578], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:08:54,790 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.121e+02 2.565e+02 3.086e+02 6.326e+02, threshold=5.129e+02, percent-clipped=1.0 2023-05-01 00:09:04,440 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:09:15,088 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:10:14,417 INFO [train.py:904] (5/8) Epoch 19, batch 10000, loss[loss=0.175, simple_loss=0.2775, pruned_loss=0.03621, over 16504.00 frames. ], tot_loss[loss=0.169, simple_loss=0.265, pruned_loss=0.03651, over 3064316.61 frames. ], batch size: 147, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:11:22,358 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192734.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:25,776 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192735.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:50,859 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192748.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:56,967 INFO [train.py:904] (5/8) Epoch 19, batch 10050, loss[loss=0.1665, simple_loss=0.266, pruned_loss=0.03345, over 16898.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2654, pruned_loss=0.03676, over 3062445.74 frames. ], batch size: 96, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:12:02,958 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192755.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:09,733 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:27,604 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2276, 2.1156, 2.0872, 3.8383, 2.0693, 2.4360, 2.1756, 2.2517], device='cuda:5'), covar=tensor([0.1271, 0.3671, 0.3068, 0.0487, 0.4192, 0.2625, 0.3527, 0.3551], device='cuda:5'), in_proj_covar=tensor([0.0379, 0.0420, 0.0348, 0.0308, 0.0420, 0.0480, 0.0391, 0.0488], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:12:32,889 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.015e+02 2.519e+02 2.909e+02 8.183e+02, threshold=5.037e+02, percent-clipped=1.0 2023-05-01 00:12:38,162 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3647, 4.1646, 4.3339, 4.5511, 4.6810, 4.2558, 4.7104, 4.7045], device='cuda:5'), covar=tensor([0.1875, 0.1449, 0.1871, 0.0889, 0.0663, 0.1083, 0.0703, 0.0786], device='cuda:5'), in_proj_covar=tensor([0.0578, 0.0716, 0.0836, 0.0734, 0.0549, 0.0573, 0.0594, 0.0687], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:12:42,785 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6038, 3.6516, 3.4536, 3.0841, 3.2651, 3.5580, 3.3382, 3.3492], device='cuda:5'), covar=tensor([0.0633, 0.0785, 0.0285, 0.0268, 0.0517, 0.0574, 0.1208, 0.0501], device='cuda:5'), in_proj_covar=tensor([0.0266, 0.0378, 0.0309, 0.0299, 0.0315, 0.0351, 0.0213, 0.0367], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 00:13:08,139 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:21,299 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192796.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:28,580 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2437, 3.4641, 3.4812, 2.3962, 3.1497, 3.5124, 3.3217, 2.0733], device='cuda:5'), covar=tensor([0.0484, 0.0059, 0.0049, 0.0387, 0.0119, 0.0091, 0.0093, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0131, 0.0077, 0.0077, 0.0130, 0.0092, 0.0102, 0.0088, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:13:28,774 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-05-01 00:13:30,107 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 00:13:30,580 INFO [train.py:904] (5/8) Epoch 19, batch 10100, loss[loss=0.1574, simple_loss=0.2491, pruned_loss=0.03286, over 16477.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2652, pruned_loss=0.03699, over 3064454.49 frames. ], batch size: 68, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:13:35,251 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4597, 2.9127, 3.1264, 1.9738, 2.7876, 2.1167, 2.9619, 3.1173], device='cuda:5'), covar=tensor([0.0301, 0.0808, 0.0485, 0.1936, 0.0785, 0.1006, 0.0729, 0.0920], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0150, 0.0159, 0.0146, 0.0139, 0.0123, 0.0137, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:13:37,805 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3665, 4.6945, 4.5048, 4.4755, 4.2107, 4.1848, 4.1923, 4.7234], device='cuda:5'), covar=tensor([0.1212, 0.0907, 0.0980, 0.0802, 0.0814, 0.1396, 0.1061, 0.0931], device='cuda:5'), in_proj_covar=tensor([0.0613, 0.0754, 0.0612, 0.0560, 0.0473, 0.0482, 0.0628, 0.0578], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:13:39,009 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:48,429 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192811.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:14:28,848 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192831.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:14:29,692 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7055, 4.8659, 4.9871, 4.8162, 4.8630, 5.3849, 4.8865, 4.6431], device='cuda:5'), covar=tensor([0.1003, 0.1603, 0.1776, 0.1867, 0.2259, 0.0918, 0.1542, 0.2231], device='cuda:5'), in_proj_covar=tensor([0.0372, 0.0537, 0.0596, 0.0449, 0.0599, 0.0627, 0.0470, 0.0597], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:15:13,703 INFO [train.py:904] (5/8) Epoch 20, batch 0, loss[loss=0.2434, simple_loss=0.302, pruned_loss=0.09241, over 16432.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.302, pruned_loss=0.09241, over 16432.00 frames. ], batch size: 146, lr: 3.43e-03, grad_scale: 8.0 2023-05-01 00:15:13,703 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 00:15:21,163 INFO [train.py:938] (5/8) Epoch 20, validation: loss=0.146, simple_loss=0.2496, pruned_loss=0.02121, over 944034.00 frames. 2023-05-01 00:15:21,164 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 00:15:32,534 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:36,711 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8304, 3.7841, 4.2085, 2.0694, 4.3225, 4.3271, 3.1300, 3.3370], device='cuda:5'), covar=tensor([0.0683, 0.0209, 0.0146, 0.1187, 0.0061, 0.0126, 0.0420, 0.0374], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0101, 0.0089, 0.0133, 0.0074, 0.0114, 0.0120, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 00:15:49,198 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.412e+02 2.946e+02 3.638e+02 7.163e+02, threshold=5.893e+02, percent-clipped=6.0 2023-05-01 00:15:53,696 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192875.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:56,058 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:16:17,614 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192892.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:16:31,054 INFO [train.py:904] (5/8) Epoch 20, batch 50, loss[loss=0.1964, simple_loss=0.2791, pruned_loss=0.05686, over 15459.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2764, pruned_loss=0.0532, over 743921.20 frames. ], batch size: 190, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:17:00,247 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192923.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:03,081 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:21,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7656, 4.6831, 4.6309, 4.2338, 4.6781, 1.8232, 4.4283, 4.3875], device='cuda:5'), covar=tensor([0.0140, 0.0103, 0.0213, 0.0295, 0.0129, 0.2722, 0.0157, 0.0243], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0142, 0.0184, 0.0163, 0.0161, 0.0198, 0.0174, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:17:38,941 INFO [train.py:904] (5/8) Epoch 20, batch 100, loss[loss=0.1741, simple_loss=0.2604, pruned_loss=0.04386, over 17228.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2696, pruned_loss=0.04882, over 1316173.71 frames. ], batch size: 43, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:17:48,086 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2131, 5.5592, 5.3124, 5.3201, 5.0049, 4.9777, 4.9736, 5.6877], device='cuda:5'), covar=tensor([0.1143, 0.0987, 0.1297, 0.0935, 0.0946, 0.0917, 0.1204, 0.0974], device='cuda:5'), in_proj_covar=tensor([0.0627, 0.0774, 0.0627, 0.0573, 0.0484, 0.0494, 0.0645, 0.0590], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:18:07,335 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.214e+02 2.624e+02 3.260e+02 6.598e+02, threshold=5.249e+02, percent-clipped=3.0 2023-05-01 00:18:07,676 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192972.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:18:34,904 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1806, 5.1109, 5.0286, 4.4515, 4.6342, 5.0456, 5.0701, 4.6402], device='cuda:5'), covar=tensor([0.0575, 0.0481, 0.0308, 0.0392, 0.1090, 0.0498, 0.0286, 0.0752], device='cuda:5'), in_proj_covar=tensor([0.0271, 0.0387, 0.0316, 0.0306, 0.0324, 0.0359, 0.0218, 0.0377], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:18:37,558 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 00:18:41,878 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6608, 3.6284, 2.1766, 3.8738, 2.8428, 3.7789, 2.2647, 2.9606], device='cuda:5'), covar=tensor([0.0249, 0.0474, 0.1575, 0.0315, 0.0746, 0.0768, 0.1435, 0.0639], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0171, 0.0189, 0.0152, 0.0172, 0.0207, 0.0198, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:18:48,442 INFO [train.py:904] (5/8) Epoch 20, batch 150, loss[loss=0.165, simple_loss=0.2517, pruned_loss=0.03918, over 17228.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2678, pruned_loss=0.04811, over 1764694.62 frames. ], batch size: 44, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:19:01,787 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 00:19:27,938 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193029.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:43,986 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:52,843 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:58,179 INFO [train.py:904] (5/8) Epoch 20, batch 200, loss[loss=0.1343, simple_loss=0.2166, pruned_loss=0.02598, over 16783.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.268, pruned_loss=0.04821, over 2104632.45 frames. ], batch size: 39, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:20:03,488 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193055.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:05,074 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-01 00:20:27,499 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.769e+02 2.323e+02 2.688e+02 3.539e+02 1.444e+03, threshold=5.377e+02, percent-clipped=5.0 2023-05-01 00:20:50,832 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:53,804 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193091.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:00,959 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:08,807 INFO [train.py:904] (5/8) Epoch 20, batch 250, loss[loss=0.1638, simple_loss=0.2472, pruned_loss=0.04024, over 15681.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2639, pruned_loss=0.04732, over 2377578.60 frames. ], batch size: 191, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:21:09,311 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193102.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:10,234 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:10,587 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.09 vs. limit=5.0 2023-05-01 00:21:12,327 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5270, 3.4871, 2.7529, 2.1657, 2.2393, 2.1919, 3.5421, 3.0300], device='cuda:5'), covar=tensor([0.2840, 0.0671, 0.1776, 0.3036, 0.2822, 0.2375, 0.0510, 0.1585], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0260, 0.0295, 0.0300, 0.0283, 0.0248, 0.0282, 0.0324], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:21:58,202 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193137.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:22:17,134 INFO [train.py:904] (5/8) Epoch 20, batch 300, loss[loss=0.1639, simple_loss=0.2435, pruned_loss=0.04217, over 16990.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2607, pruned_loss=0.04577, over 2591909.48 frames. ], batch size: 41, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:22:43,901 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1824, 5.9149, 6.0045, 5.7170, 5.7610, 6.3760, 5.8535, 5.5836], device='cuda:5'), covar=tensor([0.0906, 0.1918, 0.2180, 0.1983, 0.2657, 0.0855, 0.1505, 0.2276], device='cuda:5'), in_proj_covar=tensor([0.0394, 0.0570, 0.0631, 0.0475, 0.0635, 0.0658, 0.0495, 0.0631], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:22:46,395 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.161e+02 2.496e+02 2.884e+02 5.011e+02, threshold=4.993e+02, percent-clipped=0.0 2023-05-01 00:23:05,608 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193187.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:23:10,486 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7789, 3.7990, 2.4854, 4.4457, 2.9557, 4.4047, 2.4932, 3.0788], device='cuda:5'), covar=tensor([0.0312, 0.0431, 0.1540, 0.0314, 0.0927, 0.0575, 0.1548, 0.0829], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0157, 0.0176, 0.0212, 0.0202, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:23:28,169 INFO [train.py:904] (5/8) Epoch 20, batch 350, loss[loss=0.1837, simple_loss=0.2515, pruned_loss=0.05791, over 16940.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2577, pruned_loss=0.04433, over 2756398.97 frames. ], batch size: 109, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:23:39,620 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:24:13,916 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8395, 5.1584, 5.2777, 5.1075, 5.1319, 5.7256, 5.1818, 4.9338], device='cuda:5'), covar=tensor([0.1252, 0.2121, 0.2616, 0.2186, 0.2653, 0.1066, 0.1821, 0.2668], device='cuda:5'), in_proj_covar=tensor([0.0398, 0.0575, 0.0639, 0.0480, 0.0642, 0.0667, 0.0501, 0.0639], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:24:33,452 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3753, 4.3986, 4.7287, 4.7415, 4.7921, 4.4578, 4.4636, 4.3307], device='cuda:5'), covar=tensor([0.0386, 0.0777, 0.0465, 0.0412, 0.0420, 0.0470, 0.0865, 0.0673], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0429, 0.0417, 0.0390, 0.0463, 0.0439, 0.0529, 0.0355], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 00:24:37,906 INFO [train.py:904] (5/8) Epoch 20, batch 400, loss[loss=0.1618, simple_loss=0.2377, pruned_loss=0.04295, over 16802.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2565, pruned_loss=0.0444, over 2881240.01 frames. ], batch size: 83, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:25:05,927 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193272.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:06,033 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193272.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:25:06,673 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.085e+02 2.558e+02 2.961e+02 5.266e+02, threshold=5.116e+02, percent-clipped=1.0 2023-05-01 00:25:23,385 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193284.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:46,658 INFO [train.py:904] (5/8) Epoch 20, batch 450, loss[loss=0.1518, simple_loss=0.2408, pruned_loss=0.03141, over 16013.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2557, pruned_loss=0.04334, over 2977439.34 frames. ], batch size: 35, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:26:01,935 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5506, 3.4955, 3.7835, 2.6694, 3.4548, 3.8401, 3.6144, 2.2035], device='cuda:5'), covar=tensor([0.0462, 0.0176, 0.0048, 0.0359, 0.0103, 0.0090, 0.0087, 0.0464], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0081, 0.0080, 0.0134, 0.0095, 0.0106, 0.0092, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:26:14,331 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:26,083 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:48,834 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:57,293 INFO [train.py:904] (5/8) Epoch 20, batch 500, loss[loss=0.1445, simple_loss=0.232, pruned_loss=0.02849, over 17252.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2543, pruned_loss=0.04251, over 3054713.38 frames. ], batch size: 45, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:27:26,054 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.420e+02 2.231e+02 2.613e+02 3.252e+02 5.197e+02, threshold=5.226e+02, percent-clipped=2.0 2023-05-01 00:27:31,762 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:51,578 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193391.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:59,426 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:07,100 INFO [train.py:904] (5/8) Epoch 20, batch 550, loss[loss=0.1609, simple_loss=0.2501, pruned_loss=0.03581, over 16677.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2534, pruned_loss=0.04211, over 3100291.66 frames. ], batch size: 62, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:28:28,549 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0607, 2.0481, 2.2376, 3.6269, 2.1138, 2.3526, 2.1758, 2.1879], device='cuda:5'), covar=tensor([0.1448, 0.3486, 0.2898, 0.0679, 0.3866, 0.2406, 0.3579, 0.3211], device='cuda:5'), in_proj_covar=tensor([0.0392, 0.0435, 0.0361, 0.0321, 0.0432, 0.0499, 0.0405, 0.0507], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:28:32,341 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:42,644 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6274, 4.5245, 4.5424, 4.2493, 4.1953, 4.5814, 4.3889, 4.3180], device='cuda:5'), covar=tensor([0.0937, 0.1263, 0.0408, 0.0367, 0.1060, 0.0723, 0.0546, 0.0804], device='cuda:5'), in_proj_covar=tensor([0.0286, 0.0410, 0.0332, 0.0325, 0.0343, 0.0380, 0.0231, 0.0399], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:28:56,784 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:29:14,814 INFO [train.py:904] (5/8) Epoch 20, batch 600, loss[loss=0.1473, simple_loss=0.2298, pruned_loss=0.03242, over 16768.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2524, pruned_loss=0.04248, over 3130626.96 frames. ], batch size: 39, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:29:18,283 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193454.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:29:43,214 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.108e+02 2.456e+02 2.826e+02 6.166e+02, threshold=4.913e+02, percent-clipped=1.0 2023-05-01 00:29:55,260 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193482.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:30:03,489 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193487.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:30:21,023 INFO [train.py:904] (5/8) Epoch 20, batch 650, loss[loss=0.1596, simple_loss=0.2543, pruned_loss=0.03246, over 17284.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2513, pruned_loss=0.04183, over 3170704.93 frames. ], batch size: 52, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:30:39,994 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193515.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:31:07,551 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:31:11,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1481, 3.1975, 3.4442, 2.1399, 3.0267, 2.3793, 3.6944, 3.4749], device='cuda:5'), covar=tensor([0.0225, 0.0920, 0.0602, 0.1944, 0.0788, 0.1019, 0.0480, 0.0928], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0158, 0.0165, 0.0151, 0.0144, 0.0127, 0.0142, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:31:29,639 INFO [train.py:904] (5/8) Epoch 20, batch 700, loss[loss=0.1636, simple_loss=0.2567, pruned_loss=0.0352, over 17197.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2512, pruned_loss=0.04147, over 3199834.67 frames. ], batch size: 46, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:31:48,973 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193567.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:31:57,465 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.109e+02 2.476e+02 3.086e+02 5.855e+02, threshold=4.951e+02, percent-clipped=1.0 2023-05-01 00:32:35,679 INFO [train.py:904] (5/8) Epoch 20, batch 750, loss[loss=0.1716, simple_loss=0.2462, pruned_loss=0.04853, over 16865.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2511, pruned_loss=0.04115, over 3228157.84 frames. ], batch size: 96, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:32:54,601 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:28,438 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:42,958 INFO [train.py:904] (5/8) Epoch 20, batch 800, loss[loss=0.1621, simple_loss=0.2436, pruned_loss=0.04032, over 16446.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.251, pruned_loss=0.04059, over 3249078.07 frames. ], batch size: 68, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:34:10,662 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.187e+02 2.610e+02 3.095e+02 1.121e+03, threshold=5.220e+02, percent-clipped=2.0 2023-05-01 00:34:16,630 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:31,754 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4005, 5.3522, 5.1703, 4.6978, 5.1833, 2.2171, 4.9909, 5.1467], device='cuda:5'), covar=tensor([0.0062, 0.0072, 0.0191, 0.0332, 0.0094, 0.2367, 0.0129, 0.0180], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0147, 0.0192, 0.0170, 0.0167, 0.0203, 0.0181, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:34:45,140 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:52,181 INFO [train.py:904] (5/8) Epoch 20, batch 850, loss[loss=0.1677, simple_loss=0.2597, pruned_loss=0.0379, over 16720.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2499, pruned_loss=0.04009, over 3268553.14 frames. ], batch size: 57, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:35:21,441 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4498, 5.4297, 5.1886, 4.7191, 5.2499, 2.1151, 5.0235, 5.1870], device='cuda:5'), covar=tensor([0.0072, 0.0061, 0.0201, 0.0334, 0.0092, 0.2392, 0.0117, 0.0165], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0148, 0.0192, 0.0171, 0.0168, 0.0204, 0.0181, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:35:51,454 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:35:58,977 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 00:35:59,341 INFO [train.py:904] (5/8) Epoch 20, batch 900, loss[loss=0.1998, simple_loss=0.2737, pruned_loss=0.06298, over 16694.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2497, pruned_loss=0.03973, over 3276826.02 frames. ], batch size: 124, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:36:25,175 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3645, 3.5251, 3.6837, 3.6558, 3.6600, 3.5002, 3.5193, 3.5647], device='cuda:5'), covar=tensor([0.0472, 0.0829, 0.0543, 0.0496, 0.0576, 0.0527, 0.0768, 0.0510], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0441, 0.0428, 0.0400, 0.0478, 0.0449, 0.0542, 0.0363], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 00:36:28,230 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.015e+02 2.385e+02 2.719e+02 5.424e+02, threshold=4.769e+02, percent-clipped=3.0 2023-05-01 00:36:33,735 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193777.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:36:47,601 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5920, 1.7472, 2.2574, 2.3705, 2.5463, 2.5294, 1.8869, 2.7210], device='cuda:5'), covar=tensor([0.0185, 0.0472, 0.0306, 0.0305, 0.0296, 0.0303, 0.0458, 0.0156], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0191, 0.0177, 0.0178, 0.0193, 0.0150, 0.0192, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:36:48,557 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2018, 5.9274, 6.0040, 5.7134, 5.8242, 6.3846, 5.8764, 5.5537], device='cuda:5'), covar=tensor([0.0961, 0.2078, 0.2430, 0.2270, 0.2628, 0.0884, 0.1515, 0.2405], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0585, 0.0644, 0.0487, 0.0651, 0.0674, 0.0504, 0.0646], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 00:37:09,298 INFO [train.py:904] (5/8) Epoch 20, batch 950, loss[loss=0.1557, simple_loss=0.2466, pruned_loss=0.03239, over 17189.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2502, pruned_loss=0.03999, over 3282153.47 frames. ], batch size: 46, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:37:20,497 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193810.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:37:58,520 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 00:38:17,863 INFO [train.py:904] (5/8) Epoch 20, batch 1000, loss[loss=0.1543, simple_loss=0.251, pruned_loss=0.02878, over 17122.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2483, pruned_loss=0.03963, over 3296187.41 frames. ], batch size: 47, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:38:19,511 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7425, 3.9174, 2.7166, 4.4701, 3.0905, 4.4858, 2.6633, 3.3322], device='cuda:5'), covar=tensor([0.0334, 0.0409, 0.1374, 0.0317, 0.0801, 0.0484, 0.1460, 0.0645], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0180, 0.0197, 0.0163, 0.0179, 0.0218, 0.0206, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:38:39,286 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193867.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:38:45,941 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.059e+02 2.451e+02 3.046e+02 5.383e+02, threshold=4.901e+02, percent-clipped=2.0 2023-05-01 00:39:00,650 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193884.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:39:18,273 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7025, 4.6593, 4.6385, 4.0771, 4.6507, 2.0919, 4.4152, 4.3562], device='cuda:5'), covar=tensor([0.0119, 0.0096, 0.0173, 0.0334, 0.0110, 0.2334, 0.0156, 0.0204], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0149, 0.0193, 0.0172, 0.0169, 0.0205, 0.0182, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:39:24,781 INFO [train.py:904] (5/8) Epoch 20, batch 1050, loss[loss=0.1682, simple_loss=0.263, pruned_loss=0.03665, over 17111.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2489, pruned_loss=0.03986, over 3292053.65 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:39:43,719 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:03,543 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-05-01 00:40:18,148 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:25,150 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:35,471 INFO [train.py:904] (5/8) Epoch 20, batch 1100, loss[loss=0.1831, simple_loss=0.2516, pruned_loss=0.05727, over 16881.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2484, pruned_loss=0.03969, over 3287424.35 frames. ], batch size: 109, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:41:01,763 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193971.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:04,222 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.207e+02 2.619e+02 3.357e+02 2.000e+03, threshold=5.237e+02, percent-clipped=3.0 2023-05-01 00:41:09,289 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8498, 4.7986, 4.6980, 4.2048, 4.7775, 2.0043, 4.5363, 4.4819], device='cuda:5'), covar=tensor([0.0110, 0.0103, 0.0194, 0.0350, 0.0108, 0.2582, 0.0140, 0.0218], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0149, 0.0194, 0.0173, 0.0170, 0.0205, 0.0183, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:41:25,202 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:27,851 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 00:41:38,338 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9925, 5.0596, 5.4925, 5.5050, 5.4975, 5.1535, 5.0631, 4.9095], device='cuda:5'), covar=tensor([0.0345, 0.0598, 0.0386, 0.0371, 0.0480, 0.0418, 0.0985, 0.0473], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0444, 0.0429, 0.0401, 0.0480, 0.0451, 0.0544, 0.0365], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 00:41:46,957 INFO [train.py:904] (5/8) Epoch 20, batch 1150, loss[loss=0.1494, simple_loss=0.2442, pruned_loss=0.02726, over 17185.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2485, pruned_loss=0.03916, over 3301466.08 frames. ], batch size: 46, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:42:10,980 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2339, 2.2201, 2.3114, 4.0241, 2.2238, 2.5761, 2.2688, 2.3976], device='cuda:5'), covar=tensor([0.1388, 0.3694, 0.3006, 0.0570, 0.3996, 0.2509, 0.3844, 0.3230], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0438, 0.0363, 0.0325, 0.0434, 0.0505, 0.0409, 0.0515], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:42:35,809 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-05-01 00:42:56,047 INFO [train.py:904] (5/8) Epoch 20, batch 1200, loss[loss=0.137, simple_loss=0.2218, pruned_loss=0.02615, over 15815.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.249, pruned_loss=0.03924, over 3307441.98 frames. ], batch size: 35, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:43:25,127 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194072.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:43:25,794 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.319e+02 2.146e+02 2.588e+02 3.131e+02 5.620e+02, threshold=5.177e+02, percent-clipped=2.0 2023-05-01 00:43:33,106 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194077.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:44:06,932 INFO [train.py:904] (5/8) Epoch 20, batch 1250, loss[loss=0.1677, simple_loss=0.2517, pruned_loss=0.04191, over 15736.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2485, pruned_loss=0.03962, over 3301992.96 frames. ], batch size: 190, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:44:17,632 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194110.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:44:38,184 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:44:49,835 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:45:15,928 INFO [train.py:904] (5/8) Epoch 20, batch 1300, loss[loss=0.2042, simple_loss=0.2767, pruned_loss=0.06584, over 16911.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2483, pruned_loss=0.03967, over 3304366.85 frames. ], batch size: 109, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:45:26,571 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194158.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:45:46,649 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.297e+02 2.207e+02 2.541e+02 3.000e+02 4.873e+02, threshold=5.083e+02, percent-clipped=0.0 2023-05-01 00:46:04,743 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0603, 3.1358, 3.3527, 2.0310, 2.8830, 2.1745, 3.6079, 3.4968], device='cuda:5'), covar=tensor([0.0208, 0.0909, 0.0628, 0.1988, 0.0842, 0.1102, 0.0490, 0.0862], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0161, 0.0166, 0.0152, 0.0144, 0.0128, 0.0144, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:46:27,323 INFO [train.py:904] (5/8) Epoch 20, batch 1350, loss[loss=0.1713, simple_loss=0.2479, pruned_loss=0.04734, over 16888.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2489, pruned_loss=0.03969, over 3309888.64 frames. ], batch size: 90, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:47:21,265 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194240.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:47:36,947 INFO [train.py:904] (5/8) Epoch 20, batch 1400, loss[loss=0.1691, simple_loss=0.2636, pruned_loss=0.0373, over 16771.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2494, pruned_loss=0.03992, over 3312595.44 frames. ], batch size: 57, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:48:03,487 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194271.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:48:06,431 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.050e+02 2.370e+02 3.098e+02 5.438e+02, threshold=4.739e+02, percent-clipped=2.0 2023-05-01 00:48:06,898 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8124, 4.9052, 5.2989, 5.3311, 5.3001, 4.9426, 4.9086, 4.7609], device='cuda:5'), covar=tensor([0.0367, 0.0572, 0.0499, 0.0414, 0.0516, 0.0420, 0.0969, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0445, 0.0429, 0.0403, 0.0481, 0.0453, 0.0545, 0.0364], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 00:48:44,302 INFO [train.py:904] (5/8) Epoch 20, batch 1450, loss[loss=0.1532, simple_loss=0.243, pruned_loss=0.03165, over 17209.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2489, pruned_loss=0.03988, over 3323016.82 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:06,330 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194318.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:08,665 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:20,646 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0260, 4.1217, 2.7500, 4.8079, 3.3002, 4.7281, 2.8612, 3.5130], device='cuda:5'), covar=tensor([0.0290, 0.0383, 0.1584, 0.0303, 0.0789, 0.0468, 0.1444, 0.0644], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0181, 0.0198, 0.0164, 0.0179, 0.0220, 0.0206, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:49:54,002 INFO [train.py:904] (5/8) Epoch 20, batch 1500, loss[loss=0.1475, simple_loss=0.2384, pruned_loss=0.02829, over 17183.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2494, pruned_loss=0.04065, over 3322716.19 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:55,484 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:50:07,949 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0723, 5.0517, 4.8220, 4.3306, 4.9197, 1.7882, 4.6593, 4.7077], device='cuda:5'), covar=tensor([0.0092, 0.0085, 0.0207, 0.0437, 0.0117, 0.2862, 0.0152, 0.0240], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0150, 0.0195, 0.0175, 0.0172, 0.0206, 0.0185, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:50:07,992 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194362.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:50:24,393 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.182e+02 2.566e+02 3.379e+02 8.536e+02, threshold=5.133e+02, percent-clipped=4.0 2023-05-01 00:50:31,263 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:51:03,665 INFO [train.py:904] (5/8) Epoch 20, batch 1550, loss[loss=0.1608, simple_loss=0.2493, pruned_loss=0.03611, over 17234.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2506, pruned_loss=0.04111, over 3315212.55 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:51:20,836 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:51:34,194 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194423.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:51:40,655 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:52:03,733 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9411, 3.0958, 3.1142, 2.1498, 2.6968, 2.1872, 3.4199, 3.4287], device='cuda:5'), covar=tensor([0.0271, 0.0932, 0.0677, 0.1891, 0.0983, 0.1100, 0.0615, 0.0874], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0161, 0.0166, 0.0152, 0.0144, 0.0128, 0.0145, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:52:13,200 INFO [train.py:904] (5/8) Epoch 20, batch 1600, loss[loss=0.1481, simple_loss=0.2336, pruned_loss=0.03133, over 17015.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2521, pruned_loss=0.0419, over 3323590.52 frames. ], batch size: 41, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:52:36,619 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3440, 3.3452, 2.1863, 3.5842, 2.6029, 3.5396, 2.1620, 2.7260], device='cuda:5'), covar=tensor([0.0293, 0.0443, 0.1512, 0.0282, 0.0794, 0.0726, 0.1447, 0.0733], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0181, 0.0198, 0.0165, 0.0180, 0.0221, 0.0206, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:52:43,858 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.299e+02 2.983e+02 3.595e+02 1.383e+03, threshold=5.966e+02, percent-clipped=6.0 2023-05-01 00:53:12,224 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 00:53:22,058 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4388, 3.6965, 3.9851, 2.2385, 3.0932, 2.5517, 3.8073, 3.8265], device='cuda:5'), covar=tensor([0.0288, 0.0869, 0.0459, 0.1862, 0.0833, 0.0910, 0.0630, 0.1035], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0160, 0.0166, 0.0151, 0.0144, 0.0128, 0.0144, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:53:22,708 INFO [train.py:904] (5/8) Epoch 20, batch 1650, loss[loss=0.1528, simple_loss=0.2512, pruned_loss=0.02726, over 17110.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.253, pruned_loss=0.04167, over 3312432.84 frames. ], batch size: 49, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:54:16,630 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194540.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:54:18,913 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8798, 3.8555, 4.3557, 1.9821, 4.5864, 4.6796, 3.2573, 3.5786], device='cuda:5'), covar=tensor([0.0745, 0.0279, 0.0220, 0.1299, 0.0084, 0.0164, 0.0441, 0.0386], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0141, 0.0080, 0.0125, 0.0128, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:54:32,849 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5074, 3.7224, 4.0769, 2.1556, 3.2381, 2.6260, 3.9125, 3.8406], device='cuda:5'), covar=tensor([0.0277, 0.0931, 0.0511, 0.2078, 0.0807, 0.0969, 0.0672, 0.1132], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0160, 0.0166, 0.0151, 0.0143, 0.0128, 0.0144, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:54:33,468 INFO [train.py:904] (5/8) Epoch 20, batch 1700, loss[loss=0.1403, simple_loss=0.2348, pruned_loss=0.02292, over 17265.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2547, pruned_loss=0.04216, over 3317316.08 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:54:45,449 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3067, 2.1883, 1.7574, 1.9149, 2.4520, 2.1729, 2.2897, 2.5622], device='cuda:5'), covar=tensor([0.0261, 0.0366, 0.0521, 0.0469, 0.0241, 0.0346, 0.0211, 0.0259], device='cuda:5'), in_proj_covar=tensor([0.0208, 0.0239, 0.0228, 0.0231, 0.0241, 0.0239, 0.0239, 0.0232], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:54:48,087 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-05-01 00:55:05,702 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.259e+02 2.640e+02 3.354e+02 1.280e+03, threshold=5.281e+02, percent-clipped=2.0 2023-05-01 00:55:25,665 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194588.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:55:43,640 INFO [train.py:904] (5/8) Epoch 20, batch 1750, loss[loss=0.1701, simple_loss=0.2592, pruned_loss=0.04045, over 16492.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2555, pruned_loss=0.04181, over 3327466.17 frames. ], batch size: 68, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:55:44,020 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8458, 4.6996, 4.8782, 5.1161, 5.2601, 4.6225, 5.2448, 5.2823], device='cuda:5'), covar=tensor([0.2210, 0.1598, 0.2160, 0.0968, 0.0799, 0.0993, 0.0768, 0.0735], device='cuda:5'), in_proj_covar=tensor([0.0648, 0.0803, 0.0940, 0.0823, 0.0612, 0.0642, 0.0663, 0.0766], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:56:27,125 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194634.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:56:51,757 INFO [train.py:904] (5/8) Epoch 20, batch 1800, loss[loss=0.194, simple_loss=0.287, pruned_loss=0.05052, over 17051.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2578, pruned_loss=0.0428, over 3322906.37 frames. ], batch size: 53, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:57:06,047 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4893, 3.5137, 1.9005, 3.7031, 2.7214, 3.6864, 2.0770, 2.7364], device='cuda:5'), covar=tensor([0.0252, 0.0392, 0.1831, 0.0328, 0.0806, 0.0692, 0.1634, 0.0744], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0180, 0.0198, 0.0165, 0.0179, 0.0220, 0.0205, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 00:57:10,216 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2348, 4.0035, 4.0048, 4.3842, 4.5388, 4.1774, 4.3244, 4.5148], device='cuda:5'), covar=tensor([0.1629, 0.1404, 0.2112, 0.1024, 0.0894, 0.1512, 0.2142, 0.1215], device='cuda:5'), in_proj_covar=tensor([0.0644, 0.0797, 0.0933, 0.0818, 0.0610, 0.0639, 0.0659, 0.0762], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:57:22,536 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:57:23,425 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.398e+02 2.901e+02 3.350e+02 9.637e+02, threshold=5.801e+02, percent-clipped=10.0 2023-05-01 00:57:37,268 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6306, 3.9395, 4.0533, 2.7879, 3.5719, 4.0919, 3.6577, 2.4523], device='cuda:5'), covar=tensor([0.0480, 0.0193, 0.0054, 0.0371, 0.0130, 0.0096, 0.0099, 0.0426], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0082, 0.0082, 0.0135, 0.0097, 0.0108, 0.0094, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:5') 2023-05-01 00:57:50,535 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194695.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:57:58,030 INFO [train.py:904] (5/8) Epoch 20, batch 1850, loss[loss=0.1572, simple_loss=0.255, pruned_loss=0.02966, over 16728.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2587, pruned_loss=0.04319, over 3329145.86 frames. ], batch size: 57, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:58:08,701 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:58:21,446 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194718.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:58:34,481 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:58:56,397 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3446, 5.7302, 5.4867, 5.5519, 5.1066, 5.1778, 5.1319, 5.8761], device='cuda:5'), covar=tensor([0.1344, 0.0983, 0.1065, 0.0855, 0.0929, 0.0852, 0.1168, 0.0943], device='cuda:5'), in_proj_covar=tensor([0.0672, 0.0828, 0.0675, 0.0616, 0.0518, 0.0525, 0.0693, 0.0635], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 00:59:06,828 INFO [train.py:904] (5/8) Epoch 20, batch 1900, loss[loss=0.1569, simple_loss=0.2448, pruned_loss=0.03452, over 17198.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2577, pruned_loss=0.0425, over 3330599.95 frames. ], batch size: 44, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:59:38,383 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.030e+02 2.433e+02 3.010e+02 6.888e+02, threshold=4.866e+02, percent-clipped=1.0 2023-05-01 00:59:41,139 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:00:16,323 INFO [train.py:904] (5/8) Epoch 20, batch 1950, loss[loss=0.1833, simple_loss=0.2656, pruned_loss=0.05049, over 16695.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2571, pruned_loss=0.04199, over 3327259.43 frames. ], batch size: 134, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 01:00:20,116 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 01:00:44,436 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 01:00:49,241 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8228, 4.7053, 4.6978, 4.3312, 4.3561, 4.7568, 4.6515, 4.4457], device='cuda:5'), covar=tensor([0.0695, 0.0763, 0.0338, 0.0357, 0.0953, 0.0565, 0.0438, 0.0739], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0428, 0.0346, 0.0340, 0.0357, 0.0399, 0.0239, 0.0414], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:01:17,143 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:01:23,592 INFO [train.py:904] (5/8) Epoch 20, batch 2000, loss[loss=0.1809, simple_loss=0.2507, pruned_loss=0.05552, over 16880.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2572, pruned_loss=0.04204, over 3323368.04 frames. ], batch size: 116, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:01:54,987 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.120e+02 2.565e+02 3.052e+02 4.628e+02, threshold=5.130e+02, percent-clipped=0.0 2023-05-01 01:02:32,415 INFO [train.py:904] (5/8) Epoch 20, batch 2050, loss[loss=0.1696, simple_loss=0.25, pruned_loss=0.04455, over 16844.00 frames. ], tot_loss[loss=0.171, simple_loss=0.257, pruned_loss=0.0425, over 3322888.25 frames. ], batch size: 116, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:02:41,474 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194908.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:02:55,021 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6344, 3.7264, 2.8555, 2.1601, 2.4668, 2.3638, 3.8335, 3.2960], device='cuda:5'), covar=tensor([0.2745, 0.0646, 0.1706, 0.2889, 0.2682, 0.2025, 0.0551, 0.1397], device='cuda:5'), in_proj_covar=tensor([0.0323, 0.0267, 0.0300, 0.0304, 0.0293, 0.0253, 0.0289, 0.0332], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:03:15,417 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 01:03:41,681 INFO [train.py:904] (5/8) Epoch 20, batch 2100, loss[loss=0.1857, simple_loss=0.2602, pruned_loss=0.05554, over 16453.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2573, pruned_loss=0.04267, over 3314609.65 frames. ], batch size: 146, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:04:12,814 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194974.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:04:13,685 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.067e+02 2.444e+02 3.012e+02 5.275e+02, threshold=4.887e+02, percent-clipped=1.0 2023-05-01 01:04:34,705 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:04:50,815 INFO [train.py:904] (5/8) Epoch 20, batch 2150, loss[loss=0.155, simple_loss=0.242, pruned_loss=0.03398, over 17214.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2584, pruned_loss=0.04348, over 3321324.73 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:05:01,408 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:04,411 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:13,430 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195018.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:05:18,550 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:48,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5833, 3.6470, 2.8374, 2.1386, 2.3804, 2.3640, 3.7780, 3.1793], device='cuda:5'), covar=tensor([0.2821, 0.0640, 0.1684, 0.3177, 0.2920, 0.2061, 0.0533, 0.1596], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0268, 0.0300, 0.0304, 0.0294, 0.0253, 0.0290, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:05:49,724 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195044.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:06:01,909 INFO [train.py:904] (5/8) Epoch 20, batch 2200, loss[loss=0.166, simple_loss=0.2532, pruned_loss=0.03942, over 17060.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2589, pruned_loss=0.04366, over 3317407.73 frames. ], batch size: 55, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:06:09,195 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:21,179 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195066.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:06:28,907 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195071.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:30,164 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195072.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:06:33,764 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.410e+02 2.761e+02 3.300e+02 5.006e+02, threshold=5.521e+02, percent-clipped=1.0 2023-05-01 01:07:10,896 INFO [train.py:904] (5/8) Epoch 20, batch 2250, loss[loss=0.2081, simple_loss=0.2882, pruned_loss=0.06405, over 16202.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2597, pruned_loss=0.04394, over 3306884.45 frames. ], batch size: 164, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:07:15,469 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195105.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 01:07:54,771 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:08:21,153 INFO [train.py:904] (5/8) Epoch 20, batch 2300, loss[loss=0.1769, simple_loss=0.2636, pruned_loss=0.04506, over 16854.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2602, pruned_loss=0.04428, over 3304999.28 frames. ], batch size: 102, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:08:23,637 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-05-01 01:08:51,385 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.280e+02 2.663e+02 3.175e+02 5.300e+02, threshold=5.327e+02, percent-clipped=0.0 2023-05-01 01:09:29,613 INFO [train.py:904] (5/8) Epoch 20, batch 2350, loss[loss=0.1622, simple_loss=0.2508, pruned_loss=0.03683, over 17107.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2603, pruned_loss=0.04447, over 3316204.00 frames. ], batch size: 48, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:09:31,709 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:09:34,568 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 01:09:39,583 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8603, 2.8077, 2.4605, 2.7488, 3.0676, 2.8813, 3.5057, 3.3357], device='cuda:5'), covar=tensor([0.0157, 0.0416, 0.0500, 0.0412, 0.0296, 0.0395, 0.0233, 0.0290], device='cuda:5'), in_proj_covar=tensor([0.0209, 0.0240, 0.0227, 0.0230, 0.0240, 0.0240, 0.0241, 0.0233], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:10:09,038 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 01:10:36,699 INFO [train.py:904] (5/8) Epoch 20, batch 2400, loss[loss=0.1612, simple_loss=0.2421, pruned_loss=0.04015, over 16807.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2603, pruned_loss=0.04425, over 3319301.09 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:11:07,015 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.253e+02 2.639e+02 3.077e+02 5.540e+02, threshold=5.278e+02, percent-clipped=1.0 2023-05-01 01:11:29,204 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:11:35,022 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 01:11:45,807 INFO [train.py:904] (5/8) Epoch 20, batch 2450, loss[loss=0.1546, simple_loss=0.2477, pruned_loss=0.03074, over 17245.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2605, pruned_loss=0.04364, over 3316281.04 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:12:20,907 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 01:12:35,687 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:12:40,708 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 01:12:54,082 INFO [train.py:904] (5/8) Epoch 20, batch 2500, loss[loss=0.2072, simple_loss=0.2972, pruned_loss=0.05863, over 16692.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2599, pruned_loss=0.04308, over 3325674.55 frames. ], batch size: 62, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:13:16,675 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195367.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 01:13:26,927 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.185e+02 2.562e+02 3.023e+02 6.708e+02, threshold=5.124e+02, percent-clipped=4.0 2023-05-01 01:13:50,909 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:14:01,552 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195400.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 01:14:04,096 INFO [train.py:904] (5/8) Epoch 20, batch 2550, loss[loss=0.166, simple_loss=0.2594, pruned_loss=0.03633, over 16491.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.26, pruned_loss=0.04349, over 3306081.16 frames. ], batch size: 75, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:14:38,111 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:15:05,087 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 01:15:07,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1474, 2.1825, 2.3012, 3.9390, 2.1563, 2.4646, 2.2350, 2.3244], device='cuda:5'), covar=tensor([0.1489, 0.3696, 0.2884, 0.0602, 0.3927, 0.2564, 0.3764, 0.3169], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0442, 0.0365, 0.0327, 0.0435, 0.0510, 0.0412, 0.0519], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:15:11,783 INFO [train.py:904] (5/8) Epoch 20, batch 2600, loss[loss=0.1618, simple_loss=0.2512, pruned_loss=0.03623, over 16433.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2598, pruned_loss=0.04272, over 3304231.09 frames. ], batch size: 75, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:15:13,384 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:15:39,638 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 01:15:42,993 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.224e+02 2.625e+02 3.381e+02 7.141e+02, threshold=5.251e+02, percent-clipped=4.0 2023-05-01 01:15:55,389 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8136, 2.5892, 2.4362, 3.8901, 3.1607, 3.9044, 1.5230, 2.8388], device='cuda:5'), covar=tensor([0.1369, 0.0703, 0.1194, 0.0168, 0.0137, 0.0369, 0.1635, 0.0841], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0173, 0.0191, 0.0186, 0.0203, 0.0214, 0.0197, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 01:16:20,747 INFO [train.py:904] (5/8) Epoch 20, batch 2650, loss[loss=0.182, simple_loss=0.2632, pruned_loss=0.05036, over 15739.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2605, pruned_loss=0.04245, over 3313752.28 frames. ], batch size: 192, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:16:22,304 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:17:28,738 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:17:29,630 INFO [train.py:904] (5/8) Epoch 20, batch 2700, loss[loss=0.1809, simple_loss=0.2603, pruned_loss=0.05079, over 16835.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2608, pruned_loss=0.04248, over 3320459.60 frames. ], batch size: 96, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:18:00,662 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.029e+02 2.571e+02 3.039e+02 8.808e+02, threshold=5.142e+02, percent-clipped=4.0 2023-05-01 01:18:34,037 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0414, 5.0510, 5.5577, 5.5228, 5.5438, 5.1924, 5.1196, 4.9201], device='cuda:5'), covar=tensor([0.0336, 0.0542, 0.0371, 0.0448, 0.0492, 0.0397, 0.0953, 0.0473], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0454, 0.0437, 0.0410, 0.0485, 0.0462, 0.0554, 0.0371], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 01:18:34,657 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 01:18:39,451 INFO [train.py:904] (5/8) Epoch 20, batch 2750, loss[loss=0.1429, simple_loss=0.2373, pruned_loss=0.02423, over 16791.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2606, pruned_loss=0.04211, over 3316266.74 frames. ], batch size: 39, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:18:51,117 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 01:19:20,493 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3034, 2.3363, 2.2287, 4.1558, 2.2622, 2.6555, 2.3079, 2.4412], device='cuda:5'), covar=tensor([0.1324, 0.3455, 0.3007, 0.0539, 0.3885, 0.2490, 0.3550, 0.3364], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0441, 0.0365, 0.0326, 0.0434, 0.0509, 0.0412, 0.0518], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:19:47,497 INFO [train.py:904] (5/8) Epoch 20, batch 2800, loss[loss=0.1836, simple_loss=0.2625, pruned_loss=0.05235, over 16791.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2608, pruned_loss=0.04207, over 3324697.26 frames. ], batch size: 124, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:20:07,234 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195667.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:20:18,616 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 2.150e+02 2.557e+02 3.074e+02 6.713e+02, threshold=5.114e+02, percent-clipped=2.0 2023-05-01 01:20:20,243 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1459, 5.7172, 5.9293, 5.5467, 5.7212, 6.2713, 5.6535, 5.3804], device='cuda:5'), covar=tensor([0.1040, 0.1934, 0.2096, 0.2302, 0.2936, 0.1007, 0.1600, 0.2772], device='cuda:5'), in_proj_covar=tensor([0.0418, 0.0608, 0.0665, 0.0506, 0.0673, 0.0698, 0.0516, 0.0675], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:20:28,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8762, 4.3880, 3.0844, 2.3927, 2.8397, 2.6526, 4.7704, 3.8147], device='cuda:5'), covar=tensor([0.2905, 0.0554, 0.1758, 0.2927, 0.2753, 0.1975, 0.0331, 0.1287], device='cuda:5'), in_proj_covar=tensor([0.0323, 0.0269, 0.0302, 0.0306, 0.0295, 0.0255, 0.0291, 0.0335], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:20:38,168 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0719, 5.5408, 5.7818, 5.3836, 5.5162, 6.1375, 5.5712, 5.3199], device='cuda:5'), covar=tensor([0.0962, 0.2005, 0.2377, 0.2137, 0.2689, 0.0962, 0.1497, 0.2339], device='cuda:5'), in_proj_covar=tensor([0.0417, 0.0608, 0.0665, 0.0506, 0.0673, 0.0698, 0.0516, 0.0674], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:20:52,388 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195700.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:20:54,322 INFO [train.py:904] (5/8) Epoch 20, batch 2850, loss[loss=0.1772, simple_loss=0.2829, pruned_loss=0.03575, over 17070.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2608, pruned_loss=0.04171, over 3325294.16 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:21:13,262 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195715.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:29,675 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:56,481 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195748.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:21:56,492 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195748.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:22:02,488 INFO [train.py:904] (5/8) Epoch 20, batch 2900, loss[loss=0.1698, simple_loss=0.2532, pruned_loss=0.04322, over 16117.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2596, pruned_loss=0.04258, over 3323822.06 frames. ], batch size: 35, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:22:33,093 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.294e+02 2.743e+02 3.539e+02 8.268e+02, threshold=5.486e+02, percent-clipped=5.0 2023-05-01 01:22:34,112 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:22:38,733 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 01:23:10,995 INFO [train.py:904] (5/8) Epoch 20, batch 2950, loss[loss=0.1616, simple_loss=0.2562, pruned_loss=0.03352, over 17109.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2588, pruned_loss=0.04298, over 3322466.71 frames. ], batch size: 49, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:23:27,722 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7581, 2.8673, 2.7622, 4.9539, 3.9800, 4.4286, 1.5971, 3.1424], device='cuda:5'), covar=tensor([0.1360, 0.0768, 0.1149, 0.0162, 0.0227, 0.0344, 0.1643, 0.0760], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0173, 0.0192, 0.0188, 0.0205, 0.0215, 0.0198, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 01:23:57,606 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8719, 5.2015, 4.9058, 4.9406, 4.7283, 4.7138, 4.5851, 5.2731], device='cuda:5'), covar=tensor([0.1225, 0.0858, 0.1165, 0.0920, 0.0859, 0.1020, 0.1289, 0.0855], device='cuda:5'), in_proj_covar=tensor([0.0664, 0.0822, 0.0668, 0.0613, 0.0515, 0.0516, 0.0686, 0.0629], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:24:18,942 INFO [train.py:904] (5/8) Epoch 20, batch 3000, loss[loss=0.1788, simple_loss=0.2692, pruned_loss=0.04418, over 16648.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2601, pruned_loss=0.04375, over 3314050.90 frames. ], batch size: 62, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,942 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 01:24:27,133 INFO [train.py:938] (5/8) Epoch 20, validation: loss=0.1354, simple_loss=0.2409, pruned_loss=0.01492, over 944034.00 frames. 2023-05-01 01:24:27,134 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 01:24:58,718 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.268e+02 2.640e+02 3.074e+02 4.713e+02, threshold=5.280e+02, percent-clipped=0.0 2023-05-01 01:25:38,247 INFO [train.py:904] (5/8) Epoch 20, batch 3050, loss[loss=0.1492, simple_loss=0.2411, pruned_loss=0.02864, over 16817.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2589, pruned_loss=0.0433, over 3321696.68 frames. ], batch size: 42, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:26:46,791 INFO [train.py:904] (5/8) Epoch 20, batch 3100, loss[loss=0.1663, simple_loss=0.2423, pruned_loss=0.04512, over 16706.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2584, pruned_loss=0.04301, over 3330800.57 frames. ], batch size: 83, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:27:16,970 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.170e+02 2.517e+02 3.010e+02 4.589e+02, threshold=5.034e+02, percent-clipped=0.0 2023-05-01 01:27:55,511 INFO [train.py:904] (5/8) Epoch 20, batch 3150, loss[loss=0.1814, simple_loss=0.2619, pruned_loss=0.05044, over 16869.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2579, pruned_loss=0.04246, over 3329314.38 frames. ], batch size: 116, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:28:02,768 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 01:28:15,958 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 01:28:18,521 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2404, 5.1756, 5.0283, 4.5249, 4.6439, 5.1265, 5.1005, 4.7064], device='cuda:5'), covar=tensor([0.0604, 0.0553, 0.0337, 0.0385, 0.1170, 0.0456, 0.0375, 0.0783], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0434, 0.0352, 0.0347, 0.0364, 0.0404, 0.0242, 0.0421], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:28:20,238 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 01:28:57,193 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:29:03,222 INFO [train.py:904] (5/8) Epoch 20, batch 3200, loss[loss=0.1798, simple_loss=0.2699, pruned_loss=0.04488, over 17058.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2569, pruned_loss=0.0417, over 3328926.48 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:29:35,522 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.205e+02 2.543e+02 3.042e+02 4.560e+02, threshold=5.087e+02, percent-clipped=0.0 2023-05-01 01:30:04,478 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=196096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:30:11,579 INFO [train.py:904] (5/8) Epoch 20, batch 3250, loss[loss=0.2042, simple_loss=0.2976, pruned_loss=0.05543, over 17029.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2576, pruned_loss=0.04233, over 3316158.93 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:31:07,786 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7062, 3.6547, 3.9289, 3.6332, 3.8058, 4.2667, 3.8118, 3.5031], device='cuda:5'), covar=tensor([0.2686, 0.2535, 0.2259, 0.2770, 0.2746, 0.2052, 0.1767, 0.2755], device='cuda:5'), in_proj_covar=tensor([0.0420, 0.0611, 0.0670, 0.0510, 0.0677, 0.0705, 0.0518, 0.0679], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:31:19,994 INFO [train.py:904] (5/8) Epoch 20, batch 3300, loss[loss=0.1512, simple_loss=0.2413, pruned_loss=0.03057, over 17205.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2591, pruned_loss=0.04349, over 3310959.48 frames. ], batch size: 44, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:31:29,375 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4790, 5.8924, 5.6335, 5.6952, 5.2792, 5.2681, 5.2903, 6.0099], device='cuda:5'), covar=tensor([0.1293, 0.0939, 0.0968, 0.0852, 0.0880, 0.0791, 0.1289, 0.0924], device='cuda:5'), in_proj_covar=tensor([0.0675, 0.0834, 0.0680, 0.0622, 0.0523, 0.0526, 0.0696, 0.0637], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:31:52,355 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.123e+02 2.550e+02 3.106e+02 4.546e+02, threshold=5.100e+02, percent-clipped=0.0 2023-05-01 01:32:28,250 INFO [train.py:904] (5/8) Epoch 20, batch 3350, loss[loss=0.1619, simple_loss=0.2555, pruned_loss=0.03416, over 17012.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2578, pruned_loss=0.04271, over 3315295.05 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:33:04,022 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 01:33:24,696 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1822, 5.7010, 5.8260, 5.4834, 5.6334, 6.2122, 5.7246, 5.4805], device='cuda:5'), covar=tensor([0.0829, 0.2000, 0.2279, 0.2286, 0.2482, 0.0976, 0.1557, 0.2346], device='cuda:5'), in_proj_covar=tensor([0.0419, 0.0607, 0.0667, 0.0508, 0.0673, 0.0703, 0.0518, 0.0678], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:33:33,472 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0856, 2.1646, 2.3113, 3.8327, 2.1705, 2.5198, 2.2483, 2.3825], device='cuda:5'), covar=tensor([0.1454, 0.3722, 0.2843, 0.0598, 0.3850, 0.2436, 0.3833, 0.2968], device='cuda:5'), in_proj_covar=tensor([0.0401, 0.0442, 0.0366, 0.0329, 0.0437, 0.0511, 0.0413, 0.0519], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:33:35,781 INFO [train.py:904] (5/8) Epoch 20, batch 3400, loss[loss=0.1793, simple_loss=0.2693, pruned_loss=0.0446, over 17082.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.258, pruned_loss=0.04286, over 3326689.18 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:34:06,840 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.504e+02 2.175e+02 2.675e+02 3.072e+02 5.322e+02, threshold=5.351e+02, percent-clipped=1.0 2023-05-01 01:34:26,079 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5443, 3.5855, 3.9470, 2.0879, 3.1752, 2.5738, 3.9591, 3.7869], device='cuda:5'), covar=tensor([0.0260, 0.0957, 0.0505, 0.2118, 0.0805, 0.0957, 0.0621, 0.1192], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 01:34:44,352 INFO [train.py:904] (5/8) Epoch 20, batch 3450, loss[loss=0.1587, simple_loss=0.2584, pruned_loss=0.02949, over 17052.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2567, pruned_loss=0.04219, over 3319915.66 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:35:10,239 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:35:50,065 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-05-01 01:35:50,467 INFO [train.py:904] (5/8) Epoch 20, batch 3500, loss[loss=0.1701, simple_loss=0.2502, pruned_loss=0.04494, over 16849.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2556, pruned_loss=0.04169, over 3316877.56 frames. ], batch size: 102, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:36:03,313 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196360.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:36:23,664 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 2.052e+02 2.465e+02 2.837e+02 5.038e+02, threshold=4.930e+02, percent-clipped=0.0 2023-05-01 01:36:35,808 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:36:52,564 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6428, 4.3679, 4.1649, 4.7237, 4.8409, 4.4643, 4.6907, 4.8479], device='cuda:5'), covar=tensor([0.1772, 0.1617, 0.3257, 0.1447, 0.1361, 0.1509, 0.1851, 0.1660], device='cuda:5'), in_proj_covar=tensor([0.0665, 0.0824, 0.0964, 0.0844, 0.0628, 0.0658, 0.0673, 0.0782], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:37:01,674 INFO [train.py:904] (5/8) Epoch 20, batch 3550, loss[loss=0.1809, simple_loss=0.2835, pruned_loss=0.03919, over 16706.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2545, pruned_loss=0.04115, over 3325728.94 frames. ], batch size: 57, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:37:28,427 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:37:37,965 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 01:38:06,987 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1327, 5.5523, 5.7256, 5.4707, 5.5544, 6.1276, 5.6022, 5.3060], device='cuda:5'), covar=tensor([0.0891, 0.1882, 0.2068, 0.1969, 0.2472, 0.0899, 0.1518, 0.2482], device='cuda:5'), in_proj_covar=tensor([0.0417, 0.0606, 0.0665, 0.0507, 0.0671, 0.0701, 0.0516, 0.0676], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:38:10,222 INFO [train.py:904] (5/8) Epoch 20, batch 3600, loss[loss=0.1477, simple_loss=0.2378, pruned_loss=0.02878, over 17223.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2537, pruned_loss=0.04099, over 3319778.89 frames. ], batch size: 45, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:38:41,983 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.224e+02 2.585e+02 3.008e+02 4.978e+02, threshold=5.169e+02, percent-clipped=1.0 2023-05-01 01:39:20,694 INFO [train.py:904] (5/8) Epoch 20, batch 3650, loss[loss=0.1595, simple_loss=0.2464, pruned_loss=0.03627, over 17235.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2523, pruned_loss=0.04102, over 3312923.21 frames. ], batch size: 43, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:39:40,444 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 01:40:32,643 INFO [train.py:904] (5/8) Epoch 20, batch 3700, loss[loss=0.1605, simple_loss=0.2391, pruned_loss=0.04098, over 16406.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2507, pruned_loss=0.04205, over 3300906.31 frames. ], batch size: 146, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:41:07,139 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.139e+02 2.467e+02 2.956e+02 4.680e+02, threshold=4.935e+02, percent-clipped=0.0 2023-05-01 01:41:11,832 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:41:47,095 INFO [train.py:904] (5/8) Epoch 20, batch 3750, loss[loss=0.1665, simple_loss=0.2425, pruned_loss=0.04526, over 16817.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2521, pruned_loss=0.04359, over 3287123.22 frames. ], batch size: 83, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:42:10,491 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-01 01:42:13,565 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4469, 4.3566, 4.3351, 4.1013, 4.1318, 4.4192, 4.1126, 4.2291], device='cuda:5'), covar=tensor([0.0656, 0.0725, 0.0302, 0.0253, 0.0696, 0.0481, 0.0694, 0.0521], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0438, 0.0353, 0.0349, 0.0365, 0.0405, 0.0243, 0.0422], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:42:39,030 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:42:58,323 INFO [train.py:904] (5/8) Epoch 20, batch 3800, loss[loss=0.1745, simple_loss=0.2515, pruned_loss=0.04876, over 16837.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2531, pruned_loss=0.04486, over 3285774.53 frames. ], batch size: 102, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:43:31,147 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.212e+02 2.504e+02 2.913e+02 5.553e+02, threshold=5.009e+02, percent-clipped=1.0 2023-05-01 01:43:32,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3282, 3.3063, 3.5028, 2.2332, 2.9993, 2.4115, 3.8053, 3.7770], device='cuda:5'), covar=tensor([0.0224, 0.0893, 0.0605, 0.1911, 0.0859, 0.0972, 0.0452, 0.0795], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 01:43:35,326 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:44:01,064 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 01:44:10,774 INFO [train.py:904] (5/8) Epoch 20, batch 3850, loss[loss=0.1812, simple_loss=0.27, pruned_loss=0.04622, over 16597.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2542, pruned_loss=0.04603, over 3275978.29 frames. ], batch size: 62, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:44:32,377 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:45:12,983 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8527, 4.9876, 5.1726, 4.9793, 5.0489, 5.6048, 5.0644, 4.7908], device='cuda:5'), covar=tensor([0.1225, 0.1943, 0.1874, 0.2052, 0.2443, 0.0910, 0.1557, 0.2312], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0602, 0.0661, 0.0504, 0.0666, 0.0695, 0.0513, 0.0670], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:45:24,178 INFO [train.py:904] (5/8) Epoch 20, batch 3900, loss[loss=0.1568, simple_loss=0.2324, pruned_loss=0.04057, over 16764.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2545, pruned_loss=0.04679, over 3272841.77 frames. ], batch size: 89, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:45:24,725 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5994, 5.6042, 5.4019, 4.8001, 5.5869, 2.2565, 5.3240, 5.0225], device='cuda:5'), covar=tensor([0.0045, 0.0038, 0.0154, 0.0300, 0.0053, 0.2539, 0.0086, 0.0200], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0151, 0.0198, 0.0178, 0.0175, 0.0206, 0.0187, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:45:39,608 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 01:45:40,536 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5478, 3.2732, 3.6667, 1.9185, 3.7644, 3.7994, 3.1367, 2.8656], device='cuda:5'), covar=tensor([0.0709, 0.0272, 0.0159, 0.1144, 0.0106, 0.0178, 0.0352, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0108, 0.0097, 0.0138, 0.0079, 0.0124, 0.0126, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 01:45:57,454 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 2.027e+02 2.552e+02 2.919e+02 4.484e+02, threshold=5.103e+02, percent-clipped=0.0 2023-05-01 01:46:02,103 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 01:46:04,902 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196779.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:46:26,354 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-05-01 01:46:36,475 INFO [train.py:904] (5/8) Epoch 20, batch 3950, loss[loss=0.1858, simple_loss=0.263, pruned_loss=0.05427, over 16488.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2537, pruned_loss=0.04725, over 3271701.19 frames. ], batch size: 146, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:47:16,339 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 01:47:19,388 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1034, 3.2712, 3.4538, 2.3357, 3.1722, 3.5506, 3.2319, 1.7996], device='cuda:5'), covar=tensor([0.0524, 0.0121, 0.0066, 0.0392, 0.0107, 0.0111, 0.0105, 0.0523], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0082, 0.0082, 0.0133, 0.0097, 0.0107, 0.0093, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:47:32,080 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196840.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:47:48,970 INFO [train.py:904] (5/8) Epoch 20, batch 4000, loss[loss=0.1656, simple_loss=0.2472, pruned_loss=0.04199, over 16867.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2532, pruned_loss=0.04747, over 3284443.48 frames. ], batch size: 42, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:47:52,457 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:47:58,245 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-05-01 01:48:21,948 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.289e+02 2.579e+02 3.040e+02 5.791e+02, threshold=5.158e+02, percent-clipped=1.0 2023-05-01 01:48:32,499 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196881.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:01,034 INFO [train.py:904] (5/8) Epoch 20, batch 4050, loss[loss=0.1721, simple_loss=0.263, pruned_loss=0.04058, over 16715.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2547, pruned_loss=0.04704, over 3289393.12 frames. ], batch size: 89, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:49:19,962 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:47,760 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:47,870 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3692, 4.3364, 4.2336, 3.5274, 4.2926, 1.8229, 4.0158, 3.6941], device='cuda:5'), covar=tensor([0.0093, 0.0077, 0.0177, 0.0291, 0.0081, 0.2830, 0.0127, 0.0256], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0153, 0.0199, 0.0180, 0.0176, 0.0208, 0.0189, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:49:57,999 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196942.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:50:12,379 INFO [train.py:904] (5/8) Epoch 20, batch 4100, loss[loss=0.1988, simple_loss=0.2875, pruned_loss=0.05503, over 16745.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2561, pruned_loss=0.04644, over 3285017.80 frames. ], batch size: 124, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:50:39,849 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 01:50:44,594 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 1.746e+02 2.102e+02 2.519e+02 4.672e+02, threshold=4.204e+02, percent-clipped=0.0 2023-05-01 01:50:48,515 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:50:55,484 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-01 01:51:16,086 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2501, 4.3159, 4.4576, 4.2105, 4.3439, 4.8336, 4.3694, 4.0468], device='cuda:5'), covar=tensor([0.1658, 0.2043, 0.1984, 0.2291, 0.2604, 0.1074, 0.1484, 0.2608], device='cuda:5'), in_proj_covar=tensor([0.0414, 0.0601, 0.0658, 0.0503, 0.0664, 0.0694, 0.0513, 0.0670], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:51:23,822 INFO [train.py:904] (5/8) Epoch 20, batch 4150, loss[loss=0.2111, simple_loss=0.2963, pruned_loss=0.06297, over 16703.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2625, pruned_loss=0.04894, over 3224977.29 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:51:29,343 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4262, 4.4866, 4.6226, 4.4294, 4.4711, 5.0277, 4.5734, 4.2789], device='cuda:5'), covar=tensor([0.1305, 0.1947, 0.1976, 0.2162, 0.2739, 0.0935, 0.1357, 0.2467], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0600, 0.0657, 0.0502, 0.0663, 0.0693, 0.0511, 0.0669], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:51:45,584 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197016.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:52:00,075 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:52:11,943 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6490, 3.6986, 2.3879, 4.3666, 2.7993, 4.3056, 2.4293, 3.0216], device='cuda:5'), covar=tensor([0.0277, 0.0379, 0.1556, 0.0159, 0.0812, 0.0448, 0.1471, 0.0724], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0177, 0.0193, 0.0162, 0.0176, 0.0217, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 01:52:23,998 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3875, 4.4834, 4.6333, 4.3151, 4.3541, 4.9985, 4.4932, 4.1809], device='cuda:5'), covar=tensor([0.1422, 0.2016, 0.1955, 0.2339, 0.2967, 0.1082, 0.1678, 0.2819], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0594, 0.0649, 0.0496, 0.0656, 0.0686, 0.0506, 0.0661], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 01:52:24,314 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 01:52:40,732 INFO [train.py:904] (5/8) Epoch 20, batch 4200, loss[loss=0.2088, simple_loss=0.3038, pruned_loss=0.05688, over 16477.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2698, pruned_loss=0.05083, over 3187174.63 frames. ], batch size: 75, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:52:55,649 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7886, 4.8687, 5.1538, 5.1430, 5.1951, 4.8325, 4.7888, 4.5330], device='cuda:5'), covar=tensor([0.0295, 0.0442, 0.0355, 0.0361, 0.0407, 0.0353, 0.1018, 0.0515], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0446, 0.0428, 0.0402, 0.0476, 0.0455, 0.0542, 0.0364], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 01:52:58,852 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:53:12,971 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.329e+02 2.798e+02 3.249e+02 9.837e+02, threshold=5.595e+02, percent-clipped=2.0 2023-05-01 01:53:51,291 INFO [train.py:904] (5/8) Epoch 20, batch 4250, loss[loss=0.1877, simple_loss=0.2806, pruned_loss=0.04741, over 16863.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2736, pruned_loss=0.05074, over 3175505.04 frames. ], batch size: 109, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:54:40,512 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:54:59,345 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3550, 5.3778, 5.2280, 4.8583, 4.8877, 5.3215, 5.1592, 4.9544], device='cuda:5'), covar=tensor([0.0550, 0.0338, 0.0249, 0.0265, 0.0850, 0.0297, 0.0232, 0.0578], device='cuda:5'), in_proj_covar=tensor([0.0295, 0.0423, 0.0342, 0.0338, 0.0353, 0.0391, 0.0235, 0.0409], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 01:55:04,300 INFO [train.py:904] (5/8) Epoch 20, batch 4300, loss[loss=0.1831, simple_loss=0.2694, pruned_loss=0.04841, over 16806.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.274, pruned_loss=0.04952, over 3173827.36 frames. ], batch size: 39, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:55:37,988 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.176e+02 2.515e+02 3.039e+02 5.119e+02, threshold=5.030e+02, percent-clipped=0.0 2023-05-01 01:56:18,925 INFO [train.py:904] (5/8) Epoch 20, batch 4350, loss[loss=0.2141, simple_loss=0.2982, pruned_loss=0.06498, over 16631.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2778, pruned_loss=0.05091, over 3178018.58 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:56:29,940 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:56:34,112 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6772, 3.4526, 3.9130, 1.9235, 4.2011, 4.2120, 2.9767, 3.1918], device='cuda:5'), covar=tensor([0.0774, 0.0305, 0.0240, 0.1192, 0.0064, 0.0098, 0.0464, 0.0414], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0139, 0.0080, 0.0124, 0.0127, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 01:57:05,373 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197234.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:09,115 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:31,881 INFO [train.py:904] (5/8) Epoch 20, batch 4400, loss[loss=0.205, simple_loss=0.2975, pruned_loss=0.05625, over 16622.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2792, pruned_loss=0.05146, over 3202911.80 frames. ], batch size: 75, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:58:04,312 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.097e+02 2.441e+02 2.762e+02 5.334e+02, threshold=4.883e+02, percent-clipped=1.0 2023-05-01 01:58:12,915 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3322, 3.5995, 3.8167, 2.1917, 3.1390, 2.3898, 3.6762, 3.7622], device='cuda:5'), covar=tensor([0.0213, 0.0739, 0.0460, 0.1868, 0.0748, 0.0865, 0.0564, 0.0899], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 01:58:14,501 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:58:42,810 INFO [train.py:904] (5/8) Epoch 20, batch 4450, loss[loss=0.2081, simple_loss=0.3029, pruned_loss=0.05666, over 16848.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.283, pruned_loss=0.05276, over 3220760.25 frames. ], batch size: 116, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:59:55,991 INFO [train.py:904] (5/8) Epoch 20, batch 4500, loss[loss=0.2108, simple_loss=0.2952, pruned_loss=0.06322, over 16731.00 frames. ], tot_loss[loss=0.196, simple_loss=0.284, pruned_loss=0.05399, over 3220234.34 frames. ], batch size: 134, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:00:30,614 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 1.873e+02 2.163e+02 2.505e+02 3.448e+02, threshold=4.325e+02, percent-clipped=0.0 2023-05-01 02:01:08,126 INFO [train.py:904] (5/8) Epoch 20, batch 4550, loss[loss=0.228, simple_loss=0.3125, pruned_loss=0.07181, over 16418.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2843, pruned_loss=0.0545, over 3227230.72 frames. ], batch size: 68, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:01:20,852 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5284, 2.6040, 2.5006, 3.9076, 2.9285, 3.8723, 1.5777, 2.8394], device='cuda:5'), covar=tensor([0.1624, 0.0847, 0.1279, 0.0169, 0.0291, 0.0374, 0.1861, 0.0886], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0173, 0.0192, 0.0186, 0.0206, 0.0214, 0.0199, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:01:27,167 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197415.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:01:56,132 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197435.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:02:08,586 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5085, 4.5941, 4.3747, 4.0413, 4.0446, 4.5302, 4.1612, 4.1890], device='cuda:5'), covar=tensor([0.0462, 0.0294, 0.0230, 0.0257, 0.0694, 0.0267, 0.0583, 0.0505], device='cuda:5'), in_proj_covar=tensor([0.0289, 0.0414, 0.0335, 0.0332, 0.0347, 0.0382, 0.0231, 0.0401], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:02:18,317 INFO [train.py:904] (5/8) Epoch 20, batch 4600, loss[loss=0.2027, simple_loss=0.2983, pruned_loss=0.05349, over 16678.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2855, pruned_loss=0.05491, over 3225435.29 frames. ], batch size: 134, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:02:20,987 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-05-01 02:02:42,516 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6208, 3.5684, 2.2214, 4.3949, 2.8068, 4.3133, 2.2749, 2.9312], device='cuda:5'), covar=tensor([0.0288, 0.0379, 0.1715, 0.0157, 0.0822, 0.0366, 0.1620, 0.0812], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0159, 0.0174, 0.0214, 0.0199, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:02:52,208 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 1.798e+02 2.178e+02 2.547e+02 5.479e+02, threshold=4.357e+02, percent-clipped=1.0 2023-05-01 02:02:52,770 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197476.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:03:03,000 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:30,131 INFO [train.py:904] (5/8) Epoch 20, batch 4650, loss[loss=0.1865, simple_loss=0.2749, pruned_loss=0.04911, over 17218.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2843, pruned_loss=0.05451, over 3236716.28 frames. ], batch size: 45, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:03:41,248 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:49,588 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197516.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:05,608 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9869, 1.9438, 2.1550, 3.5436, 1.9328, 2.1622, 2.0833, 2.0563], device='cuda:5'), covar=tensor([0.1600, 0.4218, 0.3086, 0.0701, 0.5300, 0.3173, 0.3699, 0.4333], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0441, 0.0362, 0.0325, 0.0434, 0.0510, 0.0411, 0.0517], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:04:20,867 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:32,475 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197545.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:42,156 INFO [train.py:904] (5/8) Epoch 20, batch 4700, loss[loss=0.1751, simple_loss=0.2648, pruned_loss=0.04276, over 16499.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2818, pruned_loss=0.05352, over 3222538.58 frames. ], batch size: 75, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:04:52,430 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:05:18,235 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 1.893e+02 2.227e+02 2.649e+02 4.562e+02, threshold=4.454e+02, percent-clipped=1.0 2023-05-01 02:05:19,879 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197577.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:05:31,367 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:05:55,405 INFO [train.py:904] (5/8) Epoch 20, batch 4750, loss[loss=0.174, simple_loss=0.2596, pruned_loss=0.04418, over 16745.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2775, pruned_loss=0.05145, over 3231162.25 frames. ], batch size: 124, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:06:01,870 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:06:06,421 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8917, 2.1001, 2.3590, 3.1245, 2.1703, 2.3045, 2.2743, 2.2174], device='cuda:5'), covar=tensor([0.1410, 0.3265, 0.2569, 0.0693, 0.3920, 0.2290, 0.3290, 0.3425], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0441, 0.0362, 0.0324, 0.0433, 0.0509, 0.0410, 0.0517], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:07:08,969 INFO [train.py:904] (5/8) Epoch 20, batch 4800, loss[loss=0.173, simple_loss=0.275, pruned_loss=0.03552, over 16867.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2732, pruned_loss=0.04908, over 3235213.21 frames. ], batch size: 96, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:07:27,711 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-01 02:07:45,096 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.191e+02 1.779e+02 2.011e+02 2.336e+02 4.336e+02, threshold=4.022e+02, percent-clipped=0.0 2023-05-01 02:08:24,130 INFO [train.py:904] (5/8) Epoch 20, batch 4850, loss[loss=0.1657, simple_loss=0.2533, pruned_loss=0.03904, over 16274.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2746, pruned_loss=0.04837, over 3227740.46 frames. ], batch size: 35, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:08:35,732 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.54 vs. limit=5.0 2023-05-01 02:09:40,079 INFO [train.py:904] (5/8) Epoch 20, batch 4900, loss[loss=0.1912, simple_loss=0.2833, pruned_loss=0.0496, over 16355.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2735, pruned_loss=0.04703, over 3207455.27 frames. ], batch size: 146, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:10:08,664 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197771.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:10:15,936 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.929e+02 2.192e+02 2.700e+02 5.551e+02, threshold=4.384e+02, percent-clipped=1.0 2023-05-01 02:10:25,600 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-01 02:10:52,826 INFO [train.py:904] (5/8) Epoch 20, batch 4950, loss[loss=0.2092, simple_loss=0.2954, pruned_loss=0.06152, over 16568.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2737, pruned_loss=0.04685, over 3184983.94 frames. ], batch size: 57, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:04,526 INFO [train.py:904] (5/8) Epoch 20, batch 5000, loss[loss=0.1673, simple_loss=0.2599, pruned_loss=0.03738, over 16591.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2752, pruned_loss=0.04696, over 3203227.43 frames. ], batch size: 62, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:33,418 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197872.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:12:38,860 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.376e+02 1.953e+02 2.351e+02 2.789e+02 5.042e+02, threshold=4.702e+02, percent-clipped=2.0 2023-05-01 02:13:01,239 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197892.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:13:13,695 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197901.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:13:14,329 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 02:13:14,491 INFO [train.py:904] (5/8) Epoch 20, batch 5050, loss[loss=0.1893, simple_loss=0.2815, pruned_loss=0.04854, over 15234.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2756, pruned_loss=0.04673, over 3207983.02 frames. ], batch size: 190, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:14:09,931 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7907, 5.1277, 5.3104, 5.0013, 5.1064, 5.7128, 5.1615, 4.8603], device='cuda:5'), covar=tensor([0.0940, 0.1700, 0.1464, 0.1838, 0.2167, 0.0779, 0.1329, 0.2082], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0571, 0.0624, 0.0475, 0.0635, 0.0664, 0.0491, 0.0640], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:14:25,056 INFO [train.py:904] (5/8) Epoch 20, batch 5100, loss[loss=0.1485, simple_loss=0.2421, pruned_loss=0.02745, over 16732.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2735, pruned_loss=0.046, over 3217661.95 frames. ], batch size: 83, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:14:27,875 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197953.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:15:00,612 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 1.957e+02 2.274e+02 2.606e+02 4.454e+02, threshold=4.549e+02, percent-clipped=0.0 2023-05-01 02:15:05,169 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 02:15:15,333 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5923, 3.3593, 3.9851, 1.9853, 4.0789, 4.1561, 3.0643, 2.8977], device='cuda:5'), covar=tensor([0.0746, 0.0294, 0.0132, 0.1165, 0.0064, 0.0102, 0.0355, 0.0491], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0108, 0.0096, 0.0138, 0.0079, 0.0123, 0.0127, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:15:41,439 INFO [train.py:904] (5/8) Epoch 20, batch 5150, loss[loss=0.2022, simple_loss=0.2961, pruned_loss=0.05417, over 16752.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2731, pruned_loss=0.0452, over 3211994.65 frames. ], batch size: 124, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:16:22,289 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2249, 3.9395, 3.9208, 2.6021, 3.5280, 3.9157, 3.4603, 2.1710], device='cuda:5'), covar=tensor([0.0552, 0.0046, 0.0043, 0.0369, 0.0091, 0.0100, 0.0094, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0131, 0.0095, 0.0106, 0.0092, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:16:51,056 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 02:16:52,363 INFO [train.py:904] (5/8) Epoch 20, batch 5200, loss[loss=0.1717, simple_loss=0.2519, pruned_loss=0.04573, over 17008.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2724, pruned_loss=0.04483, over 3205732.29 frames. ], batch size: 55, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:17:06,768 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 02:17:20,274 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198071.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:17:27,748 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 1.904e+02 2.236e+02 2.771e+02 4.329e+02, threshold=4.472e+02, percent-clipped=0.0 2023-05-01 02:17:35,506 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198081.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:18:03,847 INFO [train.py:904] (5/8) Epoch 20, batch 5250, loss[loss=0.1806, simple_loss=0.2621, pruned_loss=0.04959, over 12440.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2693, pruned_loss=0.04437, over 3206986.14 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:18:26,393 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 02:18:28,864 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198119.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:19:01,581 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198142.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:19:15,065 INFO [train.py:904] (5/8) Epoch 20, batch 5300, loss[loss=0.1519, simple_loss=0.2385, pruned_loss=0.03262, over 16605.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2657, pruned_loss=0.04325, over 3219913.71 frames. ], batch size: 75, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:19:44,949 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198172.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:19:49,804 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 1.825e+02 2.153e+02 2.534e+02 4.591e+02, threshold=4.307e+02, percent-clipped=1.0 2023-05-01 02:20:27,629 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198201.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:20:28,369 INFO [train.py:904] (5/8) Epoch 20, batch 5350, loss[loss=0.1721, simple_loss=0.2623, pruned_loss=0.04095, over 16742.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2647, pruned_loss=0.04272, over 3228973.45 frames. ], batch size: 39, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:20:28,945 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0058, 3.4320, 3.3809, 2.2101, 3.1118, 3.3877, 3.1348, 1.9231], device='cuda:5'), covar=tensor([0.0569, 0.0053, 0.0055, 0.0416, 0.0108, 0.0103, 0.0108, 0.0508], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0081, 0.0080, 0.0132, 0.0095, 0.0106, 0.0092, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:20:30,901 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0488, 3.2174, 3.4489, 2.0732, 2.9515, 2.2532, 3.5274, 3.5456], device='cuda:5'), covar=tensor([0.0261, 0.0805, 0.0599, 0.1961, 0.0816, 0.0985, 0.0603, 0.0850], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0162, 0.0166, 0.0151, 0.0143, 0.0128, 0.0144, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:20:54,794 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198220.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:20:55,428 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 02:21:13,345 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1700, 4.2539, 4.0598, 3.7814, 3.7683, 4.1820, 3.8845, 3.9266], device='cuda:5'), covar=tensor([0.0664, 0.0545, 0.0308, 0.0290, 0.0827, 0.0476, 0.0804, 0.0531], device='cuda:5'), in_proj_covar=tensor([0.0289, 0.0417, 0.0336, 0.0332, 0.0348, 0.0387, 0.0232, 0.0403], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:21:35,347 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198248.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:37,094 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198249.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:40,987 INFO [train.py:904] (5/8) Epoch 20, batch 5400, loss[loss=0.1749, simple_loss=0.2754, pruned_loss=0.03718, over 16759.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2673, pruned_loss=0.0434, over 3234986.17 frames. ], batch size: 83, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:22:15,987 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 1.945e+02 2.236e+02 2.644e+02 4.931e+02, threshold=4.472e+02, percent-clipped=2.0 2023-05-01 02:22:57,916 INFO [train.py:904] (5/8) Epoch 20, batch 5450, loss[loss=0.1798, simple_loss=0.2717, pruned_loss=0.04401, over 16446.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2703, pruned_loss=0.04472, over 3220615.20 frames. ], batch size: 68, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:23:04,357 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0169, 3.9340, 4.6172, 2.4247, 4.8850, 4.8827, 3.4238, 3.5268], device='cuda:5'), covar=tensor([0.0773, 0.0299, 0.0147, 0.1110, 0.0043, 0.0087, 0.0367, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0110, 0.0098, 0.0141, 0.0081, 0.0125, 0.0129, 0.0134], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:24:08,107 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 02:24:14,766 INFO [train.py:904] (5/8) Epoch 20, batch 5500, loss[loss=0.2233, simple_loss=0.3071, pruned_loss=0.06975, over 15274.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2768, pruned_loss=0.04873, over 3184176.78 frames. ], batch size: 190, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:24:51,695 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.873e+02 3.524e+02 4.449e+02 7.452e+02, threshold=7.049e+02, percent-clipped=24.0 2023-05-01 02:25:34,176 INFO [train.py:904] (5/8) Epoch 20, batch 5550, loss[loss=0.2342, simple_loss=0.3176, pruned_loss=0.07543, over 15190.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2841, pruned_loss=0.05354, over 3169185.22 frames. ], batch size: 190, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:26:02,767 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4797, 3.5331, 4.0328, 1.7193, 4.1841, 4.2523, 3.1602, 2.9110], device='cuda:5'), covar=tensor([0.0972, 0.0291, 0.0193, 0.1482, 0.0080, 0.0144, 0.0411, 0.0578], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0109, 0.0098, 0.0140, 0.0080, 0.0125, 0.0129, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:26:30,795 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198437.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:26:50,284 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 02:26:53,772 INFO [train.py:904] (5/8) Epoch 20, batch 5600, loss[loss=0.1929, simple_loss=0.2791, pruned_loss=0.05341, over 16769.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2894, pruned_loss=0.05831, over 3112244.72 frames. ], batch size: 83, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:27:10,924 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-01 02:27:34,785 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.278e+02 3.283e+02 3.655e+02 4.380e+02 7.107e+02, threshold=7.309e+02, percent-clipped=1.0 2023-05-01 02:28:14,195 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198499.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:28:18,613 INFO [train.py:904] (5/8) Epoch 20, batch 5650, loss[loss=0.1923, simple_loss=0.2835, pruned_loss=0.05051, over 16449.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2947, pruned_loss=0.06261, over 3072493.32 frames. ], batch size: 68, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:28:29,793 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6091, 2.5408, 1.8691, 2.6930, 2.1413, 2.7413, 2.1698, 2.3657], device='cuda:5'), covar=tensor([0.0364, 0.0401, 0.1262, 0.0256, 0.0624, 0.0493, 0.1091, 0.0621], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0175, 0.0193, 0.0157, 0.0175, 0.0213, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:28:51,771 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-01 02:29:13,589 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3690, 5.3917, 5.1414, 4.5111, 5.2952, 1.8175, 4.9869, 4.9595], device='cuda:5'), covar=tensor([0.0058, 0.0054, 0.0173, 0.0333, 0.0070, 0.2665, 0.0106, 0.0180], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0147, 0.0191, 0.0173, 0.0168, 0.0201, 0.0181, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:29:32,571 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198548.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:29:36,989 INFO [train.py:904] (5/8) Epoch 20, batch 5700, loss[loss=0.205, simple_loss=0.2962, pruned_loss=0.05691, over 16521.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2967, pruned_loss=0.06477, over 3061152.07 frames. ], batch size: 75, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:29:50,417 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:30:14,614 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 3.352e+02 4.045e+02 5.075e+02 1.137e+03, threshold=8.090e+02, percent-clipped=5.0 2023-05-01 02:30:47,282 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198596.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:30:55,558 INFO [train.py:904] (5/8) Epoch 20, batch 5750, loss[loss=0.206, simple_loss=0.2949, pruned_loss=0.05851, over 16925.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2992, pruned_loss=0.066, over 3038613.26 frames. ], batch size: 109, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:31:17,265 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:32:03,646 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5399, 3.5982, 3.3812, 3.0696, 3.1979, 3.5199, 3.3362, 3.3550], device='cuda:5'), covar=tensor([0.0623, 0.0634, 0.0279, 0.0270, 0.0535, 0.0439, 0.1270, 0.0472], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0410, 0.0331, 0.0325, 0.0342, 0.0379, 0.0228, 0.0396], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:32:16,961 INFO [train.py:904] (5/8) Epoch 20, batch 5800, loss[loss=0.1836, simple_loss=0.2802, pruned_loss=0.04354, over 16816.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2987, pruned_loss=0.06481, over 3038727.99 frames. ], batch size: 102, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:32:53,806 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.730e+02 3.384e+02 4.113e+02 5.841e+02, threshold=6.768e+02, percent-clipped=0.0 2023-05-01 02:32:54,454 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:33:34,967 INFO [train.py:904] (5/8) Epoch 20, batch 5850, loss[loss=0.2116, simple_loss=0.2938, pruned_loss=0.06471, over 16394.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2967, pruned_loss=0.06327, over 3053086.48 frames. ], batch size: 146, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:34:03,541 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3769, 4.0594, 3.9531, 2.7475, 3.6185, 4.0582, 3.6513, 2.3191], device='cuda:5'), covar=tensor([0.0518, 0.0041, 0.0054, 0.0343, 0.0096, 0.0092, 0.0087, 0.0415], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0132, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:34:30,937 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198737.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:34:34,054 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9753, 3.4968, 3.4361, 2.2120, 3.2262, 3.5192, 3.2980, 1.9169], device='cuda:5'), covar=tensor([0.0582, 0.0052, 0.0067, 0.0423, 0.0105, 0.0119, 0.0092, 0.0476], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0133, 0.0096, 0.0107, 0.0093, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:34:53,519 INFO [train.py:904] (5/8) Epoch 20, batch 5900, loss[loss=0.2364, simple_loss=0.3026, pruned_loss=0.08513, over 11461.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2961, pruned_loss=0.06251, over 3065171.99 frames. ], batch size: 248, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:35:34,252 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.541e+02 2.837e+02 3.258e+02 4.009e+02 8.301e+02, threshold=6.515e+02, percent-clipped=1.0 2023-05-01 02:35:48,535 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:36:14,217 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 02:36:14,580 INFO [train.py:904] (5/8) Epoch 20, batch 5950, loss[loss=0.2088, simple_loss=0.2951, pruned_loss=0.06122, over 16317.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2969, pruned_loss=0.06178, over 3072197.99 frames. ], batch size: 165, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:31,026 INFO [train.py:904] (5/8) Epoch 20, batch 6000, loss[loss=0.2038, simple_loss=0.2895, pruned_loss=0.05901, over 15345.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2964, pruned_loss=0.06187, over 3061774.07 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:31,027 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 02:37:41,828 INFO [train.py:938] (5/8) Epoch 20, validation: loss=0.1516, simple_loss=0.2644, pruned_loss=0.01942, over 944034.00 frames. 2023-05-01 02:37:41,829 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 02:37:42,292 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:37:46,512 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198855.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:37:46,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7242, 4.7238, 5.1333, 5.0980, 5.1245, 4.7901, 4.7391, 4.5177], device='cuda:5'), covar=tensor([0.0347, 0.0623, 0.0351, 0.0408, 0.0458, 0.0419, 0.0999, 0.0556], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0442, 0.0426, 0.0397, 0.0475, 0.0450, 0.0540, 0.0362], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 02:37:59,696 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5594, 3.5519, 3.5320, 2.8342, 3.4665, 2.0560, 3.2376, 2.8738], device='cuda:5'), covar=tensor([0.0142, 0.0121, 0.0171, 0.0231, 0.0095, 0.2307, 0.0131, 0.0245], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0148, 0.0192, 0.0174, 0.0168, 0.0202, 0.0182, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:38:06,956 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2958, 1.6507, 1.9558, 2.1744, 2.3049, 2.5426, 1.7387, 2.4387], device='cuda:5'), covar=tensor([0.0212, 0.0465, 0.0301, 0.0334, 0.0296, 0.0183, 0.0501, 0.0158], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0191, 0.0175, 0.0179, 0.0192, 0.0149, 0.0192, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:38:12,427 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8129, 1.4370, 1.6995, 1.6718, 1.8491, 1.9135, 1.6696, 1.7762], device='cuda:5'), covar=tensor([0.0242, 0.0360, 0.0206, 0.0298, 0.0252, 0.0161, 0.0372, 0.0135], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0191, 0.0175, 0.0179, 0.0192, 0.0149, 0.0192, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:38:17,162 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 2.764e+02 3.257e+02 3.996e+02 6.001e+02, threshold=6.515e+02, percent-clipped=0.0 2023-05-01 02:38:28,059 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:38:52,736 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 02:38:56,348 INFO [train.py:904] (5/8) Epoch 20, batch 6050, loss[loss=0.204, simple_loss=0.2789, pruned_loss=0.06461, over 11974.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2948, pruned_loss=0.06165, over 3058055.05 frames. ], batch size: 246, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:39:12,795 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-05-01 02:39:14,629 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198913.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:39:16,014 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1156, 4.8844, 5.0907, 5.2890, 5.5394, 4.7993, 5.4948, 5.5078], device='cuda:5'), covar=tensor([0.1893, 0.1356, 0.1834, 0.0845, 0.0598, 0.0986, 0.0754, 0.0696], device='cuda:5'), in_proj_covar=tensor([0.0620, 0.0765, 0.0890, 0.0778, 0.0585, 0.0611, 0.0627, 0.0724], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:39:36,319 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198927.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:39:57,972 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-05-01 02:40:05,098 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198944.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:17,601 INFO [train.py:904] (5/8) Epoch 20, batch 6100, loss[loss=0.2491, simple_loss=0.3144, pruned_loss=0.09185, over 11559.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2941, pruned_loss=0.06042, over 3082106.63 frames. ], batch size: 247, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:40:18,501 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3676, 4.4588, 4.6174, 4.4411, 4.5575, 5.0431, 4.5770, 4.3298], device='cuda:5'), covar=tensor([0.1525, 0.1985, 0.2309, 0.1938, 0.2409, 0.0966, 0.1623, 0.2410], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0583, 0.0639, 0.0482, 0.0643, 0.0671, 0.0499, 0.0650], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:40:39,689 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 02:40:46,264 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198971.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:53,274 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.636e+02 3.021e+02 3.731e+02 7.049e+02, threshold=6.042e+02, percent-clipped=1.0 2023-05-01 02:41:12,471 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:41:28,413 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 02:41:32,345 INFO [train.py:904] (5/8) Epoch 20, batch 6150, loss[loss=0.1796, simple_loss=0.2761, pruned_loss=0.04158, over 16828.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2923, pruned_loss=0.05973, over 3088965.16 frames. ], batch size: 102, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:41:55,138 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3763, 3.3471, 1.9524, 3.7813, 2.6205, 3.7839, 2.1450, 2.6546], device='cuda:5'), covar=tensor([0.0274, 0.0404, 0.1735, 0.0191, 0.0777, 0.0528, 0.1612, 0.0779], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0175, 0.0192, 0.0157, 0.0174, 0.0213, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:42:22,296 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8486, 2.7058, 2.8370, 2.0987, 2.6421, 2.1789, 2.6939, 2.9447], device='cuda:5'), covar=tensor([0.0298, 0.0783, 0.0543, 0.1844, 0.0815, 0.0904, 0.0602, 0.0781], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0161, 0.0165, 0.0150, 0.0142, 0.0128, 0.0143, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:42:49,580 INFO [train.py:904] (5/8) Epoch 20, batch 6200, loss[loss=0.1737, simple_loss=0.2618, pruned_loss=0.0428, over 16716.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2897, pruned_loss=0.0588, over 3097855.97 frames. ], batch size: 134, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:43:17,645 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-05-01 02:43:28,025 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.934e+02 3.552e+02 4.446e+02 1.093e+03, threshold=7.104e+02, percent-clipped=2.0 2023-05-01 02:44:02,000 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9200, 4.9064, 4.8068, 3.5240, 4.8656, 1.6241, 4.4615, 4.4630], device='cuda:5'), covar=tensor([0.0168, 0.0135, 0.0241, 0.0688, 0.0150, 0.3461, 0.0229, 0.0343], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0147, 0.0191, 0.0173, 0.0168, 0.0202, 0.0181, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:44:06,526 INFO [train.py:904] (5/8) Epoch 20, batch 6250, loss[loss=0.2092, simple_loss=0.2863, pruned_loss=0.06602, over 11374.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2891, pruned_loss=0.05853, over 3088403.71 frames. ], batch size: 248, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:44:57,092 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6107, 4.7838, 4.9450, 4.7450, 4.8045, 5.3464, 4.8331, 4.6289], device='cuda:5'), covar=tensor([0.1274, 0.1844, 0.2541, 0.1990, 0.2366, 0.0997, 0.1741, 0.2489], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0583, 0.0642, 0.0484, 0.0643, 0.0672, 0.0501, 0.0651], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:45:21,382 INFO [train.py:904] (5/8) Epoch 20, batch 6300, loss[loss=0.1999, simple_loss=0.2851, pruned_loss=0.05737, over 16968.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2888, pruned_loss=0.05774, over 3097296.93 frames. ], batch size: 109, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:45:26,533 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199155.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:45:59,350 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:00,085 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.632e+02 3.175e+02 3.943e+02 7.347e+02, threshold=6.350e+02, percent-clipped=2.0 2023-05-01 02:46:22,903 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-05-01 02:46:38,974 INFO [train.py:904] (5/8) Epoch 20, batch 6350, loss[loss=0.2072, simple_loss=0.2907, pruned_loss=0.06186, over 16926.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2896, pruned_loss=0.05898, over 3096968.68 frames. ], batch size: 109, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:46:40,594 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:46,515 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6037, 2.6083, 1.8866, 2.6870, 2.1743, 2.7544, 2.1206, 2.3404], device='cuda:5'), covar=tensor([0.0281, 0.0345, 0.1154, 0.0286, 0.0599, 0.0542, 0.1117, 0.0583], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0175, 0.0193, 0.0157, 0.0174, 0.0213, 0.0200, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:46:48,066 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199208.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:30,906 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199237.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:47:34,243 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199239.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:44,720 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0266, 3.7651, 3.7343, 2.3260, 3.4150, 3.7939, 3.4673, 2.0089], device='cuda:5'), covar=tensor([0.0607, 0.0049, 0.0051, 0.0427, 0.0098, 0.0106, 0.0088, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0133, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:47:52,866 INFO [train.py:904] (5/8) Epoch 20, batch 6400, loss[loss=0.169, simple_loss=0.2652, pruned_loss=0.03642, over 16880.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2907, pruned_loss=0.06013, over 3090400.38 frames. ], batch size: 96, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:48:21,620 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199271.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:48:29,165 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.234e+02 2.937e+02 3.403e+02 4.348e+02 9.192e+02, threshold=6.807e+02, percent-clipped=3.0 2023-05-01 02:48:39,469 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199283.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:48:47,671 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0682, 3.2708, 3.4620, 1.7066, 3.6194, 3.7760, 2.9479, 2.6354], device='cuda:5'), covar=tensor([0.1156, 0.0213, 0.0184, 0.1365, 0.0091, 0.0177, 0.0404, 0.0581], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0108, 0.0097, 0.0138, 0.0079, 0.0123, 0.0127, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:49:07,113 INFO [train.py:904] (5/8) Epoch 20, batch 6450, loss[loss=0.1977, simple_loss=0.288, pruned_loss=0.05372, over 17058.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2908, pruned_loss=0.05989, over 3069978.98 frames. ], batch size: 50, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:49:19,592 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1446, 2.1378, 2.2450, 3.7330, 2.1196, 2.5075, 2.2599, 2.3438], device='cuda:5'), covar=tensor([0.1314, 0.3511, 0.2888, 0.0561, 0.4050, 0.2372, 0.3298, 0.3332], device='cuda:5'), in_proj_covar=tensor([0.0395, 0.0438, 0.0361, 0.0321, 0.0432, 0.0505, 0.0408, 0.0513], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:49:33,334 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:50:24,682 INFO [train.py:904] (5/8) Epoch 20, batch 6500, loss[loss=0.2022, simple_loss=0.2891, pruned_loss=0.0577, over 16467.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2889, pruned_loss=0.0591, over 3085645.62 frames. ], batch size: 68, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:50:35,490 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-05-01 02:51:02,008 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.867e+02 3.442e+02 4.204e+02 8.359e+02, threshold=6.884e+02, percent-clipped=2.0 2023-05-01 02:51:21,781 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 02:51:39,177 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199400.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:51:41,757 INFO [train.py:904] (5/8) Epoch 20, batch 6550, loss[loss=0.2256, simple_loss=0.3021, pruned_loss=0.07453, over 11883.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2915, pruned_loss=0.05985, over 3108081.22 frames. ], batch size: 246, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:52:56,172 INFO [train.py:904] (5/8) Epoch 20, batch 6600, loss[loss=0.2504, simple_loss=0.3178, pruned_loss=0.09144, over 11976.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2943, pruned_loss=0.06099, over 3090801.30 frames. ], batch size: 247, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:53:05,819 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:53:10,074 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199461.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:53:31,644 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-05-01 02:53:33,380 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.722e+02 3.309e+02 3.917e+02 9.551e+02, threshold=6.618e+02, percent-clipped=2.0 2023-05-01 02:53:59,411 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199494.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:54:11,574 INFO [train.py:904] (5/8) Epoch 20, batch 6650, loss[loss=0.2698, simple_loss=0.3267, pruned_loss=0.1064, over 11625.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2947, pruned_loss=0.06174, over 3082900.76 frames. ], batch size: 246, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:54:20,083 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199508.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:54:28,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4629, 3.3288, 2.7183, 2.1189, 2.2793, 2.3087, 3.5789, 3.1142], device='cuda:5'), covar=tensor([0.3062, 0.0968, 0.1951, 0.2790, 0.2625, 0.2147, 0.0510, 0.1396], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0267, 0.0302, 0.0306, 0.0294, 0.0253, 0.0292, 0.0331], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:54:37,311 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199519.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:54:56,465 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199532.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:55:00,519 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:55:06,300 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199539.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:55:06,514 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4227, 2.9957, 2.5720, 2.2289, 2.2969, 2.2071, 2.9529, 2.8574], device='cuda:5'), covar=tensor([0.2626, 0.0770, 0.1744, 0.2487, 0.2481, 0.2174, 0.0578, 0.1298], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0267, 0.0302, 0.0306, 0.0295, 0.0253, 0.0292, 0.0331], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 02:55:25,180 INFO [train.py:904] (5/8) Epoch 20, batch 6700, loss[loss=0.2036, simple_loss=0.2944, pruned_loss=0.05645, over 16740.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2939, pruned_loss=0.06221, over 3059562.91 frames. ], batch size: 124, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:55:29,383 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1090, 5.0948, 4.9670, 4.5403, 4.6434, 5.0152, 4.9288, 4.6627], device='cuda:5'), covar=tensor([0.0644, 0.0438, 0.0274, 0.0320, 0.0966, 0.0415, 0.0298, 0.0675], device='cuda:5'), in_proj_covar=tensor([0.0282, 0.0407, 0.0327, 0.0322, 0.0337, 0.0376, 0.0227, 0.0391], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:55:31,396 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199555.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:55:32,680 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199556.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:02,761 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.649e+02 2.838e+02 3.650e+02 4.570e+02 8.139e+02, threshold=7.299e+02, percent-clipped=7.0 2023-05-01 02:56:13,372 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:18,219 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:31,568 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199596.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:56:38,615 INFO [train.py:904] (5/8) Epoch 20, batch 6750, loss[loss=0.1849, simple_loss=0.271, pruned_loss=0.04937, over 16778.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2929, pruned_loss=0.06236, over 3053002.70 frames. ], batch size: 102, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:56:47,922 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1578, 1.5452, 1.9494, 2.0971, 2.2192, 2.3606, 1.7870, 2.2498], device='cuda:5'), covar=tensor([0.0237, 0.0473, 0.0270, 0.0298, 0.0297, 0.0177, 0.0450, 0.0141], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0191, 0.0175, 0.0180, 0.0192, 0.0149, 0.0192, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 02:57:01,106 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-01 02:57:20,747 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:57:53,104 INFO [train.py:904] (5/8) Epoch 20, batch 6800, loss[loss=0.2364, simple_loss=0.3066, pruned_loss=0.08314, over 11922.00 frames. ], tot_loss[loss=0.209, simple_loss=0.293, pruned_loss=0.0625, over 3032962.21 frames. ], batch size: 246, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:58:29,266 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.711e+02 3.238e+02 3.835e+02 6.988e+02, threshold=6.476e+02, percent-clipped=0.0 2023-05-01 02:59:05,443 INFO [train.py:904] (5/8) Epoch 20, batch 6850, loss[loss=0.1949, simple_loss=0.302, pruned_loss=0.04389, over 16902.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2934, pruned_loss=0.0625, over 3032358.24 frames. ], batch size: 109, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:59:09,712 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7144, 2.5707, 2.4270, 3.9031, 2.6348, 3.9880, 1.4532, 2.8917], device='cuda:5'), covar=tensor([0.1353, 0.0810, 0.1212, 0.0154, 0.0200, 0.0382, 0.1792, 0.0765], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0173, 0.0194, 0.0184, 0.0206, 0.0213, 0.0199, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 02:59:15,814 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5462, 5.5000, 5.3701, 4.5160, 5.3601, 1.8755, 5.0773, 5.0331], device='cuda:5'), covar=tensor([0.0128, 0.0125, 0.0205, 0.0484, 0.0147, 0.2873, 0.0243, 0.0218], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0149, 0.0193, 0.0176, 0.0170, 0.0204, 0.0183, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:00:20,629 INFO [train.py:904] (5/8) Epoch 20, batch 6900, loss[loss=0.2007, simple_loss=0.2959, pruned_loss=0.05277, over 16471.00 frames. ], tot_loss[loss=0.209, simple_loss=0.295, pruned_loss=0.06155, over 3051156.72 frames. ], batch size: 75, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:00:26,889 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199756.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:00:59,799 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 2.558e+02 3.121e+02 3.929e+02 6.424e+02, threshold=6.242e+02, percent-clipped=0.0 2023-05-01 03:01:06,185 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8093, 2.7631, 2.8803, 2.1526, 2.6964, 2.2095, 2.7557, 2.9423], device='cuda:5'), covar=tensor([0.0262, 0.0698, 0.0457, 0.1585, 0.0705, 0.0844, 0.0541, 0.0639], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0161, 0.0165, 0.0150, 0.0142, 0.0128, 0.0142, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:01:36,164 INFO [train.py:904] (5/8) Epoch 20, batch 6950, loss[loss=0.1981, simple_loss=0.2944, pruned_loss=0.05085, over 16696.00 frames. ], tot_loss[loss=0.212, simple_loss=0.297, pruned_loss=0.06346, over 3036949.12 frames. ], batch size: 89, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:01:55,040 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199814.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:02:19,390 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6751, 4.6294, 4.4855, 3.7460, 4.5552, 1.7946, 4.3477, 4.2911], device='cuda:5'), covar=tensor([0.0102, 0.0089, 0.0190, 0.0372, 0.0101, 0.2701, 0.0137, 0.0221], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0148, 0.0192, 0.0174, 0.0168, 0.0202, 0.0181, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:02:22,434 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:02:47,652 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199850.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:02:49,634 INFO [train.py:904] (5/8) Epoch 20, batch 7000, loss[loss=0.1881, simple_loss=0.2896, pruned_loss=0.04328, over 16787.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2967, pruned_loss=0.06209, over 3053283.96 frames. ], batch size: 83, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:03:29,996 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.829e+02 3.376e+02 4.341e+02 9.860e+02, threshold=6.751e+02, percent-clipped=7.0 2023-05-01 03:03:33,463 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199880.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:03:49,934 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199891.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:04:07,208 INFO [train.py:904] (5/8) Epoch 20, batch 7050, loss[loss=0.2002, simple_loss=0.2945, pruned_loss=0.05294, over 16288.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2974, pruned_loss=0.06171, over 3063195.76 frames. ], batch size: 165, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:04:18,555 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:04:40,071 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0047, 2.3043, 2.2133, 2.6981, 1.7746, 3.0976, 1.8017, 2.6157], device='cuda:5'), covar=tensor([0.1217, 0.0739, 0.1250, 0.0225, 0.0145, 0.0362, 0.1621, 0.0800], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0174, 0.0195, 0.0186, 0.0208, 0.0214, 0.0201, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:05:19,975 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6329, 4.2475, 4.1831, 2.9799, 3.6946, 4.1672, 3.7114, 2.3590], device='cuda:5'), covar=tensor([0.0480, 0.0037, 0.0043, 0.0340, 0.0108, 0.0111, 0.0091, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0133, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 03:05:25,781 INFO [train.py:904] (5/8) Epoch 20, batch 7100, loss[loss=0.2058, simple_loss=0.2973, pruned_loss=0.05713, over 16887.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2964, pruned_loss=0.06192, over 3030473.40 frames. ], batch size: 96, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:05:54,844 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199970.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:06:05,959 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.782e+02 3.561e+02 4.163e+02 7.041e+02, threshold=7.121e+02, percent-clipped=1.0 2023-05-01 03:06:37,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8523, 3.2341, 3.3283, 2.0825, 2.9011, 2.1808, 3.4160, 3.4522], device='cuda:5'), covar=tensor([0.0269, 0.0770, 0.0603, 0.2040, 0.0844, 0.1018, 0.0670, 0.1035], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0162, 0.0166, 0.0152, 0.0143, 0.0129, 0.0144, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:06:45,339 INFO [train.py:904] (5/8) Epoch 20, batch 7150, loss[loss=0.2465, simple_loss=0.3086, pruned_loss=0.09221, over 11128.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2935, pruned_loss=0.06072, over 3066737.90 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:07:05,520 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 03:07:59,656 INFO [train.py:904] (5/8) Epoch 20, batch 7200, loss[loss=0.2108, simple_loss=0.3015, pruned_loss=0.06007, over 11663.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2917, pruned_loss=0.05968, over 3048489.72 frames. ], batch size: 247, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:08:06,606 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200056.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:08:27,789 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-01 03:08:40,500 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.812e+02 2.598e+02 3.072e+02 3.513e+02 6.350e+02, threshold=6.144e+02, percent-clipped=0.0 2023-05-01 03:09:19,974 INFO [train.py:904] (5/8) Epoch 20, batch 7250, loss[loss=0.1663, simple_loss=0.2543, pruned_loss=0.03914, over 16581.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2894, pruned_loss=0.05866, over 3035612.43 frames. ], batch size: 75, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:09:23,331 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200104.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:09:37,495 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:10:31,975 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200150.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:10:35,114 INFO [train.py:904] (5/8) Epoch 20, batch 7300, loss[loss=0.2005, simple_loss=0.2913, pruned_loss=0.05484, over 15349.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.289, pruned_loss=0.05818, over 3048363.62 frames. ], batch size: 191, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:10:50,982 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200162.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:11:14,682 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200177.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:11:15,883 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.904e+02 3.413e+02 4.614e+02 7.836e+02, threshold=6.825e+02, percent-clipped=9.0 2023-05-01 03:11:37,025 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200191.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:11:47,791 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200198.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:11:52,595 INFO [train.py:904] (5/8) Epoch 20, batch 7350, loss[loss=0.2326, simple_loss=0.3026, pruned_loss=0.0813, over 11363.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2901, pruned_loss=0.05908, over 3045661.38 frames. ], batch size: 246, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:12:51,303 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200238.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:12:52,209 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200239.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:13:12,414 INFO [train.py:904] (5/8) Epoch 20, batch 7400, loss[loss=0.2684, simple_loss=0.332, pruned_loss=0.1024, over 11117.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2903, pruned_loss=0.05942, over 3053260.10 frames. ], batch size: 247, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:13:33,081 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:13:52,750 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.844e+02 2.893e+02 3.356e+02 4.091e+02 7.315e+02, threshold=6.713e+02, percent-clipped=2.0 2023-05-01 03:13:55,413 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200279.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:14:32,173 INFO [train.py:904] (5/8) Epoch 20, batch 7450, loss[loss=0.1799, simple_loss=0.2584, pruned_loss=0.05076, over 16704.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2915, pruned_loss=0.06048, over 3061418.26 frames. ], batch size: 39, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:15:36,342 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200340.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:15:54,172 INFO [train.py:904] (5/8) Epoch 20, batch 7500, loss[loss=0.1914, simple_loss=0.2732, pruned_loss=0.05484, over 16970.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2913, pruned_loss=0.0592, over 3091662.91 frames. ], batch size: 55, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:16:34,103 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 2.766e+02 3.444e+02 4.330e+02 6.961e+02, threshold=6.888e+02, percent-clipped=1.0 2023-05-01 03:17:10,897 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1099, 2.0449, 1.7782, 1.7766, 2.2888, 1.9540, 1.9555, 2.3671], device='cuda:5'), covar=tensor([0.0196, 0.0397, 0.0522, 0.0430, 0.0250, 0.0352, 0.0226, 0.0245], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0227, 0.0220, 0.0220, 0.0229, 0.0228, 0.0227, 0.0224], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:17:11,461 INFO [train.py:904] (5/8) Epoch 20, batch 7550, loss[loss=0.199, simple_loss=0.2838, pruned_loss=0.05715, over 15284.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2904, pruned_loss=0.05946, over 3083743.67 frames. ], batch size: 191, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:17:57,395 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3545, 5.3936, 5.1910, 4.7262, 4.7986, 5.2459, 5.2444, 4.9005], device='cuda:5'), covar=tensor([0.0678, 0.0439, 0.0289, 0.0327, 0.1026, 0.0462, 0.0244, 0.0687], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0402, 0.0324, 0.0320, 0.0333, 0.0373, 0.0224, 0.0388], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:18:00,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 03:18:26,307 INFO [train.py:904] (5/8) Epoch 20, batch 7600, loss[loss=0.2442, simple_loss=0.3193, pruned_loss=0.08459, over 16380.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2898, pruned_loss=0.05984, over 3083971.87 frames. ], batch size: 146, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:19:06,493 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.777e+02 3.348e+02 4.097e+02 9.491e+02, threshold=6.697e+02, percent-clipped=3.0 2023-05-01 03:19:33,402 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-05-01 03:19:44,057 INFO [train.py:904] (5/8) Epoch 20, batch 7650, loss[loss=0.2149, simple_loss=0.2984, pruned_loss=0.06574, over 16387.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2893, pruned_loss=0.05906, over 3122752.12 frames. ], batch size: 146, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:19:48,460 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6063, 3.8655, 3.9625, 2.1280, 3.3596, 2.5500, 3.9024, 4.0373], device='cuda:5'), covar=tensor([0.0228, 0.0739, 0.0535, 0.2176, 0.0799, 0.0985, 0.0624, 0.0982], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0161, 0.0166, 0.0152, 0.0143, 0.0129, 0.0144, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:20:33,328 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200533.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:21:02,185 INFO [train.py:904] (5/8) Epoch 20, batch 7700, loss[loss=0.2066, simple_loss=0.2977, pruned_loss=0.05771, over 16737.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2901, pruned_loss=0.05963, over 3121951.22 frames. ], batch size: 89, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:21:22,907 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200565.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:21:44,269 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.918e+02 3.681e+02 4.507e+02 8.362e+02, threshold=7.363e+02, percent-clipped=4.0 2023-05-01 03:22:21,785 INFO [train.py:904] (5/8) Epoch 20, batch 7750, loss[loss=0.2059, simple_loss=0.2924, pruned_loss=0.05977, over 17004.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2909, pruned_loss=0.06048, over 3089768.72 frames. ], batch size: 53, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:22:38,650 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200613.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:12,030 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200635.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:23:28,166 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:36,977 INFO [train.py:904] (5/8) Epoch 20, batch 7800, loss[loss=0.2214, simple_loss=0.3046, pruned_loss=0.06911, over 15340.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2921, pruned_loss=0.06096, over 3103777.10 frames. ], batch size: 190, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:24:18,683 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.831e+02 3.422e+02 4.035e+02 8.624e+02, threshold=6.845e+02, percent-clipped=1.0 2023-05-01 03:24:53,417 INFO [train.py:904] (5/8) Epoch 20, batch 7850, loss[loss=0.2135, simple_loss=0.2858, pruned_loss=0.0706, over 10956.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2924, pruned_loss=0.06026, over 3097658.59 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:25:00,320 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4772, 4.1831, 4.1232, 2.6325, 3.6820, 4.1410, 3.6751, 2.3182], device='cuda:5'), covar=tensor([0.0511, 0.0035, 0.0042, 0.0380, 0.0094, 0.0093, 0.0087, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0080, 0.0080, 0.0132, 0.0095, 0.0107, 0.0091, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 03:25:00,371 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:25:11,517 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-01 03:26:01,805 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5189, 3.4950, 3.4702, 2.7075, 3.3466, 2.1006, 3.1087, 2.7703], device='cuda:5'), covar=tensor([0.0152, 0.0125, 0.0177, 0.0208, 0.0099, 0.2261, 0.0139, 0.0212], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0147, 0.0190, 0.0173, 0.0167, 0.0201, 0.0180, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:26:08,722 INFO [train.py:904] (5/8) Epoch 20, batch 7900, loss[loss=0.2157, simple_loss=0.3158, pruned_loss=0.05778, over 16762.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2909, pruned_loss=0.05925, over 3116667.18 frames. ], batch size: 124, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:26:49,222 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 2.835e+02 3.346e+02 4.054e+02 6.354e+02, threshold=6.693e+02, percent-clipped=0.0 2023-05-01 03:27:27,138 INFO [train.py:904] (5/8) Epoch 20, batch 7950, loss[loss=0.2502, simple_loss=0.3146, pruned_loss=0.09287, over 11725.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2912, pruned_loss=0.05976, over 3114157.14 frames. ], batch size: 250, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:28:09,236 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5073, 2.7254, 2.2388, 2.5746, 3.0157, 2.7305, 3.0962, 3.2246], device='cuda:5'), covar=tensor([0.0113, 0.0337, 0.0505, 0.0345, 0.0245, 0.0327, 0.0226, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0195, 0.0226, 0.0218, 0.0219, 0.0228, 0.0227, 0.0226, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:28:09,466 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-05-01 03:28:16,170 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200833.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:28:40,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2806, 4.1291, 4.3830, 4.4857, 4.6424, 4.2583, 4.5847, 4.6556], device='cuda:5'), covar=tensor([0.1846, 0.1357, 0.1480, 0.0750, 0.0583, 0.1178, 0.0734, 0.0706], device='cuda:5'), in_proj_covar=tensor([0.0616, 0.0761, 0.0886, 0.0772, 0.0586, 0.0607, 0.0626, 0.0724], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:28:44,606 INFO [train.py:904] (5/8) Epoch 20, batch 8000, loss[loss=0.2542, simple_loss=0.3188, pruned_loss=0.09482, over 11449.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2917, pruned_loss=0.06032, over 3113261.59 frames. ], batch size: 246, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:28:50,965 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-01 03:29:24,826 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 3.046e+02 3.376e+02 4.028e+02 7.694e+02, threshold=6.753e+02, percent-clipped=2.0 2023-05-01 03:29:28,342 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200881.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:29:53,599 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-01 03:29:59,860 INFO [train.py:904] (5/8) Epoch 20, batch 8050, loss[loss=0.2055, simple_loss=0.2922, pruned_loss=0.0594, over 16635.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2922, pruned_loss=0.06075, over 3105726.83 frames. ], batch size: 134, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:30:25,799 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1345, 2.1790, 2.1706, 3.7975, 2.1746, 2.5814, 2.2664, 2.3288], device='cuda:5'), covar=tensor([0.1366, 0.3683, 0.3042, 0.0560, 0.4172, 0.2560, 0.3601, 0.3390], device='cuda:5'), in_proj_covar=tensor([0.0394, 0.0441, 0.0361, 0.0323, 0.0433, 0.0508, 0.0410, 0.0516], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:30:50,131 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:30:54,498 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5352, 3.5131, 3.5014, 2.6934, 3.4163, 2.0671, 3.2071, 2.8410], device='cuda:5'), covar=tensor([0.0162, 0.0137, 0.0186, 0.0236, 0.0113, 0.2316, 0.0145, 0.0244], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0148, 0.0190, 0.0174, 0.0168, 0.0202, 0.0181, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:31:12,110 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0311, 3.3182, 3.4400, 2.0781, 3.0023, 2.2888, 3.5288, 3.5544], device='cuda:5'), covar=tensor([0.0262, 0.0922, 0.0632, 0.2163, 0.0837, 0.0994, 0.0693, 0.1005], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0163, 0.0166, 0.0152, 0.0144, 0.0130, 0.0145, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:31:15,314 INFO [train.py:904] (5/8) Epoch 20, batch 8100, loss[loss=0.2521, simple_loss=0.3253, pruned_loss=0.08947, over 11581.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2915, pruned_loss=0.06041, over 3089830.79 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:31:57,148 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.634e+02 3.129e+02 3.819e+02 6.896e+02, threshold=6.259e+02, percent-clipped=1.0 2023-05-01 03:32:04,949 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200983.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:31,974 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:33,321 INFO [train.py:904] (5/8) Epoch 20, batch 8150, loss[loss=0.178, simple_loss=0.2669, pruned_loss=0.04451, over 17243.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2894, pruned_loss=0.0596, over 3093392.65 frames. ], batch size: 45, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:33:05,623 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7993, 3.8825, 4.1587, 4.1073, 4.1180, 3.8849, 3.9165, 3.9166], device='cuda:5'), covar=tensor([0.0372, 0.0623, 0.0388, 0.0438, 0.0474, 0.0445, 0.0799, 0.0482], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0443, 0.0430, 0.0398, 0.0478, 0.0451, 0.0543, 0.0362], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 03:33:51,832 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 03:33:52,119 INFO [train.py:904] (5/8) Epoch 20, batch 8200, loss[loss=0.1741, simple_loss=0.2677, pruned_loss=0.04027, over 16628.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2867, pruned_loss=0.05867, over 3098898.79 frames. ], batch size: 62, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:34:24,414 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201071.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:34:36,691 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 2.732e+02 3.390e+02 4.593e+02 1.143e+03, threshold=6.781e+02, percent-clipped=6.0 2023-05-01 03:34:39,350 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5665, 3.7793, 2.8425, 2.0730, 2.3598, 2.3484, 4.0069, 3.3159], device='cuda:5'), covar=tensor([0.3090, 0.0639, 0.1817, 0.3220, 0.2959, 0.2281, 0.0424, 0.1355], device='cuda:5'), in_proj_covar=tensor([0.0328, 0.0269, 0.0304, 0.0310, 0.0298, 0.0256, 0.0293, 0.0333], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 03:35:15,218 INFO [train.py:904] (5/8) Epoch 20, batch 8250, loss[loss=0.1854, simple_loss=0.2731, pruned_loss=0.04884, over 12165.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2856, pruned_loss=0.05577, over 3097396.38 frames. ], batch size: 247, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:35:51,315 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1716, 2.4635, 2.6919, 1.9582, 2.8184, 2.8238, 2.5380, 2.5382], device='cuda:5'), covar=tensor([0.0628, 0.0256, 0.0233, 0.1015, 0.0112, 0.0278, 0.0421, 0.0432], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0106, 0.0096, 0.0137, 0.0077, 0.0121, 0.0125, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 03:36:06,076 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:36:38,971 INFO [train.py:904] (5/8) Epoch 20, batch 8300, loss[loss=0.1619, simple_loss=0.2679, pruned_loss=0.02795, over 16821.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2831, pruned_loss=0.05313, over 3088016.02 frames. ], batch size: 102, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:37:22,452 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.376e+02 2.857e+02 3.605e+02 7.537e+02, threshold=5.714e+02, percent-clipped=2.0 2023-05-01 03:38:00,771 INFO [train.py:904] (5/8) Epoch 20, batch 8350, loss[loss=0.1999, simple_loss=0.2773, pruned_loss=0.06129, over 12064.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2817, pruned_loss=0.05144, over 3054981.33 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:03,130 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 03:39:23,238 INFO [train.py:904] (5/8) Epoch 20, batch 8400, loss[loss=0.1645, simple_loss=0.2483, pruned_loss=0.0403, over 12080.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2788, pruned_loss=0.04905, over 3063261.86 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:26,545 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:00,760 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:08,504 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.319e+02 2.651e+02 3.336e+02 5.393e+02, threshold=5.301e+02, percent-clipped=0.0 2023-05-01 03:40:44,440 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201301.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:45,028 INFO [train.py:904] (5/8) Epoch 20, batch 8450, loss[loss=0.1885, simple_loss=0.2928, pruned_loss=0.04212, over 16284.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2768, pruned_loss=0.04714, over 3064072.65 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:40:55,414 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5068, 3.3861, 2.8274, 2.0996, 2.1869, 2.2969, 3.4344, 3.0843], device='cuda:5'), covar=tensor([0.2786, 0.0632, 0.1601, 0.3242, 0.3071, 0.2276, 0.0434, 0.1272], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0263, 0.0297, 0.0303, 0.0289, 0.0250, 0.0286, 0.0325], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 03:40:59,025 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-05-01 03:41:05,884 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:41:39,400 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:42:02,495 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:42:07,359 INFO [train.py:904] (5/8) Epoch 20, batch 8500, loss[loss=0.1736, simple_loss=0.2473, pruned_loss=0.04995, over 11932.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2734, pruned_loss=0.04536, over 3044172.51 frames. ], batch size: 250, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:42:43,747 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 03:42:50,125 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.179e+02 2.765e+02 3.393e+02 6.239e+02, threshold=5.531e+02, percent-clipped=4.0 2023-05-01 03:43:31,065 INFO [train.py:904] (5/8) Epoch 20, batch 8550, loss[loss=0.1784, simple_loss=0.2771, pruned_loss=0.03984, over 16802.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2717, pruned_loss=0.04485, over 3031939.44 frames. ], batch size: 124, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:43:45,672 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8135, 1.8553, 2.3833, 2.7771, 2.6004, 3.1440, 2.1576, 3.2101], device='cuda:5'), covar=tensor([0.0205, 0.0550, 0.0351, 0.0296, 0.0312, 0.0160, 0.0478, 0.0134], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0189, 0.0174, 0.0178, 0.0191, 0.0148, 0.0191, 0.0144], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:44:19,204 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:45:09,645 INFO [train.py:904] (5/8) Epoch 20, batch 8600, loss[loss=0.1767, simple_loss=0.2583, pruned_loss=0.04754, over 12417.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2717, pruned_loss=0.04387, over 3021869.35 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:46:02,942 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 2.180e+02 2.549e+02 3.038e+02 6.489e+02, threshold=5.098e+02, percent-clipped=1.0 2023-05-01 03:46:04,256 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9377, 2.7373, 2.6540, 2.0315, 2.5506, 2.7968, 2.5720, 1.9768], device='cuda:5'), covar=tensor([0.0413, 0.0069, 0.0060, 0.0308, 0.0113, 0.0088, 0.0091, 0.0397], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0080, 0.0079, 0.0130, 0.0095, 0.0105, 0.0090, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 03:46:48,554 INFO [train.py:904] (5/8) Epoch 20, batch 8650, loss[loss=0.1566, simple_loss=0.2548, pruned_loss=0.02923, over 16707.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2704, pruned_loss=0.04258, over 3033247.05 frames. ], batch size: 134, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:48:27,520 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6918, 2.5812, 2.4371, 3.8456, 2.3996, 3.8352, 1.4324, 2.9236], device='cuda:5'), covar=tensor([0.1401, 0.0788, 0.1176, 0.0170, 0.0125, 0.0362, 0.1806, 0.0737], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0170, 0.0192, 0.0182, 0.0204, 0.0210, 0.0198, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:48:36,197 INFO [train.py:904] (5/8) Epoch 20, batch 8700, loss[loss=0.1631, simple_loss=0.2688, pruned_loss=0.02875, over 16853.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2684, pruned_loss=0.04158, over 3047630.66 frames. ], batch size: 102, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:49:04,153 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9002, 3.8997, 2.4835, 4.4420, 2.9944, 4.4129, 2.7187, 3.1860], device='cuda:5'), covar=tensor([0.0235, 0.0323, 0.1491, 0.0188, 0.0821, 0.0357, 0.1312, 0.0714], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0169, 0.0187, 0.0150, 0.0170, 0.0205, 0.0195, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 03:49:28,970 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.047e+02 2.487e+02 3.171e+02 4.995e+02, threshold=4.975e+02, percent-clipped=0.0 2023-05-01 03:50:13,620 INFO [train.py:904] (5/8) Epoch 20, batch 8750, loss[loss=0.1921, simple_loss=0.2903, pruned_loss=0.04693, over 15358.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2676, pruned_loss=0.04051, over 3064201.41 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:50:31,558 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201609.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:22,020 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201630.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:51,050 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201644.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:52:07,685 INFO [train.py:904] (5/8) Epoch 20, batch 8800, loss[loss=0.1667, simple_loss=0.2627, pruned_loss=0.03531, over 15227.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2655, pruned_loss=0.03925, over 3058939.82 frames. ], batch size: 190, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:52:12,540 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8861, 2.2777, 2.4462, 2.9718, 1.8814, 3.2651, 1.7068, 2.7708], device='cuda:5'), covar=tensor([0.1254, 0.0661, 0.0924, 0.0150, 0.0081, 0.0430, 0.1606, 0.0694], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0180, 0.0202, 0.0208, 0.0197, 0.0188], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:52:24,885 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201661.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:53:06,520 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.243e+02 2.623e+02 3.212e+02 5.763e+02, threshold=5.246e+02, percent-clipped=4.0 2023-05-01 03:53:51,641 INFO [train.py:904] (5/8) Epoch 20, batch 8850, loss[loss=0.1694, simple_loss=0.2767, pruned_loss=0.031, over 15143.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2682, pruned_loss=0.03868, over 3054789.27 frames. ], batch size: 190, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:53:57,899 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:00,635 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:23,733 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5185, 3.8362, 2.9079, 2.1733, 2.4169, 2.4329, 4.1302, 3.3430], device='cuda:5'), covar=tensor([0.3119, 0.0559, 0.1803, 0.2934, 0.2952, 0.2112, 0.0357, 0.1235], device='cuda:5'), in_proj_covar=tensor([0.0319, 0.0262, 0.0297, 0.0301, 0.0286, 0.0250, 0.0285, 0.0324], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 03:54:30,694 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-01 03:54:34,751 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:44,048 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3795, 3.6655, 4.0084, 2.1544, 3.3755, 2.5095, 3.7808, 3.7927], device='cuda:5'), covar=tensor([0.0243, 0.0816, 0.0442, 0.2013, 0.0665, 0.0916, 0.0561, 0.0973], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0156, 0.0160, 0.0148, 0.0140, 0.0126, 0.0140, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:54:45,916 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:59,635 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8488, 4.6690, 4.8602, 4.9995, 5.2347, 4.6427, 5.2405, 5.2492], device='cuda:5'), covar=tensor([0.2016, 0.1259, 0.1713, 0.0811, 0.0516, 0.0899, 0.0455, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0598, 0.0739, 0.0862, 0.0753, 0.0570, 0.0592, 0.0610, 0.0703], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:55:38,113 INFO [train.py:904] (5/8) Epoch 20, batch 8900, loss[loss=0.1825, simple_loss=0.2772, pruned_loss=0.04394, over 16356.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2693, pruned_loss=0.03848, over 3073335.13 frames. ], batch size: 146, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:55:57,074 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9634, 2.9399, 2.7377, 4.7079, 3.4542, 4.2272, 1.7135, 3.1061], device='cuda:5'), covar=tensor([0.1182, 0.0682, 0.1018, 0.0167, 0.0161, 0.0312, 0.1531, 0.0669], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0168, 0.0189, 0.0180, 0.0201, 0.0207, 0.0196, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 03:56:08,831 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:27,848 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:42,889 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3317, 4.1721, 4.4074, 4.5085, 4.6700, 4.2576, 4.6349, 4.6853], device='cuda:5'), covar=tensor([0.2026, 0.1309, 0.1453, 0.0735, 0.0550, 0.1093, 0.0570, 0.0712], device='cuda:5'), in_proj_covar=tensor([0.0594, 0.0733, 0.0855, 0.0749, 0.0566, 0.0587, 0.0606, 0.0698], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 03:56:46,607 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.216e+02 2.705e+02 3.264e+02 5.822e+02, threshold=5.410e+02, percent-clipped=1.0 2023-05-01 03:57:41,717 INFO [train.py:904] (5/8) Epoch 20, batch 8950, loss[loss=0.174, simple_loss=0.2665, pruned_loss=0.04078, over 17081.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2688, pruned_loss=0.03867, over 3088771.93 frames. ], batch size: 50, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:59:30,702 INFO [train.py:904] (5/8) Epoch 20, batch 9000, loss[loss=0.1537, simple_loss=0.2458, pruned_loss=0.03082, over 17211.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2651, pruned_loss=0.03716, over 3101137.90 frames. ], batch size: 44, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:59:30,703 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 03:59:41,114 INFO [train.py:938] (5/8) Epoch 20, validation: loss=0.1464, simple_loss=0.2502, pruned_loss=0.02125, over 944034.00 frames. 2023-05-01 03:59:41,115 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 04:00:41,798 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.083e+02 2.552e+02 3.280e+02 1.556e+03, threshold=5.104e+02, percent-clipped=3.0 2023-05-01 04:01:24,592 INFO [train.py:904] (5/8) Epoch 20, batch 9050, loss[loss=0.1752, simple_loss=0.262, pruned_loss=0.04424, over 16644.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2662, pruned_loss=0.03776, over 3095652.98 frames. ], batch size: 134, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:01:40,825 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:02:21,299 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:03:09,207 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5019, 4.5003, 4.2765, 3.6456, 4.3317, 1.5849, 4.1480, 4.0799], device='cuda:5'), covar=tensor([0.0109, 0.0106, 0.0233, 0.0361, 0.0132, 0.2913, 0.0145, 0.0277], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0145, 0.0185, 0.0168, 0.0164, 0.0199, 0.0177, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:03:12,556 INFO [train.py:904] (5/8) Epoch 20, batch 9100, loss[loss=0.174, simple_loss=0.2756, pruned_loss=0.0362, over 16657.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2662, pruned_loss=0.03826, over 3106917.79 frames. ], batch size: 76, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:03:22,656 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201957.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:07,599 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3586, 3.3007, 2.6521, 2.1204, 2.1142, 2.3319, 3.3550, 2.9747], device='cuda:5'), covar=tensor([0.3157, 0.0776, 0.1780, 0.3058, 0.2917, 0.2195, 0.0520, 0.1485], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0262, 0.0296, 0.0301, 0.0284, 0.0250, 0.0284, 0.0323], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 04:04:14,332 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:23,057 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.025e+02 2.515e+02 2.947e+02 5.747e+02, threshold=5.029e+02, percent-clipped=1.0 2023-05-01 04:04:43,631 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:05:02,168 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-05-01 04:05:11,356 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202000.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:05:14,334 INFO [train.py:904] (5/8) Epoch 20, batch 9150, loss[loss=0.1535, simple_loss=0.245, pruned_loss=0.03104, over 12116.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2659, pruned_loss=0.03775, over 3087375.64 frames. ], batch size: 250, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:05:20,016 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7163, 4.5126, 4.7200, 4.8546, 5.0404, 4.5115, 5.0358, 5.0285], device='cuda:5'), covar=tensor([0.1597, 0.1126, 0.1436, 0.0685, 0.0517, 0.0918, 0.0498, 0.0609], device='cuda:5'), in_proj_covar=tensor([0.0594, 0.0732, 0.0855, 0.0751, 0.0566, 0.0588, 0.0604, 0.0698], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:05:48,720 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202017.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:06:59,013 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202051.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:06:59,893 INFO [train.py:904] (5/8) Epoch 20, batch 9200, loss[loss=0.1668, simple_loss=0.2645, pruned_loss=0.03451, over 15307.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2616, pruned_loss=0.03673, over 3107394.29 frames. ], batch size: 192, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:07:20,317 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:07:55,480 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.302e+02 2.670e+02 3.574e+02 1.158e+03, threshold=5.341e+02, percent-clipped=6.0 2023-05-01 04:08:37,606 INFO [train.py:904] (5/8) Epoch 20, batch 9250, loss[loss=0.1512, simple_loss=0.2361, pruned_loss=0.03318, over 12208.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2622, pruned_loss=0.03695, over 3107796.77 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:08:49,144 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2186, 3.3329, 3.3945, 2.3622, 3.1120, 3.4399, 3.2695, 2.1563], device='cuda:5'), covar=tensor([0.0507, 0.0053, 0.0054, 0.0366, 0.0109, 0.0078, 0.0073, 0.0450], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0078, 0.0078, 0.0129, 0.0094, 0.0104, 0.0089, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 04:10:29,394 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5087, 2.7616, 2.7531, 4.0129, 2.6905, 3.9589, 1.3958, 3.1862], device='cuda:5'), covar=tensor([0.1543, 0.0778, 0.1065, 0.0169, 0.0153, 0.0356, 0.1900, 0.0641], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0181, 0.0200, 0.0208, 0.0197, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 04:10:29,969 INFO [train.py:904] (5/8) Epoch 20, batch 9300, loss[loss=0.1526, simple_loss=0.2458, pruned_loss=0.02971, over 16903.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2602, pruned_loss=0.0367, over 3073346.51 frames. ], batch size: 116, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:11:18,393 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1935, 3.2381, 1.9241, 3.4678, 2.3397, 3.4710, 2.1254, 2.6263], device='cuda:5'), covar=tensor([0.0284, 0.0367, 0.1542, 0.0298, 0.0862, 0.0550, 0.1446, 0.0719], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0167, 0.0186, 0.0149, 0.0169, 0.0203, 0.0195, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 04:11:37,349 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 1.963e+02 2.278e+02 2.874e+02 5.860e+02, threshold=4.556e+02, percent-clipped=2.0 2023-05-01 04:12:16,144 INFO [train.py:904] (5/8) Epoch 20, batch 9350, loss[loss=0.178, simple_loss=0.2729, pruned_loss=0.04154, over 16762.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2611, pruned_loss=0.03711, over 3077743.80 frames. ], batch size: 83, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:12:20,163 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6099, 2.7133, 2.3681, 4.0979, 2.6115, 4.0268, 1.4140, 3.0299], device='cuda:5'), covar=tensor([0.1590, 0.0803, 0.1312, 0.0206, 0.0175, 0.0389, 0.1998, 0.0720], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0181, 0.0201, 0.0209, 0.0198, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 04:13:01,979 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 04:13:44,590 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 04:13:44,654 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 04:13:59,190 INFO [train.py:904] (5/8) Epoch 20, batch 9400, loss[loss=0.1691, simple_loss=0.2766, pruned_loss=0.03081, over 16684.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.261, pruned_loss=0.03663, over 3085537.86 frames. ], batch size: 83, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:14:04,500 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2126, 4.3792, 4.5038, 4.2876, 4.3864, 4.8567, 4.4268, 4.1043], device='cuda:5'), covar=tensor([0.1712, 0.1965, 0.2269, 0.2347, 0.2394, 0.0964, 0.1615, 0.2649], device='cuda:5'), in_proj_covar=tensor([0.0376, 0.0550, 0.0607, 0.0454, 0.0604, 0.0635, 0.0476, 0.0612], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 04:15:00,976 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.064e+02 2.476e+02 2.960e+02 4.580e+02, threshold=4.951e+02, percent-clipped=1.0 2023-05-01 04:15:39,459 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:15:41,540 INFO [train.py:904] (5/8) Epoch 20, batch 9450, loss[loss=0.1621, simple_loss=0.2592, pruned_loss=0.03253, over 15372.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2626, pruned_loss=0.03687, over 3068888.06 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:16:13,661 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202317.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:16:13,712 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6957, 1.8514, 2.1709, 2.6496, 2.5168, 2.9602, 2.1381, 2.9154], device='cuda:5'), covar=tensor([0.0232, 0.0512, 0.0367, 0.0311, 0.0346, 0.0182, 0.0449, 0.0152], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0185, 0.0171, 0.0175, 0.0188, 0.0144, 0.0188, 0.0140], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:17:14,576 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202346.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:18,059 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202348.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:25,745 INFO [train.py:904] (5/8) Epoch 20, batch 9500, loss[loss=0.1517, simple_loss=0.2379, pruned_loss=0.03275, over 12481.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2618, pruned_loss=0.03656, over 3059755.20 frames. ], batch size: 246, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:17:49,721 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202362.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:55,127 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:18:26,630 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 2.130e+02 2.550e+02 3.364e+02 1.443e+03, threshold=5.099e+02, percent-clipped=7.0 2023-05-01 04:19:13,232 INFO [train.py:904] (5/8) Epoch 20, batch 9550, loss[loss=0.18, simple_loss=0.2708, pruned_loss=0.04456, over 12382.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2612, pruned_loss=0.03647, over 3062304.47 frames. ], batch size: 248, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:19:24,391 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2861, 3.6303, 3.6631, 2.4498, 3.2476, 3.6707, 3.4624, 2.2155], device='cuda:5'), covar=tensor([0.0494, 0.0040, 0.0042, 0.0375, 0.0104, 0.0076, 0.0068, 0.0439], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0078, 0.0078, 0.0129, 0.0094, 0.0104, 0.0089, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 04:19:32,680 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202410.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:20:55,237 INFO [train.py:904] (5/8) Epoch 20, batch 9600, loss[loss=0.2081, simple_loss=0.3027, pruned_loss=0.05682, over 16721.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.262, pruned_loss=0.0371, over 3036328.18 frames. ], batch size: 134, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:21:53,899 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.243e+02 2.599e+02 3.368e+02 6.147e+02, threshold=5.198e+02, percent-clipped=3.0 2023-05-01 04:22:20,080 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5468, 3.6263, 2.0904, 4.0070, 2.6608, 3.9563, 2.2178, 2.8446], device='cuda:5'), covar=tensor([0.0298, 0.0337, 0.1794, 0.0231, 0.0941, 0.0473, 0.1754, 0.0804], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0167, 0.0187, 0.0150, 0.0169, 0.0203, 0.0196, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 04:22:41,355 INFO [train.py:904] (5/8) Epoch 20, batch 9650, loss[loss=0.1662, simple_loss=0.2613, pruned_loss=0.03555, over 16854.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.264, pruned_loss=0.03757, over 3046327.11 frames. ], batch size: 124, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:23:42,226 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2982, 4.4366, 4.6794, 4.6699, 4.7251, 4.5109, 4.4145, 4.3819], device='cuda:5'), covar=tensor([0.0464, 0.0835, 0.0597, 0.0551, 0.0557, 0.0505, 0.0973, 0.0466], device='cuda:5'), in_proj_covar=tensor([0.0380, 0.0418, 0.0408, 0.0379, 0.0452, 0.0427, 0.0510, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 04:24:31,606 INFO [train.py:904] (5/8) Epoch 20, batch 9700, loss[loss=0.1603, simple_loss=0.2595, pruned_loss=0.03058, over 16879.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2632, pruned_loss=0.03749, over 3052733.83 frames. ], batch size: 102, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:25:37,517 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.240e+02 2.594e+02 3.217e+02 6.091e+02, threshold=5.188e+02, percent-clipped=1.0 2023-05-01 04:26:15,981 INFO [train.py:904] (5/8) Epoch 20, batch 9750, loss[loss=0.1752, simple_loss=0.2682, pruned_loss=0.04116, over 15310.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2622, pruned_loss=0.03773, over 3041153.69 frames. ], batch size: 190, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:27:47,511 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202646.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:27:57,092 INFO [train.py:904] (5/8) Epoch 20, batch 9800, loss[loss=0.1671, simple_loss=0.2736, pruned_loss=0.03024, over 16464.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.262, pruned_loss=0.03674, over 3055489.61 frames. ], batch size: 68, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:28:57,878 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.120e+02 2.499e+02 2.833e+02 7.051e+02, threshold=4.998e+02, percent-clipped=2.0 2023-05-01 04:29:20,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5584, 4.1072, 4.1602, 2.9167, 3.5936, 4.1080, 3.8041, 2.4405], device='cuda:5'), covar=tensor([0.0502, 0.0036, 0.0038, 0.0331, 0.0105, 0.0078, 0.0063, 0.0431], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0078, 0.0077, 0.0129, 0.0094, 0.0104, 0.0088, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 04:29:26,526 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:29:42,369 INFO [train.py:904] (5/8) Epoch 20, batch 9850, loss[loss=0.169, simple_loss=0.2639, pruned_loss=0.03712, over 16417.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2633, pruned_loss=0.03655, over 3072249.89 frames. ], batch size: 146, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:30:36,701 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-01 04:31:34,897 INFO [train.py:904] (5/8) Epoch 20, batch 9900, loss[loss=0.1757, simple_loss=0.2748, pruned_loss=0.03835, over 15256.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2632, pruned_loss=0.03637, over 3052334.86 frames. ], batch size: 191, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:31:38,511 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2894, 5.3066, 5.0463, 4.6565, 5.0741, 2.0458, 4.8494, 4.9038], device='cuda:5'), covar=tensor([0.0065, 0.0064, 0.0181, 0.0251, 0.0088, 0.2407, 0.0105, 0.0172], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0144, 0.0183, 0.0164, 0.0164, 0.0198, 0.0175, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:32:49,190 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.306e+02 2.190e+02 2.483e+02 3.109e+02 6.841e+02, threshold=4.966e+02, percent-clipped=3.0 2023-05-01 04:33:03,472 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1881, 1.4995, 1.8241, 2.0773, 2.1823, 2.2305, 1.7615, 2.2289], device='cuda:5'), covar=tensor([0.0218, 0.0512, 0.0326, 0.0313, 0.0314, 0.0195, 0.0495, 0.0146], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0183, 0.0169, 0.0173, 0.0185, 0.0142, 0.0186, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:33:30,690 INFO [train.py:904] (5/8) Epoch 20, batch 9950, loss[loss=0.1685, simple_loss=0.2726, pruned_loss=0.03221, over 16783.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2653, pruned_loss=0.03637, over 3068838.47 frames. ], batch size: 124, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:34:04,560 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7573, 4.8010, 4.6123, 4.2016, 4.2751, 4.6835, 4.5832, 4.3794], device='cuda:5'), covar=tensor([0.0612, 0.0745, 0.0343, 0.0336, 0.0939, 0.0597, 0.0387, 0.0686], device='cuda:5'), in_proj_covar=tensor([0.0271, 0.0386, 0.0313, 0.0308, 0.0319, 0.0356, 0.0215, 0.0372], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 04:34:42,301 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202830.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:34:45,815 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2761, 3.0795, 3.3910, 1.7849, 3.5328, 3.5910, 2.9115, 2.6806], device='cuda:5'), covar=tensor([0.0790, 0.0270, 0.0144, 0.1188, 0.0075, 0.0139, 0.0348, 0.0496], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0105, 0.0091, 0.0135, 0.0076, 0.0117, 0.0124, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 04:35:14,449 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 04:35:32,311 INFO [train.py:904] (5/8) Epoch 20, batch 10000, loss[loss=0.1886, simple_loss=0.2941, pruned_loss=0.04153, over 15501.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2645, pruned_loss=0.03606, over 3078809.19 frames. ], batch size: 192, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:36:36,182 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.233e+02 2.660e+02 3.205e+02 6.054e+02, threshold=5.319e+02, percent-clipped=4.0 2023-05-01 04:36:52,325 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202891.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:37:11,286 INFO [train.py:904] (5/8) Epoch 20, batch 10050, loss[loss=0.169, simple_loss=0.2699, pruned_loss=0.03402, over 15147.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.265, pruned_loss=0.0364, over 3090960.86 frames. ], batch size: 190, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:37:32,168 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8630, 4.8625, 5.2897, 5.2571, 5.2742, 4.9965, 4.9057, 4.7485], device='cuda:5'), covar=tensor([0.0266, 0.0498, 0.0309, 0.0315, 0.0347, 0.0293, 0.0851, 0.0364], device='cuda:5'), in_proj_covar=tensor([0.0377, 0.0418, 0.0407, 0.0376, 0.0449, 0.0426, 0.0508, 0.0342], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 04:37:54,794 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8156, 1.8898, 2.2298, 2.7082, 2.5505, 2.9862, 1.9921, 3.0225], device='cuda:5'), covar=tensor([0.0225, 0.0512, 0.0393, 0.0318, 0.0346, 0.0216, 0.0567, 0.0161], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0183, 0.0170, 0.0173, 0.0185, 0.0142, 0.0187, 0.0138], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:38:41,711 INFO [train.py:904] (5/8) Epoch 20, batch 10100, loss[loss=0.164, simple_loss=0.2517, pruned_loss=0.03816, over 16919.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2652, pruned_loss=0.03643, over 3086963.46 frames. ], batch size: 116, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:39:35,132 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 04:39:40,413 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.271e+02 2.682e+02 3.301e+02 6.522e+02, threshold=5.364e+02, percent-clipped=2.0 2023-05-01 04:40:23,125 INFO [train.py:904] (5/8) Epoch 21, batch 0, loss[loss=0.1689, simple_loss=0.2492, pruned_loss=0.04427, over 16823.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2492, pruned_loss=0.04427, over 16823.00 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 8.0 2023-05-01 04:40:23,125 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 04:40:30,892 INFO [train.py:938] (5/8) Epoch 21, validation: loss=0.1451, simple_loss=0.2491, pruned_loss=0.02058, over 944034.00 frames. 2023-05-01 04:40:30,893 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 04:41:07,241 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1528, 4.0103, 4.2032, 4.3078, 4.3863, 3.9969, 4.2565, 4.3738], device='cuda:5'), covar=tensor([0.1593, 0.1153, 0.1226, 0.0708, 0.0618, 0.1194, 0.1447, 0.0810], device='cuda:5'), in_proj_covar=tensor([0.0594, 0.0726, 0.0847, 0.0750, 0.0565, 0.0582, 0.0603, 0.0699], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:41:19,622 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:41:36,911 INFO [train.py:904] (5/8) Epoch 21, batch 50, loss[loss=0.1827, simple_loss=0.2634, pruned_loss=0.05102, over 16766.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2715, pruned_loss=0.05125, over 750033.27 frames. ], batch size: 83, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:42:04,047 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1957, 3.9720, 4.4499, 2.3754, 4.6311, 4.6995, 3.3186, 3.5892], device='cuda:5'), covar=tensor([0.0698, 0.0272, 0.0218, 0.1104, 0.0092, 0.0152, 0.0456, 0.0410], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0105, 0.0092, 0.0136, 0.0076, 0.0118, 0.0124, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 04:42:25,328 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203085.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:42:26,043 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.424e+02 2.967e+02 3.842e+02 7.470e+02, threshold=5.934e+02, percent-clipped=2.0 2023-05-01 04:42:43,742 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203099.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:42:48,529 INFO [train.py:904] (5/8) Epoch 21, batch 100, loss[loss=0.1975, simple_loss=0.2824, pruned_loss=0.05628, over 15452.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2662, pruned_loss=0.04704, over 1328380.99 frames. ], batch size: 190, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:43:00,143 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7343, 4.6236, 4.5938, 4.2491, 4.3150, 4.6643, 4.5225, 4.3881], device='cuda:5'), covar=tensor([0.0707, 0.0901, 0.0368, 0.0393, 0.0966, 0.0473, 0.0463, 0.0772], device='cuda:5'), in_proj_covar=tensor([0.0275, 0.0393, 0.0317, 0.0313, 0.0324, 0.0361, 0.0218, 0.0378], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 04:43:11,155 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 04:43:50,262 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:43:57,351 INFO [train.py:904] (5/8) Epoch 21, batch 150, loss[loss=0.1833, simple_loss=0.2595, pruned_loss=0.05357, over 16870.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2653, pruned_loss=0.0468, over 1768103.35 frames. ], batch size: 102, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:44:31,475 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7842, 2.7911, 2.5113, 2.5923, 3.1422, 2.8690, 3.3777, 3.3136], device='cuda:5'), covar=tensor([0.0127, 0.0387, 0.0457, 0.0400, 0.0235, 0.0365, 0.0252, 0.0269], device='cuda:5'), in_proj_covar=tensor([0.0198, 0.0232, 0.0222, 0.0223, 0.0231, 0.0231, 0.0229, 0.0225], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:44:44,683 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.249e+02 2.595e+02 3.084e+02 6.012e+02, threshold=5.190e+02, percent-clipped=1.0 2023-05-01 04:44:45,006 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203186.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:45:04,866 INFO [train.py:904] (5/8) Epoch 21, batch 200, loss[loss=0.158, simple_loss=0.2505, pruned_loss=0.0327, over 15950.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2649, pruned_loss=0.04586, over 2121705.31 frames. ], batch size: 35, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:45:25,332 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3502, 5.2837, 5.1752, 4.6798, 4.8785, 5.2037, 5.2635, 4.8279], device='cuda:5'), covar=tensor([0.0646, 0.0554, 0.0323, 0.0339, 0.1003, 0.0476, 0.0260, 0.0768], device='cuda:5'), in_proj_covar=tensor([0.0279, 0.0399, 0.0323, 0.0318, 0.0329, 0.0368, 0.0221, 0.0385], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:46:15,306 INFO [train.py:904] (5/8) Epoch 21, batch 250, loss[loss=0.176, simple_loss=0.2522, pruned_loss=0.04994, over 16862.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2629, pruned_loss=0.04622, over 2383646.37 frames. ], batch size: 109, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:46:29,590 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 04:47:00,833 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.163e+02 2.580e+02 3.064e+02 7.723e+02, threshold=5.160e+02, percent-clipped=4.0 2023-05-01 04:47:20,229 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2427, 3.4186, 3.7134, 2.1488, 2.9709, 2.3028, 3.6315, 3.6146], device='cuda:5'), covar=tensor([0.0303, 0.0931, 0.0567, 0.2083, 0.0919, 0.1103, 0.0701, 0.1187], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0157, 0.0162, 0.0150, 0.0141, 0.0127, 0.0141, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 04:47:23,341 INFO [train.py:904] (5/8) Epoch 21, batch 300, loss[loss=0.1824, simple_loss=0.2884, pruned_loss=0.03814, over 17253.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2599, pruned_loss=0.04444, over 2591628.76 frames. ], batch size: 52, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:48:32,612 INFO [train.py:904] (5/8) Epoch 21, batch 350, loss[loss=0.1542, simple_loss=0.2399, pruned_loss=0.03422, over 16830.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2575, pruned_loss=0.04373, over 2747348.89 frames. ], batch size: 42, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:49:17,936 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.171e+02 2.523e+02 3.052e+02 8.873e+02, threshold=5.047e+02, percent-clipped=2.0 2023-05-01 04:49:30,092 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203394.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:49:41,966 INFO [train.py:904] (5/8) Epoch 21, batch 400, loss[loss=0.1971, simple_loss=0.2722, pruned_loss=0.06097, over 16792.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.257, pruned_loss=0.04319, over 2872919.54 frames. ], batch size: 102, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:49:56,656 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7286, 4.7702, 5.1148, 5.0950, 5.1419, 4.8261, 4.7980, 4.5320], device='cuda:5'), covar=tensor([0.0346, 0.0590, 0.0405, 0.0437, 0.0492, 0.0407, 0.0880, 0.0604], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0434, 0.0422, 0.0392, 0.0467, 0.0443, 0.0527, 0.0356], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 04:50:37,075 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203441.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:50:46,536 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9347, 3.2207, 3.3228, 5.1847, 4.5003, 4.4084, 1.6779, 3.4072], device='cuda:5'), covar=tensor([0.1207, 0.0643, 0.0875, 0.0141, 0.0252, 0.0459, 0.1518, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0171, 0.0192, 0.0184, 0.0201, 0.0212, 0.0199, 0.0190], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 04:50:50,737 INFO [train.py:904] (5/8) Epoch 21, batch 450, loss[loss=0.1364, simple_loss=0.226, pruned_loss=0.02338, over 16806.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2556, pruned_loss=0.04258, over 2958094.36 frames. ], batch size: 39, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:51:11,804 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:51:29,361 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9144, 5.2775, 5.0623, 5.0367, 4.7863, 4.7224, 4.6973, 5.3766], device='cuda:5'), covar=tensor([0.1298, 0.0909, 0.0988, 0.0839, 0.0897, 0.1017, 0.1123, 0.0829], device='cuda:5'), in_proj_covar=tensor([0.0654, 0.0799, 0.0655, 0.0600, 0.0508, 0.0516, 0.0673, 0.0620], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:51:37,697 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.094e+02 2.458e+02 3.011e+02 6.519e+02, threshold=4.916e+02, percent-clipped=3.0 2023-05-01 04:51:38,131 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203486.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:51:59,115 INFO [train.py:904] (5/8) Epoch 21, batch 500, loss[loss=0.14, simple_loss=0.2253, pruned_loss=0.02736, over 17047.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2543, pruned_loss=0.04181, over 3033383.83 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:52:36,267 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4990, 3.9094, 4.1087, 2.6757, 3.6249, 4.0968, 3.7954, 2.4082], device='cuda:5'), covar=tensor([0.0595, 0.0269, 0.0060, 0.0436, 0.0129, 0.0135, 0.0098, 0.0493], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0082, 0.0081, 0.0133, 0.0098, 0.0108, 0.0092, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 04:52:36,306 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203528.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:52:43,129 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203534.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:52:52,058 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4434, 5.3527, 5.2554, 4.7325, 4.8829, 5.3012, 5.3356, 4.9338], device='cuda:5'), covar=tensor([0.0533, 0.0434, 0.0321, 0.0331, 0.1138, 0.0408, 0.0231, 0.0766], device='cuda:5'), in_proj_covar=tensor([0.0286, 0.0410, 0.0330, 0.0327, 0.0340, 0.0379, 0.0228, 0.0397], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:52:53,302 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6235, 2.5655, 2.1208, 2.3299, 2.9870, 2.6986, 3.1873, 3.1714], device='cuda:5'), covar=tensor([0.0166, 0.0449, 0.0634, 0.0560, 0.0307, 0.0459, 0.0311, 0.0288], device='cuda:5'), in_proj_covar=tensor([0.0203, 0.0235, 0.0225, 0.0226, 0.0235, 0.0234, 0.0234, 0.0229], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 04:53:08,040 INFO [train.py:904] (5/8) Epoch 21, batch 550, loss[loss=0.1612, simple_loss=0.2565, pruned_loss=0.03298, over 16732.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2543, pruned_loss=0.04167, over 3103407.10 frames. ], batch size: 62, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:53:50,175 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0254, 5.0540, 5.5186, 5.4369, 5.5004, 5.1268, 5.0444, 4.8168], device='cuda:5'), covar=tensor([0.0381, 0.0570, 0.0390, 0.0472, 0.0551, 0.0401, 0.1121, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0443, 0.0429, 0.0399, 0.0474, 0.0451, 0.0537, 0.0362], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 04:53:56,913 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.204e+02 2.652e+02 3.296e+02 1.402e+03, threshold=5.303e+02, percent-clipped=5.0 2023-05-01 04:54:07,180 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8819, 2.9942, 3.2560, 2.1469, 2.8206, 2.2213, 3.4954, 3.4251], device='cuda:5'), covar=tensor([0.0246, 0.0994, 0.0609, 0.1780, 0.0846, 0.1022, 0.0524, 0.0821], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0159, 0.0164, 0.0151, 0.0142, 0.0128, 0.0143, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 04:54:17,595 INFO [train.py:904] (5/8) Epoch 21, batch 600, loss[loss=0.1556, simple_loss=0.2436, pruned_loss=0.03376, over 17209.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2542, pruned_loss=0.04245, over 3152643.60 frames. ], batch size: 44, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:55:27,199 INFO [train.py:904] (5/8) Epoch 21, batch 650, loss[loss=0.1573, simple_loss=0.2519, pruned_loss=0.03137, over 17224.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2524, pruned_loss=0.04155, over 3193248.45 frames. ], batch size: 45, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:56:14,238 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.016e+02 2.395e+02 3.216e+02 7.414e+02, threshold=4.790e+02, percent-clipped=2.0 2023-05-01 04:56:24,856 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:56:34,648 INFO [train.py:904] (5/8) Epoch 21, batch 700, loss[loss=0.203, simple_loss=0.2745, pruned_loss=0.06574, over 16269.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2519, pruned_loss=0.04077, over 3223305.22 frames. ], batch size: 165, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:57:30,255 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203741.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:31,298 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:38,324 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203747.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:44,461 INFO [train.py:904] (5/8) Epoch 21, batch 750, loss[loss=0.1682, simple_loss=0.2638, pruned_loss=0.03625, over 16560.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.252, pruned_loss=0.04096, over 3246165.76 frames. ], batch size: 62, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:58:03,791 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203766.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:58:33,073 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.172e+02 2.611e+02 3.083e+02 3.128e+03, threshold=5.221e+02, percent-clipped=7.0 2023-05-01 04:58:36,276 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:58:51,609 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2135, 3.3695, 3.6338, 2.2510, 3.1206, 2.4696, 3.7258, 3.7284], device='cuda:5'), covar=tensor([0.0249, 0.0874, 0.0587, 0.1908, 0.0785, 0.0959, 0.0533, 0.0916], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0161, 0.0166, 0.0152, 0.0144, 0.0129, 0.0144, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 04:58:54,765 INFO [train.py:904] (5/8) Epoch 21, batch 800, loss[loss=0.189, simple_loss=0.2622, pruned_loss=0.0579, over 12580.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2515, pruned_loss=0.04075, over 3266648.50 frames. ], batch size: 247, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:59:03,712 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203808.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:24,798 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203823.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:30,060 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:58,231 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:00:04,933 INFO [train.py:904] (5/8) Epoch 21, batch 850, loss[loss=0.1644, simple_loss=0.2438, pruned_loss=0.04247, over 16490.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2506, pruned_loss=0.03972, over 3283917.05 frames. ], batch size: 75, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:00:53,320 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 1.964e+02 2.295e+02 2.674e+02 5.967e+02, threshold=4.591e+02, percent-clipped=3.0 2023-05-01 05:01:10,562 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-01 05:01:11,231 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203900.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:01:12,980 INFO [train.py:904] (5/8) Epoch 21, batch 900, loss[loss=0.1461, simple_loss=0.2393, pruned_loss=0.02644, over 17228.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2498, pruned_loss=0.03927, over 3294436.17 frames. ], batch size: 46, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:01:19,359 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3070, 2.9430, 2.6128, 2.1990, 2.2185, 2.2762, 2.9986, 2.7939], device='cuda:5'), covar=tensor([0.2635, 0.0863, 0.1785, 0.2549, 0.2531, 0.2222, 0.0604, 0.1405], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0270, 0.0304, 0.0310, 0.0294, 0.0257, 0.0294, 0.0336], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:01:21,670 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203908.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 05:01:53,445 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 05:02:21,441 INFO [train.py:904] (5/8) Epoch 21, batch 950, loss[loss=0.1772, simple_loss=0.2641, pruned_loss=0.04518, over 16566.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2501, pruned_loss=0.03981, over 3292191.27 frames. ], batch size: 68, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:02:35,132 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203961.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:03:10,187 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 2.138e+02 2.475e+02 2.994e+02 7.321e+02, threshold=4.951e+02, percent-clipped=5.0 2023-05-01 05:03:15,417 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-01 05:03:34,215 INFO [train.py:904] (5/8) Epoch 21, batch 1000, loss[loss=0.1475, simple_loss=0.2291, pruned_loss=0.03299, over 15499.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2495, pruned_loss=0.03996, over 3296107.04 frames. ], batch size: 190, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:04:10,863 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9722, 4.4025, 4.4158, 3.2809, 3.6326, 4.3081, 3.9478, 2.7592], device='cuda:5'), covar=tensor([0.0446, 0.0064, 0.0041, 0.0315, 0.0135, 0.0090, 0.0090, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0084, 0.0083, 0.0135, 0.0099, 0.0110, 0.0094, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:5') 2023-05-01 05:04:43,745 INFO [train.py:904] (5/8) Epoch 21, batch 1050, loss[loss=0.1811, simple_loss=0.2589, pruned_loss=0.05167, over 16612.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2501, pruned_loss=0.03966, over 3308051.10 frames. ], batch size: 89, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:04:46,471 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5357, 4.3336, 4.5746, 4.7252, 4.8661, 4.3866, 4.7079, 4.8318], device='cuda:5'), covar=tensor([0.1707, 0.1325, 0.1506, 0.0784, 0.0582, 0.1071, 0.2433, 0.0915], device='cuda:5'), in_proj_covar=tensor([0.0645, 0.0787, 0.0920, 0.0810, 0.0604, 0.0629, 0.0651, 0.0750], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:05:19,263 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204077.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:05:22,912 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1614, 3.9352, 4.4818, 2.1781, 4.7338, 4.8107, 3.3860, 3.6453], device='cuda:5'), covar=tensor([0.0700, 0.0239, 0.0198, 0.1194, 0.0079, 0.0126, 0.0401, 0.0394], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0108, 0.0096, 0.0139, 0.0079, 0.0122, 0.0127, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:05:34,277 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.155e+02 2.486e+02 2.944e+02 8.933e+02, threshold=4.972e+02, percent-clipped=2.0 2023-05-01 05:05:54,743 INFO [train.py:904] (5/8) Epoch 21, batch 1100, loss[loss=0.1856, simple_loss=0.2545, pruned_loss=0.05837, over 16879.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2491, pruned_loss=0.03965, over 3302073.83 frames. ], batch size: 116, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:05:56,352 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:24,054 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:25,303 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204123.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:45,667 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204138.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:07:05,260 INFO [train.py:904] (5/8) Epoch 21, batch 1150, loss[loss=0.1703, simple_loss=0.2613, pruned_loss=0.03967, over 17119.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2481, pruned_loss=0.03895, over 3311119.02 frames. ], batch size: 48, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:07:20,961 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:07:30,676 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7099, 3.6755, 2.3613, 3.9864, 2.9831, 3.9588, 2.4052, 3.0019], device='cuda:5'), covar=tensor([0.0246, 0.0428, 0.1440, 0.0315, 0.0726, 0.0645, 0.1368, 0.0623], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0177, 0.0194, 0.0162, 0.0177, 0.0214, 0.0203, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:07:31,673 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204171.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:07:53,761 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.065e+02 2.500e+02 2.935e+02 5.015e+02, threshold=5.000e+02, percent-clipped=1.0 2023-05-01 05:08:07,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4394, 5.8531, 5.6080, 5.6674, 5.2807, 5.3402, 5.2343, 5.9704], device='cuda:5'), covar=tensor([0.1467, 0.1094, 0.1046, 0.0924, 0.0914, 0.0692, 0.1251, 0.0981], device='cuda:5'), in_proj_covar=tensor([0.0678, 0.0826, 0.0679, 0.0623, 0.0526, 0.0533, 0.0696, 0.0642], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:08:13,793 INFO [train.py:904] (5/8) Epoch 21, batch 1200, loss[loss=0.1669, simple_loss=0.2507, pruned_loss=0.04156, over 16700.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2475, pruned_loss=0.03876, over 3300088.53 frames. ], batch size: 89, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:08:15,261 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204203.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 05:08:15,381 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1361, 4.8444, 5.1702, 5.3487, 5.5525, 4.8373, 5.5115, 5.5043], device='cuda:5'), covar=tensor([0.1791, 0.1280, 0.1650, 0.0790, 0.0600, 0.0839, 0.0538, 0.0584], device='cuda:5'), in_proj_covar=tensor([0.0649, 0.0794, 0.0926, 0.0815, 0.0610, 0.0634, 0.0657, 0.0757], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:08:28,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6117, 4.6901, 4.8229, 4.6689, 4.6815, 5.2489, 4.7600, 4.4674], device='cuda:5'), covar=tensor([0.1549, 0.2240, 0.2468, 0.2230, 0.2953, 0.1138, 0.1737, 0.2518], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0594, 0.0654, 0.0491, 0.0656, 0.0683, 0.0515, 0.0659], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:08:44,970 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204224.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:09:06,318 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 05:09:24,364 INFO [train.py:904] (5/8) Epoch 21, batch 1250, loss[loss=0.1859, simple_loss=0.2563, pruned_loss=0.05775, over 16695.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2475, pruned_loss=0.03923, over 3312278.20 frames. ], batch size: 134, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:09:30,709 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204256.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:10:01,404 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7941, 2.8325, 2.4690, 2.7684, 3.1215, 2.9010, 3.3979, 3.3304], device='cuda:5'), covar=tensor([0.0138, 0.0391, 0.0480, 0.0398, 0.0267, 0.0358, 0.0249, 0.0266], device='cuda:5'), in_proj_covar=tensor([0.0208, 0.0238, 0.0228, 0.0228, 0.0237, 0.0236, 0.0238, 0.0234], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:10:04,716 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8821, 2.0771, 2.4311, 2.8182, 2.6453, 3.2714, 2.2174, 3.2442], device='cuda:5'), covar=tensor([0.0268, 0.0471, 0.0354, 0.0312, 0.0357, 0.0198, 0.0477, 0.0152], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0194, 0.0180, 0.0184, 0.0196, 0.0153, 0.0196, 0.0148], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:10:12,297 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.352e+02 2.804e+02 3.486e+02 9.364e+02, threshold=5.607e+02, percent-clipped=5.0 2023-05-01 05:10:21,157 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5351, 2.5936, 1.9216, 2.3186, 2.9440, 2.6131, 3.1574, 3.1621], device='cuda:5'), covar=tensor([0.0204, 0.0502, 0.0784, 0.0577, 0.0355, 0.0476, 0.0342, 0.0315], device='cuda:5'), in_proj_covar=tensor([0.0208, 0.0237, 0.0227, 0.0227, 0.0237, 0.0236, 0.0238, 0.0233], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:10:33,939 INFO [train.py:904] (5/8) Epoch 21, batch 1300, loss[loss=0.1437, simple_loss=0.2385, pruned_loss=0.02441, over 17183.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2477, pruned_loss=0.03929, over 3313977.92 frames. ], batch size: 46, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:11:17,981 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6798, 3.7474, 4.1747, 2.2049, 3.4123, 2.7039, 4.1042, 3.8938], device='cuda:5'), covar=tensor([0.0225, 0.0848, 0.0461, 0.1966, 0.0757, 0.0890, 0.0537, 0.1107], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0153, 0.0145, 0.0130, 0.0145, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:11:42,906 INFO [train.py:904] (5/8) Epoch 21, batch 1350, loss[loss=0.1535, simple_loss=0.2551, pruned_loss=0.02598, over 17116.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2483, pruned_loss=0.0393, over 3322101.99 frames. ], batch size: 48, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:11:48,260 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3306, 4.1484, 4.3947, 4.5200, 4.6251, 4.1835, 4.3909, 4.6006], device='cuda:5'), covar=tensor([0.1644, 0.1094, 0.1260, 0.0692, 0.0595, 0.1218, 0.2444, 0.0723], device='cuda:5'), in_proj_covar=tensor([0.0650, 0.0795, 0.0928, 0.0818, 0.0611, 0.0635, 0.0657, 0.0759], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:12:18,629 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 05:12:31,608 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.188e+02 2.581e+02 2.996e+02 9.160e+02, threshold=5.162e+02, percent-clipped=1.0 2023-05-01 05:12:52,518 INFO [train.py:904] (5/8) Epoch 21, batch 1400, loss[loss=0.1355, simple_loss=0.2218, pruned_loss=0.02461, over 16863.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2484, pruned_loss=0.03933, over 3321331.52 frames. ], batch size: 42, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:12:54,607 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204403.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:20,736 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204422.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:35,311 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204433.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:14:00,519 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204451.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:01,416 INFO [train.py:904] (5/8) Epoch 21, batch 1450, loss[loss=0.1493, simple_loss=0.2405, pruned_loss=0.02902, over 17239.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2472, pruned_loss=0.03883, over 3322855.34 frames. ], batch size: 43, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:14:26,856 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204470.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:28,980 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0165, 2.9455, 2.6652, 4.4466, 3.7061, 4.2116, 1.8142, 3.1222], device='cuda:5'), covar=tensor([0.1229, 0.0673, 0.1145, 0.0206, 0.0226, 0.0450, 0.1470, 0.0768], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0175, 0.0195, 0.0189, 0.0205, 0.0216, 0.0202, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:14:50,907 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.087e+02 2.507e+02 3.018e+02 6.138e+02, threshold=5.015e+02, percent-clipped=2.0 2023-05-01 05:14:57,728 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5087, 5.9244, 5.6629, 5.7371, 5.3015, 5.4449, 5.3395, 6.0331], device='cuda:5'), covar=tensor([0.1505, 0.1002, 0.1016, 0.0951, 0.0968, 0.0664, 0.1142, 0.0962], device='cuda:5'), in_proj_covar=tensor([0.0681, 0.0831, 0.0680, 0.0627, 0.0529, 0.0535, 0.0699, 0.0644], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:15:10,560 INFO [train.py:904] (5/8) Epoch 21, batch 1500, loss[loss=0.1614, simple_loss=0.2548, pruned_loss=0.03407, over 16773.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.247, pruned_loss=0.03917, over 3314339.56 frames. ], batch size: 57, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:15:12,785 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:15:33,853 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204519.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:16:17,197 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:16:18,103 INFO [train.py:904] (5/8) Epoch 21, batch 1550, loss[loss=0.1612, simple_loss=0.259, pruned_loss=0.03175, over 17046.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2494, pruned_loss=0.04032, over 3305912.52 frames. ], batch size: 50, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:16:23,642 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204556.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:16:26,477 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-01 05:17:07,007 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.391e+02 2.852e+02 3.354e+02 7.624e+02, threshold=5.704e+02, percent-clipped=5.0 2023-05-01 05:17:08,278 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 05:17:26,290 INFO [train.py:904] (5/8) Epoch 21, batch 1600, loss[loss=0.1729, simple_loss=0.2693, pruned_loss=0.03826, over 17027.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2511, pruned_loss=0.04076, over 3312280.82 frames. ], batch size: 50, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:17:29,640 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204604.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:18:06,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1527, 5.7981, 5.8748, 5.5441, 5.7429, 6.2525, 5.7510, 5.3976], device='cuda:5'), covar=tensor([0.0859, 0.1947, 0.2348, 0.1969, 0.2515, 0.0920, 0.1501, 0.2366], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0591, 0.0651, 0.0487, 0.0651, 0.0680, 0.0513, 0.0654], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:18:24,759 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8640, 1.3644, 1.7364, 1.7142, 1.7377, 1.9802, 1.6299, 1.8348], device='cuda:5'), covar=tensor([0.0249, 0.0420, 0.0247, 0.0296, 0.0325, 0.0193, 0.0425, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0193, 0.0180, 0.0184, 0.0196, 0.0154, 0.0196, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:18:35,827 INFO [train.py:904] (5/8) Epoch 21, batch 1650, loss[loss=0.2284, simple_loss=0.3102, pruned_loss=0.07331, over 12211.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2525, pruned_loss=0.04131, over 3305201.97 frames. ], batch size: 246, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:19:25,766 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.742e+02 2.251e+02 2.625e+02 3.143e+02 5.667e+02, threshold=5.251e+02, percent-clipped=0.0 2023-05-01 05:19:38,383 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-05-01 05:19:45,542 INFO [train.py:904] (5/8) Epoch 21, batch 1700, loss[loss=0.1623, simple_loss=0.2463, pruned_loss=0.0392, over 16801.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2539, pruned_loss=0.042, over 3310814.30 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:20:29,946 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204733.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:20:46,408 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3685, 3.7030, 3.8310, 2.2452, 3.1085, 2.4305, 3.7830, 3.8379], device='cuda:5'), covar=tensor([0.0279, 0.0917, 0.0545, 0.1957, 0.0861, 0.1060, 0.0623, 0.1093], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0144, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:20:55,681 INFO [train.py:904] (5/8) Epoch 21, batch 1750, loss[loss=0.1609, simple_loss=0.2495, pruned_loss=0.03611, over 17197.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2554, pruned_loss=0.04233, over 3305526.11 frames. ], batch size: 44, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:21:37,075 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204781.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:21:46,300 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.363e+02 2.773e+02 3.345e+02 7.836e+02, threshold=5.546e+02, percent-clipped=6.0 2023-05-01 05:22:06,997 INFO [train.py:904] (5/8) Epoch 21, batch 1800, loss[loss=0.1811, simple_loss=0.2609, pruned_loss=0.0507, over 16774.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2556, pruned_loss=0.04138, over 3319168.22 frames. ], batch size: 124, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:22:30,012 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:22:32,930 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204821.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:22:34,678 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1661, 4.2498, 4.3751, 4.1550, 4.2499, 4.8038, 4.3089, 3.9670], device='cuda:5'), covar=tensor([0.2036, 0.2401, 0.2350, 0.2689, 0.3126, 0.1285, 0.1810, 0.3006], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0599, 0.0660, 0.0495, 0.0661, 0.0689, 0.0521, 0.0664], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:23:03,736 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6613, 4.9620, 4.7547, 4.7659, 4.5289, 4.4913, 4.4369, 5.0101], device='cuda:5'), covar=tensor([0.1178, 0.0823, 0.1012, 0.0841, 0.0803, 0.1159, 0.1116, 0.0909], device='cuda:5'), in_proj_covar=tensor([0.0678, 0.0826, 0.0679, 0.0624, 0.0526, 0.0531, 0.0697, 0.0640], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:23:15,229 INFO [train.py:904] (5/8) Epoch 21, batch 1850, loss[loss=0.1593, simple_loss=0.2577, pruned_loss=0.03049, over 17267.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2569, pruned_loss=0.04154, over 3321789.84 frames. ], batch size: 52, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:23:37,466 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204867.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:23:57,987 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:24:06,223 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 2.082e+02 2.497e+02 3.026e+02 6.100e+02, threshold=4.995e+02, percent-clipped=2.0 2023-05-01 05:24:26,182 INFO [train.py:904] (5/8) Epoch 21, batch 1900, loss[loss=0.1461, simple_loss=0.2333, pruned_loss=0.02943, over 16813.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2561, pruned_loss=0.04067, over 3318304.76 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:25:35,965 INFO [train.py:904] (5/8) Epoch 21, batch 1950, loss[loss=0.1754, simple_loss=0.2549, pruned_loss=0.04801, over 16774.00 frames. ], tot_loss[loss=0.168, simple_loss=0.256, pruned_loss=0.04007, over 3325341.07 frames. ], batch size: 83, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:26:26,416 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.195e+02 2.614e+02 3.172e+02 4.787e+02, threshold=5.228e+02, percent-clipped=0.0 2023-05-01 05:26:36,797 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7008, 3.7296, 2.2857, 4.0183, 2.9227, 3.9602, 2.4980, 2.9876], device='cuda:5'), covar=tensor([0.0273, 0.0393, 0.1497, 0.0373, 0.0742, 0.0786, 0.1361, 0.0726], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0179, 0.0196, 0.0166, 0.0178, 0.0219, 0.0205, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:26:44,687 INFO [train.py:904] (5/8) Epoch 21, batch 2000, loss[loss=0.1573, simple_loss=0.2446, pruned_loss=0.03502, over 17233.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2552, pruned_loss=0.04001, over 3321789.65 frames. ], batch size: 45, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:27:55,533 INFO [train.py:904] (5/8) Epoch 21, batch 2050, loss[loss=0.1875, simple_loss=0.27, pruned_loss=0.05255, over 15490.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2554, pruned_loss=0.04065, over 3313531.98 frames. ], batch size: 190, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:28:18,332 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4898, 4.5056, 4.3780, 4.1326, 3.8300, 4.6436, 4.4573, 4.2579], device='cuda:5'), covar=tensor([0.1035, 0.1286, 0.0516, 0.0482, 0.1587, 0.0707, 0.0568, 0.0800], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0436, 0.0353, 0.0349, 0.0361, 0.0404, 0.0240, 0.0424], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:28:44,316 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.164e+02 2.478e+02 3.098e+02 5.896e+02, threshold=4.956e+02, percent-clipped=1.0 2023-05-01 05:28:50,315 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 05:29:04,147 INFO [train.py:904] (5/8) Epoch 21, batch 2100, loss[loss=0.1802, simple_loss=0.2638, pruned_loss=0.04833, over 11981.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2569, pruned_loss=0.04189, over 3307155.57 frames. ], batch size: 248, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:29:11,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9904, 2.1127, 2.6957, 2.9608, 2.7502, 3.3327, 2.1515, 3.3632], device='cuda:5'), covar=tensor([0.0290, 0.0540, 0.0342, 0.0339, 0.0385, 0.0223, 0.0566, 0.0202], device='cuda:5'), in_proj_covar=tensor([0.0189, 0.0195, 0.0180, 0.0186, 0.0198, 0.0155, 0.0197, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:29:14,594 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205109.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:30:14,931 INFO [train.py:904] (5/8) Epoch 21, batch 2150, loss[loss=0.1538, simple_loss=0.2378, pruned_loss=0.03493, over 16820.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2583, pruned_loss=0.04311, over 3291107.28 frames. ], batch size: 96, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:30:39,956 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205170.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 05:30:50,428 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205177.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:31:04,862 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.289e+02 2.743e+02 3.268e+02 8.955e+02, threshold=5.486e+02, percent-clipped=2.0 2023-05-01 05:31:19,696 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205198.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:31:24,883 INFO [train.py:904] (5/8) Epoch 21, batch 2200, loss[loss=0.1478, simple_loss=0.2372, pruned_loss=0.02917, over 17215.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2587, pruned_loss=0.04302, over 3299074.05 frames. ], batch size: 44, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:31:44,674 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 05:32:34,214 INFO [train.py:904] (5/8) Epoch 21, batch 2250, loss[loss=0.1753, simple_loss=0.2656, pruned_loss=0.04249, over 16749.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2591, pruned_loss=0.04344, over 3296044.79 frames. ], batch size: 57, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:32:45,064 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:32:49,085 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3760, 4.4126, 4.5783, 4.3487, 4.4131, 4.9979, 4.5271, 4.2327], device='cuda:5'), covar=tensor([0.1752, 0.2287, 0.2343, 0.2521, 0.2949, 0.1230, 0.1755, 0.2890], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0601, 0.0662, 0.0499, 0.0663, 0.0692, 0.0521, 0.0668], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:33:19,323 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9353, 2.1706, 2.5612, 2.9297, 2.7177, 3.4351, 2.3924, 3.2832], device='cuda:5'), covar=tensor([0.0242, 0.0463, 0.0331, 0.0310, 0.0344, 0.0184, 0.0461, 0.0175], device='cuda:5'), in_proj_covar=tensor([0.0189, 0.0194, 0.0180, 0.0186, 0.0198, 0.0155, 0.0197, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:33:23,479 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.149e+02 2.507e+02 2.983e+02 5.754e+02, threshold=5.014e+02, percent-clipped=1.0 2023-05-01 05:33:44,305 INFO [train.py:904] (5/8) Epoch 21, batch 2300, loss[loss=0.1778, simple_loss=0.2678, pruned_loss=0.04388, over 16770.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2584, pruned_loss=0.04298, over 3300401.58 frames. ], batch size: 57, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:34:53,181 INFO [train.py:904] (5/8) Epoch 21, batch 2350, loss[loss=0.2014, simple_loss=0.2999, pruned_loss=0.05138, over 16760.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2581, pruned_loss=0.04322, over 3306671.50 frames. ], batch size: 62, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:35:18,398 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4430, 3.7244, 4.0063, 2.2861, 3.1623, 2.4770, 3.8701, 3.8687], device='cuda:5'), covar=tensor([0.0261, 0.0876, 0.0487, 0.1892, 0.0812, 0.0967, 0.0597, 0.1015], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0164, 0.0167, 0.0153, 0.0144, 0.0129, 0.0144, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:35:42,804 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.163e+02 2.457e+02 2.980e+02 4.846e+02, threshold=4.914e+02, percent-clipped=0.0 2023-05-01 05:35:55,241 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 05:36:02,959 INFO [train.py:904] (5/8) Epoch 21, batch 2400, loss[loss=0.1778, simple_loss=0.2637, pruned_loss=0.04589, over 16418.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2577, pruned_loss=0.04216, over 3308043.79 frames. ], batch size: 75, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:36:10,934 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 05:36:51,124 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 05:37:10,850 INFO [train.py:904] (5/8) Epoch 21, batch 2450, loss[loss=0.167, simple_loss=0.2637, pruned_loss=0.03509, over 17045.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2581, pruned_loss=0.04167, over 3322086.26 frames. ], batch size: 53, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:37:25,007 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4413, 3.6691, 3.8607, 2.7445, 3.5094, 3.9569, 3.5949, 2.2073], device='cuda:5'), covar=tensor([0.0502, 0.0192, 0.0053, 0.0355, 0.0110, 0.0091, 0.0098, 0.0456], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0099, 0.0109, 0.0094, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:37:29,573 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205465.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:37:45,991 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205477.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:38:00,847 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.346e+02 2.648e+02 3.284e+02 5.369e+02, threshold=5.296e+02, percent-clipped=1.0 2023-05-01 05:38:16,909 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8467, 2.0936, 2.5868, 2.8706, 2.7307, 3.2571, 2.3747, 3.1987], device='cuda:5'), covar=tensor([0.0235, 0.0485, 0.0298, 0.0278, 0.0308, 0.0185, 0.0453, 0.0156], device='cuda:5'), in_proj_covar=tensor([0.0189, 0.0195, 0.0180, 0.0186, 0.0198, 0.0155, 0.0198, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:38:21,266 INFO [train.py:904] (5/8) Epoch 21, batch 2500, loss[loss=0.1861, simple_loss=0.2577, pruned_loss=0.05727, over 16905.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2586, pruned_loss=0.04147, over 3321395.74 frames. ], batch size: 116, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:38:40,343 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1591, 4.7190, 4.6341, 3.4401, 3.9748, 4.6154, 4.0007, 3.0042], device='cuda:5'), covar=tensor([0.0436, 0.0067, 0.0042, 0.0339, 0.0126, 0.0082, 0.0089, 0.0399], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0099, 0.0109, 0.0094, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:38:53,421 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:39:30,352 INFO [train.py:904] (5/8) Epoch 21, batch 2550, loss[loss=0.1772, simple_loss=0.2707, pruned_loss=0.04187, over 16472.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2591, pruned_loss=0.04168, over 3325028.28 frames. ], batch size: 68, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:39:33,576 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:40:17,538 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4755, 4.3069, 4.5070, 4.6523, 4.7997, 4.3405, 4.6298, 4.7476], device='cuda:5'), covar=tensor([0.1740, 0.1484, 0.1501, 0.1035, 0.0665, 0.1138, 0.2105, 0.1215], device='cuda:5'), in_proj_covar=tensor([0.0660, 0.0810, 0.0949, 0.0832, 0.0619, 0.0649, 0.0670, 0.0776], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:40:19,335 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 2.328e+02 2.679e+02 3.297e+02 1.189e+03, threshold=5.358e+02, percent-clipped=3.0 2023-05-01 05:40:21,048 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8271, 2.0502, 2.3385, 3.1326, 2.1173, 2.2091, 2.2422, 2.1085], device='cuda:5'), covar=tensor([0.1533, 0.3567, 0.2619, 0.0770, 0.4171, 0.2722, 0.3484, 0.3855], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0448, 0.0370, 0.0330, 0.0436, 0.0516, 0.0417, 0.0525], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:40:38,687 INFO [train.py:904] (5/8) Epoch 21, batch 2600, loss[loss=0.1983, simple_loss=0.2759, pruned_loss=0.06037, over 16921.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2598, pruned_loss=0.04224, over 3312721.57 frames. ], batch size: 109, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:41:49,817 INFO [train.py:904] (5/8) Epoch 21, batch 2650, loss[loss=0.1875, simple_loss=0.2761, pruned_loss=0.04941, over 16794.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2597, pruned_loss=0.04164, over 3317798.57 frames. ], batch size: 102, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:41:54,585 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205655.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:42:40,193 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.048e+02 2.349e+02 2.910e+02 5.870e+02, threshold=4.699e+02, percent-clipped=1.0 2023-05-01 05:43:00,012 INFO [train.py:904] (5/8) Epoch 21, batch 2700, loss[loss=0.1745, simple_loss=0.2653, pruned_loss=0.04182, over 16689.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2605, pruned_loss=0.04142, over 3319257.83 frames. ], batch size: 57, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:43:18,690 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:43:59,615 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205744.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:44:09,884 INFO [train.py:904] (5/8) Epoch 21, batch 2750, loss[loss=0.1431, simple_loss=0.2325, pruned_loss=0.02687, over 17196.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.26, pruned_loss=0.04094, over 3316466.97 frames. ], batch size: 43, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:44:13,724 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6522, 3.6475, 2.9528, 2.2341, 2.4247, 2.3686, 3.8148, 3.2576], device='cuda:5'), covar=tensor([0.2638, 0.0637, 0.1633, 0.2870, 0.2617, 0.2023, 0.0543, 0.1377], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0270, 0.0304, 0.0309, 0.0297, 0.0257, 0.0294, 0.0337], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:44:29,198 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205765.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:02,029 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.167e+02 2.431e+02 2.778e+02 3.742e+02, threshold=4.863e+02, percent-clipped=0.0 2023-05-01 05:45:19,232 INFO [train.py:904] (5/8) Epoch 21, batch 2800, loss[loss=0.1722, simple_loss=0.2562, pruned_loss=0.0441, over 16371.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2601, pruned_loss=0.04091, over 3320009.20 frames. ], batch size: 146, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:45:22,648 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1846, 5.1490, 4.9365, 4.3647, 5.0326, 1.8607, 4.8009, 4.8513], device='cuda:5'), covar=tensor([0.0104, 0.0080, 0.0235, 0.0415, 0.0119, 0.2888, 0.0161, 0.0221], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0157, 0.0199, 0.0180, 0.0178, 0.0210, 0.0190, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:45:24,447 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205805.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:35,318 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205813.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:46:29,085 INFO [train.py:904] (5/8) Epoch 21, batch 2850, loss[loss=0.1955, simple_loss=0.2771, pruned_loss=0.05695, over 11624.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2585, pruned_loss=0.04059, over 3320505.46 frames. ], batch size: 248, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:46:31,700 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:17,822 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 2.169e+02 2.532e+02 3.008e+02 5.585e+02, threshold=5.064e+02, percent-clipped=2.0 2023-05-01 05:47:19,336 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:36,359 INFO [train.py:904] (5/8) Epoch 21, batch 2900, loss[loss=0.1564, simple_loss=0.2483, pruned_loss=0.03224, over 17186.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2578, pruned_loss=0.04083, over 3326989.75 frames. ], batch size: 46, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:47:36,636 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205902.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:36,930 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4021, 2.3917, 2.5104, 4.3057, 2.3955, 2.7240, 2.4652, 2.5486], device='cuda:5'), covar=tensor([0.1416, 0.3670, 0.2758, 0.0552, 0.3910, 0.2601, 0.3746, 0.3196], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0449, 0.0369, 0.0331, 0.0436, 0.0516, 0.0418, 0.0526], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:48:44,777 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:48:45,587 INFO [train.py:904] (5/8) Epoch 21, batch 2950, loss[loss=0.1924, simple_loss=0.2693, pruned_loss=0.05771, over 16476.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2574, pruned_loss=0.04135, over 3328364.55 frames. ], batch size: 146, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:03,918 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9391, 2.1928, 2.5630, 2.9242, 2.7605, 3.4270, 2.3141, 3.3800], device='cuda:5'), covar=tensor([0.0262, 0.0488, 0.0341, 0.0291, 0.0348, 0.0199, 0.0488, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0191, 0.0196, 0.0181, 0.0187, 0.0198, 0.0155, 0.0198, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:49:36,951 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.251e+02 2.659e+02 3.342e+02 9.466e+02, threshold=5.318e+02, percent-clipped=4.0 2023-05-01 05:49:55,046 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8200, 3.9248, 2.7692, 4.5864, 3.0074, 4.5102, 2.6573, 3.2955], device='cuda:5'), covar=tensor([0.0318, 0.0427, 0.1349, 0.0253, 0.0862, 0.0485, 0.1394, 0.0704], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0178, 0.0194, 0.0165, 0.0177, 0.0219, 0.0202, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:49:58,059 INFO [train.py:904] (5/8) Epoch 21, batch 3000, loss[loss=0.1552, simple_loss=0.2385, pruned_loss=0.03595, over 16746.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2571, pruned_loss=0.042, over 3329157.46 frames. ], batch size: 39, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:58,059 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 05:50:06,481 INFO [train.py:938] (5/8) Epoch 21, validation: loss=0.1356, simple_loss=0.2408, pruned_loss=0.01521, over 944034.00 frames. 2023-05-01 05:50:06,483 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 05:50:18,445 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:51:03,799 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4665, 4.3143, 4.5118, 4.6864, 4.8102, 4.3021, 4.6258, 4.7843], device='cuda:5'), covar=tensor([0.1748, 0.1328, 0.1450, 0.0723, 0.0633, 0.1268, 0.2042, 0.0922], device='cuda:5'), in_proj_covar=tensor([0.0668, 0.0820, 0.0960, 0.0841, 0.0624, 0.0660, 0.0674, 0.0782], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 05:51:14,533 INFO [train.py:904] (5/8) Epoch 21, batch 3050, loss[loss=0.1658, simple_loss=0.2482, pruned_loss=0.04166, over 16731.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2564, pruned_loss=0.04141, over 3335967.09 frames. ], batch size: 83, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:51:21,151 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206056.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:52:05,524 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.000e+02 2.421e+02 3.043e+02 5.222e+02, threshold=4.843e+02, percent-clipped=0.0 2023-05-01 05:52:22,012 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206100.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:52:24,079 INFO [train.py:904] (5/8) Epoch 21, batch 3100, loss[loss=0.147, simple_loss=0.2402, pruned_loss=0.02694, over 17146.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2561, pruned_loss=0.04126, over 3338575.63 frames. ], batch size: 46, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:52:43,735 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206117.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:53:31,001 INFO [train.py:904] (5/8) Epoch 21, batch 3150, loss[loss=0.1749, simple_loss=0.2681, pruned_loss=0.04087, over 17095.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2559, pruned_loss=0.04171, over 3333752.64 frames. ], batch size: 53, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:53:51,750 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2658, 5.2271, 5.1069, 4.6191, 4.7487, 5.1782, 5.1596, 4.7488], device='cuda:5'), covar=tensor([0.0702, 0.0519, 0.0322, 0.0360, 0.1132, 0.0438, 0.0329, 0.0876], device='cuda:5'), in_proj_covar=tensor([0.0308, 0.0440, 0.0358, 0.0352, 0.0367, 0.0409, 0.0245, 0.0430], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 05:54:22,973 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.110e+02 2.360e+02 2.759e+02 5.694e+02, threshold=4.721e+02, percent-clipped=1.0 2023-05-01 05:54:35,067 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206197.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:54:41,552 INFO [train.py:904] (5/8) Epoch 21, batch 3200, loss[loss=0.1426, simple_loss=0.2298, pruned_loss=0.02773, over 17229.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2551, pruned_loss=0.04118, over 3335563.37 frames. ], batch size: 44, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:55:37,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0600, 2.4549, 2.6238, 1.8965, 2.7669, 2.7519, 2.3625, 2.3503], device='cuda:5'), covar=tensor([0.0753, 0.0256, 0.0278, 0.0994, 0.0128, 0.0307, 0.0491, 0.0458], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0109, 0.0098, 0.0139, 0.0080, 0.0126, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 05:55:43,236 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:55:51,282 INFO [train.py:904] (5/8) Epoch 21, batch 3250, loss[loss=0.1974, simple_loss=0.2769, pruned_loss=0.05889, over 16316.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2552, pruned_loss=0.04147, over 3325542.11 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:55:59,431 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:56:33,548 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-05-01 05:56:42,636 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.332e+02 2.138e+02 2.458e+02 2.946e+02 5.794e+02, threshold=4.917e+02, percent-clipped=2.0 2023-05-01 05:57:00,591 INFO [train.py:904] (5/8) Epoch 21, batch 3300, loss[loss=0.1685, simple_loss=0.2632, pruned_loss=0.0369, over 17129.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2564, pruned_loss=0.04204, over 3320037.27 frames. ], batch size: 48, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:57:13,828 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206311.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:57:34,087 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 05:58:09,886 INFO [train.py:904] (5/8) Epoch 21, batch 3350, loss[loss=0.1681, simple_loss=0.2654, pruned_loss=0.03542, over 16598.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.257, pruned_loss=0.04186, over 3319446.17 frames. ], batch size: 62, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:58:20,174 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:00,107 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.208e+02 2.654e+02 3.136e+02 4.728e+02, threshold=5.309e+02, percent-clipped=0.0 2023-05-01 05:59:16,484 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206400.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:18,376 INFO [train.py:904] (5/8) Epoch 21, batch 3400, loss[loss=0.1688, simple_loss=0.2567, pruned_loss=0.04041, over 15567.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2569, pruned_loss=0.04158, over 3317579.83 frames. ], batch size: 191, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:59:32,345 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:43,105 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3240, 4.3224, 4.2571, 3.9439, 3.9700, 4.3503, 4.0793, 4.1003], device='cuda:5'), covar=tensor([0.0625, 0.0621, 0.0315, 0.0292, 0.0823, 0.0526, 0.0612, 0.0621], device='cuda:5'), in_proj_covar=tensor([0.0307, 0.0441, 0.0358, 0.0352, 0.0365, 0.0409, 0.0245, 0.0430], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 06:00:21,655 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:00:26,892 INFO [train.py:904] (5/8) Epoch 21, batch 3450, loss[loss=0.1557, simple_loss=0.2525, pruned_loss=0.02944, over 17249.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2556, pruned_loss=0.04139, over 3312602.33 frames. ], batch size: 52, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:01:17,176 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.080e+02 2.437e+02 2.944e+02 7.361e+02, threshold=4.875e+02, percent-clipped=3.0 2023-05-01 06:01:30,552 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6806, 3.7977, 2.3753, 4.2972, 2.9455, 4.2594, 2.5537, 3.1180], device='cuda:5'), covar=tensor([0.0308, 0.0387, 0.1559, 0.0324, 0.0788, 0.0539, 0.1397, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0178, 0.0195, 0.0166, 0.0177, 0.0220, 0.0203, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:01:33,784 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 06:01:36,852 INFO [train.py:904] (5/8) Epoch 21, batch 3500, loss[loss=0.1853, simple_loss=0.2659, pruned_loss=0.05228, over 16926.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2539, pruned_loss=0.04077, over 3312125.32 frames. ], batch size: 116, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:02:22,482 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 06:02:37,901 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:02:44,996 INFO [train.py:904] (5/8) Epoch 21, batch 3550, loss[loss=0.1731, simple_loss=0.2599, pruned_loss=0.04314, over 16339.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2529, pruned_loss=0.04, over 3310246.27 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:02:47,118 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:20,292 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 06:03:37,064 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.127e+02 2.401e+02 2.794e+02 4.680e+02, threshold=4.802e+02, percent-clipped=0.0 2023-05-01 06:03:44,322 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:55,937 INFO [train.py:904] (5/8) Epoch 21, batch 3600, loss[loss=0.1523, simple_loss=0.2316, pruned_loss=0.03646, over 16570.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2521, pruned_loss=0.03986, over 3308726.24 frames. ], batch size: 75, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:04:04,843 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7673, 4.5739, 4.8430, 5.0241, 5.1847, 4.6102, 5.1528, 5.1956], device='cuda:5'), covar=tensor([0.1862, 0.1510, 0.1871, 0.0875, 0.0678, 0.1116, 0.0724, 0.0678], device='cuda:5'), in_proj_covar=tensor([0.0667, 0.0818, 0.0960, 0.0845, 0.0627, 0.0657, 0.0676, 0.0782], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:04:19,399 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 06:05:06,987 INFO [train.py:904] (5/8) Epoch 21, batch 3650, loss[loss=0.1658, simple_loss=0.2371, pruned_loss=0.04721, over 16810.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2506, pruned_loss=0.04059, over 3306850.31 frames. ], batch size: 124, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:05:29,039 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0548, 3.0138, 2.9037, 5.1631, 4.2913, 4.5371, 1.8348, 3.4812], device='cuda:5'), covar=tensor([0.1272, 0.0795, 0.1112, 0.0203, 0.0258, 0.0365, 0.1582, 0.0720], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0193, 0.0208, 0.0217, 0.0202, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:05:40,754 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 06:05:59,786 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.224e+02 2.566e+02 3.274e+02 6.534e+02, threshold=5.132e+02, percent-clipped=5.0 2023-05-01 06:06:18,967 INFO [train.py:904] (5/8) Epoch 21, batch 3700, loss[loss=0.1793, simple_loss=0.2585, pruned_loss=0.05006, over 16830.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2496, pruned_loss=0.04221, over 3297312.26 frames. ], batch size: 116, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:06:34,615 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:07:32,573 INFO [train.py:904] (5/8) Epoch 21, batch 3750, loss[loss=0.1719, simple_loss=0.254, pruned_loss=0.04486, over 16520.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.25, pruned_loss=0.04346, over 3292331.62 frames. ], batch size: 75, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:07:45,682 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206760.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:08:26,978 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.267e+02 2.726e+02 3.357e+02 9.074e+02, threshold=5.452e+02, percent-clipped=3.0 2023-05-01 06:08:45,460 INFO [train.py:904] (5/8) Epoch 21, batch 3800, loss[loss=0.1854, simple_loss=0.264, pruned_loss=0.05343, over 16472.00 frames. ], tot_loss[loss=0.171, simple_loss=0.252, pruned_loss=0.04502, over 3282691.74 frames. ], batch size: 75, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:09:32,842 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206834.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:09:58,316 INFO [train.py:904] (5/8) Epoch 21, batch 3850, loss[loss=0.1591, simple_loss=0.2404, pruned_loss=0.03886, over 16243.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2528, pruned_loss=0.04599, over 3274628.99 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:10:00,714 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:10:26,536 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 06:10:52,987 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.673e+02 2.267e+02 2.812e+02 3.321e+02 1.032e+03, threshold=5.623e+02, percent-clipped=3.0 2023-05-01 06:11:00,926 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206895.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:11:09,938 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206901.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:11:10,723 INFO [train.py:904] (5/8) Epoch 21, batch 3900, loss[loss=0.1579, simple_loss=0.2385, pruned_loss=0.03867, over 16714.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2529, pruned_loss=0.04647, over 3272533.45 frames. ], batch size: 89, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:11:56,618 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7959, 3.9018, 4.1264, 4.0995, 4.1347, 3.8896, 3.9256, 3.8936], device='cuda:5'), covar=tensor([0.0403, 0.0637, 0.0460, 0.0462, 0.0528, 0.0482, 0.0752, 0.0567], device='cuda:5'), in_proj_covar=tensor([0.0420, 0.0467, 0.0451, 0.0419, 0.0498, 0.0474, 0.0564, 0.0380], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 06:12:24,778 INFO [train.py:904] (5/8) Epoch 21, batch 3950, loss[loss=0.1843, simple_loss=0.2514, pruned_loss=0.05859, over 16919.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2525, pruned_loss=0.04701, over 3276134.50 frames. ], batch size: 116, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:13:00,850 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9938, 4.9339, 4.8760, 4.2038, 4.9800, 1.9730, 4.7168, 4.5872], device='cuda:5'), covar=tensor([0.0118, 0.0130, 0.0184, 0.0392, 0.0096, 0.2735, 0.0154, 0.0244], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0158, 0.0201, 0.0182, 0.0180, 0.0210, 0.0191, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:13:16,333 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.243e+02 2.750e+02 3.370e+02 5.125e+02, threshold=5.500e+02, percent-clipped=0.0 2023-05-01 06:13:35,020 INFO [train.py:904] (5/8) Epoch 21, batch 4000, loss[loss=0.1776, simple_loss=0.2582, pruned_loss=0.04852, over 16245.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2531, pruned_loss=0.04751, over 3275549.30 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:14:45,538 INFO [train.py:904] (5/8) Epoch 21, batch 4050, loss[loss=0.1916, simple_loss=0.2714, pruned_loss=0.05588, over 12132.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2542, pruned_loss=0.04673, over 3274608.07 frames. ], batch size: 247, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:15:15,914 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0796, 2.5652, 1.9925, 2.3021, 2.8706, 2.4482, 2.8408, 3.0151], device='cuda:5'), covar=tensor([0.0132, 0.0375, 0.0545, 0.0432, 0.0227, 0.0363, 0.0187, 0.0232], device='cuda:5'), in_proj_covar=tensor([0.0211, 0.0237, 0.0226, 0.0226, 0.0237, 0.0235, 0.0240, 0.0234], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:15:34,158 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 06:15:36,910 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1122, 2.5339, 1.9615, 2.3400, 2.8392, 2.5071, 2.8499, 2.9956], device='cuda:5'), covar=tensor([0.0127, 0.0382, 0.0578, 0.0428, 0.0252, 0.0345, 0.0190, 0.0227], device='cuda:5'), in_proj_covar=tensor([0.0211, 0.0237, 0.0226, 0.0227, 0.0237, 0.0236, 0.0241, 0.0235], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:15:37,542 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.811e+02 2.022e+02 2.362e+02 4.277e+02, threshold=4.045e+02, percent-clipped=0.0 2023-05-01 06:15:56,002 INFO [train.py:904] (5/8) Epoch 21, batch 4100, loss[loss=0.1909, simple_loss=0.2798, pruned_loss=0.05101, over 16641.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2554, pruned_loss=0.04604, over 3274097.25 frames. ], batch size: 62, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:17:10,037 INFO [train.py:904] (5/8) Epoch 21, batch 4150, loss[loss=0.2525, simple_loss=0.3213, pruned_loss=0.0919, over 11574.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2623, pruned_loss=0.04895, over 3208448.97 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:17:45,653 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:18:04,519 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.144e+02 2.646e+02 3.287e+02 6.191e+02, threshold=5.292e+02, percent-clipped=8.0 2023-05-01 06:18:06,776 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207190.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:18:24,671 INFO [train.py:904] (5/8) Epoch 21, batch 4200, loss[loss=0.2206, simple_loss=0.3111, pruned_loss=0.06506, over 16660.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2692, pruned_loss=0.05042, over 3191676.57 frames. ], batch size: 57, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:19:18,506 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:19:40,066 INFO [train.py:904] (5/8) Epoch 21, batch 4250, loss[loss=0.197, simple_loss=0.2734, pruned_loss=0.06034, over 12027.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2723, pruned_loss=0.04988, over 3191175.30 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:20:35,825 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.204e+02 2.470e+02 3.047e+02 4.735e+02, threshold=4.941e+02, percent-clipped=0.0 2023-05-01 06:20:45,553 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4055, 2.4652, 2.2962, 4.1785, 2.3438, 2.8361, 2.4883, 2.5679], device='cuda:5'), covar=tensor([0.1238, 0.3398, 0.2799, 0.0481, 0.3676, 0.2279, 0.3190, 0.3187], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0451, 0.0369, 0.0332, 0.0437, 0.0520, 0.0419, 0.0527], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:20:55,795 INFO [train.py:904] (5/8) Epoch 21, batch 4300, loss[loss=0.1817, simple_loss=0.2792, pruned_loss=0.04211, over 16738.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2737, pruned_loss=0.0488, over 3198184.42 frames. ], batch size: 89, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:22:07,522 INFO [train.py:904] (5/8) Epoch 21, batch 4350, loss[loss=0.2006, simple_loss=0.2904, pruned_loss=0.05537, over 16761.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2774, pruned_loss=0.0502, over 3196196.83 frames. ], batch size: 83, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:23:02,766 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.219e+02 2.598e+02 3.036e+02 7.307e+02, threshold=5.195e+02, percent-clipped=1.0 2023-05-01 06:23:22,053 INFO [train.py:904] (5/8) Epoch 21, batch 4400, loss[loss=0.1879, simple_loss=0.2798, pruned_loss=0.04797, over 16645.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2792, pruned_loss=0.05127, over 3173441.40 frames. ], batch size: 62, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:23:48,521 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.4884, 2.3196, 2.4505, 3.5311, 2.8956, 3.7986, 1.4166, 2.7215], device='cuda:5'), covar=tensor([0.1391, 0.0933, 0.1188, 0.0147, 0.0244, 0.0335, 0.1789, 0.0838], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0174, 0.0193, 0.0190, 0.0206, 0.0214, 0.0200, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:24:36,523 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4561, 3.6351, 3.6819, 2.0324, 3.1217, 2.3581, 3.7082, 3.8552], device='cuda:5'), covar=tensor([0.0181, 0.0694, 0.0512, 0.1932, 0.0776, 0.0955, 0.0515, 0.0778], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0162, 0.0165, 0.0150, 0.0143, 0.0128, 0.0142, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:24:37,207 INFO [train.py:904] (5/8) Epoch 21, batch 4450, loss[loss=0.2207, simple_loss=0.3042, pruned_loss=0.06864, over 16695.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2824, pruned_loss=0.05237, over 3191080.42 frames. ], batch size: 57, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:24:40,660 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6995, 2.8031, 2.6797, 4.3689, 3.4020, 3.9901, 1.6218, 3.0035], device='cuda:5'), covar=tensor([0.1313, 0.0784, 0.1077, 0.0120, 0.0225, 0.0330, 0.1638, 0.0773], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0176, 0.0194, 0.0191, 0.0208, 0.0215, 0.0201, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:25:24,142 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 06:25:31,517 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.026e+02 2.353e+02 2.743e+02 3.990e+02, threshold=4.706e+02, percent-clipped=0.0 2023-05-01 06:25:33,788 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207490.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 06:25:50,493 INFO [train.py:904] (5/8) Epoch 21, batch 4500, loss[loss=0.1898, simple_loss=0.2801, pruned_loss=0.04973, over 16535.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2836, pruned_loss=0.05337, over 3195220.07 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:26:19,368 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0310, 2.0288, 2.5850, 2.9718, 2.8505, 3.3527, 2.2175, 3.3281], device='cuda:5'), covar=tensor([0.0176, 0.0501, 0.0296, 0.0272, 0.0288, 0.0155, 0.0523, 0.0109], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0193, 0.0178, 0.0185, 0.0196, 0.0153, 0.0197, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:26:35,640 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207532.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:26:43,761 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=207538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:27:03,833 INFO [train.py:904] (5/8) Epoch 21, batch 4550, loss[loss=0.2253, simple_loss=0.2915, pruned_loss=0.07957, over 11984.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2846, pruned_loss=0.05473, over 3206453.22 frames. ], batch size: 248, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:27:57,550 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.324e+02 1.874e+02 2.187e+02 2.648e+02 6.639e+02, threshold=4.374e+02, percent-clipped=1.0 2023-05-01 06:28:16,281 INFO [train.py:904] (5/8) Epoch 21, batch 4600, loss[loss=0.1921, simple_loss=0.2832, pruned_loss=0.05051, over 16855.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2856, pruned_loss=0.05487, over 3227958.56 frames. ], batch size: 39, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:28:19,878 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5779, 2.6354, 2.4354, 3.5675, 2.5489, 3.7131, 1.5768, 2.7561], device='cuda:5'), covar=tensor([0.1435, 0.0789, 0.1262, 0.0168, 0.0234, 0.0404, 0.1754, 0.0859], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0175, 0.0193, 0.0189, 0.0207, 0.0214, 0.0200, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:29:19,272 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4917, 4.6109, 4.3671, 4.0491, 4.0916, 4.4767, 4.1825, 4.1867], device='cuda:5'), covar=tensor([0.0507, 0.0322, 0.0232, 0.0227, 0.0617, 0.0329, 0.0555, 0.0490], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0414, 0.0337, 0.0332, 0.0347, 0.0385, 0.0231, 0.0405], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:29:29,138 INFO [train.py:904] (5/8) Epoch 21, batch 4650, loss[loss=0.1783, simple_loss=0.2695, pruned_loss=0.04356, over 16860.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2848, pruned_loss=0.05513, over 3224471.91 frames. ], batch size: 83, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:30:22,909 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6083, 2.6401, 2.3848, 3.9089, 2.9479, 3.9013, 1.4993, 2.7748], device='cuda:5'), covar=tensor([0.1370, 0.0803, 0.1255, 0.0178, 0.0234, 0.0342, 0.1714, 0.0831], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0190, 0.0208, 0.0215, 0.0201, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:30:23,480 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 1.925e+02 2.221e+02 2.572e+02 4.933e+02, threshold=4.441e+02, percent-clipped=1.0 2023-05-01 06:30:42,424 INFO [train.py:904] (5/8) Epoch 21, batch 4700, loss[loss=0.175, simple_loss=0.2581, pruned_loss=0.04594, over 16603.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2817, pruned_loss=0.05393, over 3214499.02 frames. ], batch size: 57, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:31:00,490 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9856, 5.4066, 5.5924, 5.3639, 5.3905, 5.9805, 5.4286, 5.1991], device='cuda:5'), covar=tensor([0.0870, 0.1535, 0.1710, 0.1801, 0.2288, 0.0739, 0.1250, 0.2057], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0584, 0.0636, 0.0487, 0.0648, 0.0672, 0.0499, 0.0649], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 06:31:56,757 INFO [train.py:904] (5/8) Epoch 21, batch 4750, loss[loss=0.1548, simple_loss=0.2476, pruned_loss=0.03096, over 16805.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2766, pruned_loss=0.05125, over 3226839.97 frames. ], batch size: 102, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:32:08,420 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2804, 5.5912, 5.3381, 5.3995, 5.1381, 5.0314, 5.0406, 5.6837], device='cuda:5'), covar=tensor([0.1087, 0.0747, 0.0902, 0.0717, 0.0700, 0.0744, 0.1073, 0.0817], device='cuda:5'), in_proj_covar=tensor([0.0663, 0.0807, 0.0672, 0.0613, 0.0514, 0.0522, 0.0679, 0.0633], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:32:35,989 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7398, 3.6137, 3.5452, 3.8736, 3.9725, 3.6089, 3.9345, 4.0055], device='cuda:5'), covar=tensor([0.1624, 0.1358, 0.2296, 0.0994, 0.0860, 0.2531, 0.1117, 0.1015], device='cuda:5'), in_proj_covar=tensor([0.0624, 0.0771, 0.0902, 0.0789, 0.0591, 0.0620, 0.0635, 0.0739], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:32:50,132 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 1.797e+02 2.146e+02 2.444e+02 4.649e+02, threshold=4.292e+02, percent-clipped=1.0 2023-05-01 06:33:11,168 INFO [train.py:904] (5/8) Epoch 21, batch 4800, loss[loss=0.1765, simple_loss=0.2628, pruned_loss=0.04512, over 16480.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.273, pruned_loss=0.04928, over 3223453.43 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:33:11,464 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1231, 5.5893, 5.7730, 5.4736, 5.5415, 6.0929, 5.5652, 5.3195], device='cuda:5'), covar=tensor([0.0821, 0.1394, 0.1667, 0.1866, 0.2150, 0.0770, 0.1202, 0.2026], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0583, 0.0634, 0.0485, 0.0645, 0.0670, 0.0498, 0.0647], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 06:33:55,487 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:34:24,542 INFO [train.py:904] (5/8) Epoch 21, batch 4850, loss[loss=0.2227, simple_loss=0.3038, pruned_loss=0.07081, over 11966.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.274, pruned_loss=0.0486, over 3203654.67 frames. ], batch size: 248, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:34:32,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9232, 3.2282, 3.2159, 2.0345, 2.9143, 3.2231, 3.0191, 1.8750], device='cuda:5'), covar=tensor([0.0551, 0.0060, 0.0060, 0.0432, 0.0118, 0.0114, 0.0120, 0.0478], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0083, 0.0082, 0.0133, 0.0097, 0.0108, 0.0093, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 06:34:39,566 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.62 vs. limit=5.0 2023-05-01 06:35:00,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3398, 3.4987, 3.6294, 3.6101, 3.6140, 3.4689, 3.4868, 3.4850], device='cuda:5'), covar=tensor([0.0398, 0.0574, 0.0452, 0.0413, 0.0481, 0.0445, 0.0758, 0.0477], device='cuda:5'), in_proj_covar=tensor([0.0401, 0.0445, 0.0432, 0.0401, 0.0475, 0.0453, 0.0542, 0.0363], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 06:35:08,244 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=207880.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:35:22,076 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 1.907e+02 2.222e+02 2.565e+02 4.310e+02, threshold=4.444e+02, percent-clipped=1.0 2023-05-01 06:35:40,336 INFO [train.py:904] (5/8) Epoch 21, batch 4900, loss[loss=0.1811, simple_loss=0.2767, pruned_loss=0.04277, over 16771.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2734, pruned_loss=0.04759, over 3191577.61 frames. ], batch size: 124, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:36:52,793 INFO [train.py:904] (5/8) Epoch 21, batch 4950, loss[loss=0.1632, simple_loss=0.2617, pruned_loss=0.03234, over 16725.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2727, pruned_loss=0.04671, over 3204910.83 frames. ], batch size: 89, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:37:03,146 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1393, 5.1383, 4.9454, 4.1843, 5.0680, 1.6443, 4.7295, 4.7141], device='cuda:5'), covar=tensor([0.0097, 0.0083, 0.0172, 0.0516, 0.0118, 0.2953, 0.0154, 0.0232], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0152, 0.0194, 0.0176, 0.0173, 0.0204, 0.0184, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:37:06,375 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1684, 4.3054, 4.0701, 3.8042, 3.8172, 4.1758, 3.9033, 3.9165], device='cuda:5'), covar=tensor([0.0636, 0.0475, 0.0285, 0.0282, 0.0761, 0.0433, 0.0772, 0.0562], device='cuda:5'), in_proj_covar=tensor([0.0286, 0.0413, 0.0335, 0.0330, 0.0344, 0.0384, 0.0228, 0.0402], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:37:47,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 1.964e+02 2.322e+02 2.960e+02 4.980e+02, threshold=4.644e+02, percent-clipped=1.0 2023-05-01 06:38:08,413 INFO [train.py:904] (5/8) Epoch 21, batch 5000, loss[loss=0.1754, simple_loss=0.2753, pruned_loss=0.03772, over 16751.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2749, pruned_loss=0.04697, over 3205181.40 frames. ], batch size: 83, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:38:49,193 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7753, 3.8594, 2.3060, 4.4787, 3.0407, 4.3419, 2.3818, 2.9386], device='cuda:5'), covar=tensor([0.0270, 0.0325, 0.1712, 0.0114, 0.0782, 0.0473, 0.1511, 0.0807], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0176, 0.0193, 0.0160, 0.0176, 0.0214, 0.0200, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:38:52,101 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208032.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:39:21,159 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 06:39:21,558 INFO [train.py:904] (5/8) Epoch 21, batch 5050, loss[loss=0.1756, simple_loss=0.2646, pruned_loss=0.04332, over 16616.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2747, pruned_loss=0.04662, over 3208005.69 frames. ], batch size: 57, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:40:18,554 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 1.998e+02 2.352e+02 2.773e+02 5.877e+02, threshold=4.704e+02, percent-clipped=4.0 2023-05-01 06:40:23,297 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208093.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 06:40:35,290 INFO [train.py:904] (5/8) Epoch 21, batch 5100, loss[loss=0.1679, simple_loss=0.2611, pruned_loss=0.03732, over 16875.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2725, pruned_loss=0.04581, over 3204646.62 frames. ], batch size: 109, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:41:48,637 INFO [train.py:904] (5/8) Epoch 21, batch 5150, loss[loss=0.1583, simple_loss=0.2549, pruned_loss=0.03087, over 16448.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2732, pruned_loss=0.04547, over 3199819.04 frames. ], batch size: 75, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:42:11,835 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 06:42:43,680 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 1.920e+02 2.264e+02 2.595e+02 3.716e+02, threshold=4.527e+02, percent-clipped=0.0 2023-05-01 06:42:59,321 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6335, 3.5938, 2.7705, 2.2512, 2.4875, 2.4730, 3.7950, 3.3312], device='cuda:5'), covar=tensor([0.2736, 0.0720, 0.1911, 0.2890, 0.2427, 0.1951, 0.0585, 0.1228], device='cuda:5'), in_proj_covar=tensor([0.0325, 0.0269, 0.0303, 0.0309, 0.0295, 0.0256, 0.0296, 0.0334], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 06:43:01,061 INFO [train.py:904] (5/8) Epoch 21, batch 5200, loss[loss=0.1697, simple_loss=0.2568, pruned_loss=0.04134, over 16516.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2722, pruned_loss=0.04527, over 3197365.80 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:11,873 INFO [train.py:904] (5/8) Epoch 21, batch 5250, loss[loss=0.174, simple_loss=0.2642, pruned_loss=0.04195, over 15335.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2697, pruned_loss=0.04491, over 3186887.12 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:23,764 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2939, 5.5780, 5.2970, 5.3662, 5.1226, 5.0160, 4.9799, 5.6578], device='cuda:5'), covar=tensor([0.1236, 0.0803, 0.1002, 0.0823, 0.0742, 0.0734, 0.1092, 0.0811], device='cuda:5'), in_proj_covar=tensor([0.0661, 0.0806, 0.0670, 0.0611, 0.0512, 0.0519, 0.0675, 0.0629], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:44:31,864 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:44:38,622 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-05-01 06:45:07,262 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 1.864e+02 2.120e+02 2.516e+02 5.355e+02, threshold=4.239e+02, percent-clipped=2.0 2023-05-01 06:45:15,578 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7803, 4.9931, 5.1551, 4.9573, 5.0398, 5.5606, 5.0005, 4.6732], device='cuda:5'), covar=tensor([0.0969, 0.1636, 0.1677, 0.1746, 0.2099, 0.0696, 0.1479, 0.2488], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0575, 0.0626, 0.0482, 0.0640, 0.0663, 0.0493, 0.0642], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 06:45:25,339 INFO [train.py:904] (5/8) Epoch 21, batch 5300, loss[loss=0.1511, simple_loss=0.2395, pruned_loss=0.03134, over 16765.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2665, pruned_loss=0.04376, over 3195237.24 frames. ], batch size: 76, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:45:57,503 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7144, 3.4918, 3.9471, 1.9177, 4.1632, 4.1426, 3.1107, 2.9924], device='cuda:5'), covar=tensor([0.0698, 0.0258, 0.0146, 0.1161, 0.0049, 0.0106, 0.0386, 0.0472], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0109, 0.0096, 0.0139, 0.0080, 0.0123, 0.0129, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:46:00,485 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208326.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:46:23,561 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 06:46:33,905 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 06:46:37,966 INFO [train.py:904] (5/8) Epoch 21, batch 5350, loss[loss=0.1912, simple_loss=0.2842, pruned_loss=0.04915, over 16513.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2645, pruned_loss=0.04296, over 3192627.40 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:47:09,028 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7947, 3.9078, 2.3304, 4.5674, 3.0865, 4.4412, 2.5320, 3.1159], device='cuda:5'), covar=tensor([0.0257, 0.0341, 0.1684, 0.0127, 0.0742, 0.0468, 0.1458, 0.0704], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0174, 0.0191, 0.0158, 0.0174, 0.0212, 0.0199, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:47:33,173 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208388.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 06:47:34,966 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.963e+02 2.291e+02 2.599e+02 4.161e+02, threshold=4.581e+02, percent-clipped=0.0 2023-05-01 06:47:53,658 INFO [train.py:904] (5/8) Epoch 21, batch 5400, loss[loss=0.1863, simple_loss=0.2735, pruned_loss=0.04953, over 12051.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2675, pruned_loss=0.04356, over 3176933.73 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:48:56,019 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-05-01 06:49:10,981 INFO [train.py:904] (5/8) Epoch 21, batch 5450, loss[loss=0.2046, simple_loss=0.2909, pruned_loss=0.05912, over 12224.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2703, pruned_loss=0.04484, over 3176861.90 frames. ], batch size: 247, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:49:20,130 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8508, 2.8142, 2.6445, 4.7731, 3.6402, 4.1208, 1.6129, 3.1322], device='cuda:5'), covar=tensor([0.1248, 0.0778, 0.1178, 0.0177, 0.0286, 0.0420, 0.1608, 0.0763], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0174, 0.0193, 0.0188, 0.0207, 0.0213, 0.0200, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:50:09,144 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.336e+02 2.337e+02 2.736e+02 3.594e+02 9.529e+02, threshold=5.472e+02, percent-clipped=14.0 2023-05-01 06:50:28,300 INFO [train.py:904] (5/8) Epoch 21, batch 5500, loss[loss=0.2063, simple_loss=0.2967, pruned_loss=0.05796, over 16522.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2773, pruned_loss=0.04944, over 3149606.91 frames. ], batch size: 75, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:51:47,653 INFO [train.py:904] (5/8) Epoch 21, batch 5550, loss[loss=0.2057, simple_loss=0.2948, pruned_loss=0.05823, over 16521.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2846, pruned_loss=0.05513, over 3105926.11 frames. ], batch size: 68, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:52:27,853 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208576.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:52:38,029 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9036, 4.1481, 3.9496, 4.0008, 3.6931, 3.7715, 3.8368, 4.1263], device='cuda:5'), covar=tensor([0.1073, 0.0888, 0.0990, 0.0886, 0.0796, 0.1485, 0.0870, 0.1008], device='cuda:5'), in_proj_covar=tensor([0.0654, 0.0796, 0.0664, 0.0603, 0.0505, 0.0512, 0.0667, 0.0622], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:52:40,724 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0969, 4.2673, 2.8336, 4.9591, 3.5367, 4.7926, 2.8654, 3.4127], device='cuda:5'), covar=tensor([0.0253, 0.0307, 0.1401, 0.0180, 0.0650, 0.0539, 0.1428, 0.0694], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0175, 0.0192, 0.0160, 0.0175, 0.0214, 0.0199, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:52:49,286 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.296e+02 3.115e+02 3.921e+02 4.748e+02 1.231e+03, threshold=7.841e+02, percent-clipped=12.0 2023-05-01 06:53:07,717 INFO [train.py:904] (5/8) Epoch 21, batch 5600, loss[loss=0.2165, simple_loss=0.2984, pruned_loss=0.06736, over 16450.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2892, pruned_loss=0.05911, over 3073130.13 frames. ], batch size: 75, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:53:39,438 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:07,054 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208637.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:20,065 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6369, 3.6916, 2.1449, 4.2399, 2.8051, 4.1346, 2.3075, 2.9528], device='cuda:5'), covar=tensor([0.0288, 0.0401, 0.1793, 0.0196, 0.0827, 0.0625, 0.1701, 0.0807], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0175, 0.0192, 0.0160, 0.0175, 0.0214, 0.0200, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 06:54:29,871 INFO [train.py:904] (5/8) Epoch 21, batch 5650, loss[loss=0.2291, simple_loss=0.3258, pruned_loss=0.06614, over 17193.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.294, pruned_loss=0.06287, over 3048295.39 frames. ], batch size: 44, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:55:12,183 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 06:55:28,178 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:55:30,668 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.148e+02 3.160e+02 3.845e+02 4.644e+02 8.553e+02, threshold=7.690e+02, percent-clipped=3.0 2023-05-01 06:55:50,862 INFO [train.py:904] (5/8) Epoch 21, batch 5700, loss[loss=0.2052, simple_loss=0.3058, pruned_loss=0.05226, over 16882.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2961, pruned_loss=0.06513, over 3018730.64 frames. ], batch size: 116, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:56:45,735 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:56:52,892 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 06:57:10,979 INFO [train.py:904] (5/8) Epoch 21, batch 5750, loss[loss=0.2338, simple_loss=0.2994, pruned_loss=0.08412, over 11280.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2991, pruned_loss=0.06688, over 3001080.80 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:58:13,922 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.275e+02 2.983e+02 3.733e+02 4.569e+02 7.128e+02, threshold=7.466e+02, percent-clipped=0.0 2023-05-01 06:58:33,859 INFO [train.py:904] (5/8) Epoch 21, batch 5800, loss[loss=0.1838, simple_loss=0.2817, pruned_loss=0.04295, over 17231.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.299, pruned_loss=0.0656, over 2992221.17 frames. ], batch size: 52, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:58:39,931 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 06:58:49,265 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2140, 4.0757, 4.2701, 4.4050, 4.5365, 4.1285, 4.4925, 4.5583], device='cuda:5'), covar=tensor([0.1864, 0.1270, 0.1555, 0.0726, 0.0612, 0.1243, 0.0860, 0.0649], device='cuda:5'), in_proj_covar=tensor([0.0619, 0.0763, 0.0893, 0.0780, 0.0585, 0.0617, 0.0630, 0.0731], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 06:59:52,593 INFO [train.py:904] (5/8) Epoch 21, batch 5850, loss[loss=0.1909, simple_loss=0.2851, pruned_loss=0.04842, over 16739.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.297, pruned_loss=0.06387, over 3004108.25 frames. ], batch size: 83, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:00:53,530 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.740e+02 3.068e+02 3.692e+02 6.527e+02, threshold=6.136e+02, percent-clipped=0.0 2023-05-01 07:01:12,721 INFO [train.py:904] (5/8) Epoch 21, batch 5900, loss[loss=0.192, simple_loss=0.2879, pruned_loss=0.04799, over 17205.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2964, pruned_loss=0.06347, over 3017921.93 frames. ], batch size: 45, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:01:48,446 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:02:05,305 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208932.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:02:07,765 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3863, 3.2605, 2.6724, 2.1408, 2.2436, 2.2934, 3.3811, 3.0537], device='cuda:5'), covar=tensor([0.2930, 0.0677, 0.1731, 0.2730, 0.2546, 0.2074, 0.0521, 0.1272], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0267, 0.0301, 0.0307, 0.0293, 0.0254, 0.0293, 0.0331], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 07:02:36,229 INFO [train.py:904] (5/8) Epoch 21, batch 5950, loss[loss=0.189, simple_loss=0.2889, pruned_loss=0.04457, over 17037.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2972, pruned_loss=0.06173, over 3037140.66 frames. ], batch size: 50, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:02,963 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:03:36,546 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.651e+02 3.328e+02 3.727e+02 9.744e+02, threshold=6.656e+02, percent-clipped=4.0 2023-05-01 07:03:56,995 INFO [train.py:904] (5/8) Epoch 21, batch 6000, loss[loss=0.2154, simple_loss=0.298, pruned_loss=0.0664, over 15212.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2957, pruned_loss=0.06077, over 3067295.58 frames. ], batch size: 190, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:56,995 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 07:04:08,266 INFO [train.py:938] (5/8) Epoch 21, validation: loss=0.1512, simple_loss=0.2639, pruned_loss=0.01924, over 944034.00 frames. 2023-05-01 07:04:08,267 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 07:05:25,682 INFO [train.py:904] (5/8) Epoch 21, batch 6050, loss[loss=0.2565, simple_loss=0.3169, pruned_loss=0.09809, over 11606.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2934, pruned_loss=0.05986, over 3073304.97 frames. ], batch size: 246, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:06:26,994 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.899e+02 2.646e+02 3.170e+02 4.163e+02 9.783e+02, threshold=6.340e+02, percent-clipped=1.0 2023-05-01 07:06:45,922 INFO [train.py:904] (5/8) Epoch 21, batch 6100, loss[loss=0.2145, simple_loss=0.2939, pruned_loss=0.06759, over 16584.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2932, pruned_loss=0.05912, over 3097577.73 frames. ], batch size: 57, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:07:05,156 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 07:08:04,727 INFO [train.py:904] (5/8) Epoch 21, batch 6150, loss[loss=0.2129, simple_loss=0.296, pruned_loss=0.06486, over 16477.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2912, pruned_loss=0.05867, over 3107014.35 frames. ], batch size: 75, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:09:04,299 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.750e+02 2.685e+02 3.365e+02 4.144e+02 7.779e+02, threshold=6.731e+02, percent-clipped=2.0 2023-05-01 07:09:23,307 INFO [train.py:904] (5/8) Epoch 21, batch 6200, loss[loss=0.1969, simple_loss=0.2801, pruned_loss=0.05684, over 16697.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2891, pruned_loss=0.05812, over 3106722.08 frames. ], batch size: 62, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:09:23,866 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:09:39,961 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7736, 4.8519, 5.2073, 5.1608, 5.2069, 4.8737, 4.8560, 4.5850], device='cuda:5'), covar=tensor([0.0305, 0.0532, 0.0362, 0.0383, 0.0432, 0.0349, 0.0879, 0.0504], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0444, 0.0432, 0.0400, 0.0476, 0.0454, 0.0539, 0.0364], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 07:09:42,527 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8858, 1.6967, 2.3816, 2.7692, 2.6601, 3.1148, 1.7069, 3.1271], device='cuda:5'), covar=tensor([0.0164, 0.0589, 0.0319, 0.0261, 0.0291, 0.0169, 0.0673, 0.0124], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0192, 0.0177, 0.0183, 0.0194, 0.0151, 0.0194, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:10:10,123 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209232.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:39,911 INFO [train.py:904] (5/8) Epoch 21, batch 6250, loss[loss=0.2173, simple_loss=0.3016, pruned_loss=0.06654, over 15255.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2885, pruned_loss=0.05748, over 3129861.54 frames. ], batch size: 191, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:10:46,262 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:57,593 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209263.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:22,154 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209280.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:35,898 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.681e+02 3.167e+02 3.896e+02 8.805e+02, threshold=6.333e+02, percent-clipped=2.0 2023-05-01 07:11:43,498 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4945, 3.5141, 2.0642, 3.8800, 2.6354, 3.8747, 2.1786, 2.7484], device='cuda:5'), covar=tensor([0.0278, 0.0406, 0.1717, 0.0253, 0.0834, 0.0639, 0.1589, 0.0795], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0161, 0.0176, 0.0215, 0.0200, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 07:11:50,469 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209298.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:54,998 INFO [train.py:904] (5/8) Epoch 21, batch 6300, loss[loss=0.1778, simple_loss=0.2649, pruned_loss=0.04536, over 16636.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2885, pruned_loss=0.05711, over 3128990.70 frames. ], batch size: 62, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:12:01,315 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 07:12:17,961 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:13:13,588 INFO [train.py:904] (5/8) Epoch 21, batch 6350, loss[loss=0.1876, simple_loss=0.274, pruned_loss=0.05055, over 16859.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2892, pruned_loss=0.05826, over 3123827.17 frames. ], batch size: 42, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:13:25,056 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:13:46,245 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 07:14:13,990 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 3.041e+02 3.551e+02 4.346e+02 9.300e+02, threshold=7.102e+02, percent-clipped=4.0 2023-05-01 07:14:31,820 INFO [train.py:904] (5/8) Epoch 21, batch 6400, loss[loss=0.1996, simple_loss=0.2806, pruned_loss=0.0593, over 17207.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2898, pruned_loss=0.05964, over 3105235.34 frames. ], batch size: 44, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:15:48,019 INFO [train.py:904] (5/8) Epoch 21, batch 6450, loss[loss=0.2144, simple_loss=0.2917, pruned_loss=0.06858, over 11519.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2893, pruned_loss=0.05834, over 3111347.29 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:15:49,609 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 07:16:10,824 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209466.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:16:52,761 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.825e+02 3.422e+02 4.079e+02 7.372e+02, threshold=6.844e+02, percent-clipped=3.0 2023-05-01 07:17:04,854 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 07:17:08,465 INFO [train.py:904] (5/8) Epoch 21, batch 6500, loss[loss=0.2, simple_loss=0.2857, pruned_loss=0.05711, over 16748.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2879, pruned_loss=0.05781, over 3110677.04 frames. ], batch size: 124, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:17:20,885 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:17:48,032 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209527.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:18:29,361 INFO [train.py:904] (5/8) Epoch 21, batch 6550, loss[loss=0.1911, simple_loss=0.2975, pruned_loss=0.04237, over 16893.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2899, pruned_loss=0.0581, over 3120555.96 frames. ], batch size: 96, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:18:40,055 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:18:59,389 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7694, 1.3100, 1.6210, 1.6245, 1.7991, 1.8574, 1.6007, 1.7737], device='cuda:5'), covar=tensor([0.0238, 0.0407, 0.0239, 0.0321, 0.0296, 0.0187, 0.0436, 0.0140], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0191, 0.0176, 0.0181, 0.0193, 0.0150, 0.0193, 0.0148], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:19:01,125 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209571.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:19:33,743 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.646e+02 3.318e+02 3.875e+02 1.016e+03, threshold=6.636e+02, percent-clipped=2.0 2023-05-01 07:19:49,526 INFO [train.py:904] (5/8) Epoch 21, batch 6600, loss[loss=0.2199, simple_loss=0.3054, pruned_loss=0.06718, over 15312.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.292, pruned_loss=0.0586, over 3122818.84 frames. ], batch size: 190, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:19:55,982 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:20:03,697 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:20:46,830 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0375, 2.3950, 2.4064, 2.8062, 2.0703, 3.1829, 1.8750, 2.7186], device='cuda:5'), covar=tensor([0.1100, 0.0587, 0.1045, 0.0187, 0.0126, 0.0373, 0.1445, 0.0699], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0174, 0.0195, 0.0189, 0.0206, 0.0214, 0.0201, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 07:21:02,613 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3980, 4.4549, 4.2781, 4.0118, 4.0054, 4.3738, 4.1289, 4.0616], device='cuda:5'), covar=tensor([0.0596, 0.0529, 0.0304, 0.0306, 0.0873, 0.0472, 0.0646, 0.0650], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0412, 0.0332, 0.0329, 0.0342, 0.0382, 0.0229, 0.0399], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:21:07,997 INFO [train.py:904] (5/8) Epoch 21, batch 6650, loss[loss=0.1997, simple_loss=0.2873, pruned_loss=0.05601, over 16885.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2923, pruned_loss=0.05947, over 3128223.37 frames. ], batch size: 109, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:21:12,197 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209654.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:21:33,134 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209667.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:04,994 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:11,941 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 3.003e+02 3.834e+02 5.149e+02 1.142e+03, threshold=7.669e+02, percent-clipped=10.0 2023-05-01 07:22:25,649 INFO [train.py:904] (5/8) Epoch 21, batch 6700, loss[loss=0.2337, simple_loss=0.2973, pruned_loss=0.08505, over 10947.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2911, pruned_loss=0.06014, over 3110114.21 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:23:40,264 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209749.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:23:44,127 INFO [train.py:904] (5/8) Epoch 21, batch 6750, loss[loss=0.1972, simple_loss=0.2876, pruned_loss=0.05335, over 16492.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2899, pruned_loss=0.05968, over 3129441.01 frames. ], batch size: 75, lr: 3.20e-03, grad_scale: 2.0 2023-05-01 07:24:40,636 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8075, 3.1449, 3.3033, 1.9005, 2.7623, 2.0803, 3.2750, 3.4447], device='cuda:5'), covar=tensor([0.0288, 0.0841, 0.0616, 0.2176, 0.0950, 0.1054, 0.0693, 0.0913], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0143, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 07:24:47,251 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.958e+02 2.606e+02 3.292e+02 3.909e+02 7.681e+02, threshold=6.584e+02, percent-clipped=1.0 2023-05-01 07:25:01,557 INFO [train.py:904] (5/8) Epoch 21, batch 6800, loss[loss=0.2108, simple_loss=0.2995, pruned_loss=0.06104, over 15378.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2906, pruned_loss=0.06008, over 3117485.68 frames. ], batch size: 190, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:25:33,967 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:04,581 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2830, 3.6566, 3.7260, 2.5042, 3.5014, 3.7762, 3.4641, 2.1173], device='cuda:5'), covar=tensor([0.0526, 0.0072, 0.0056, 0.0397, 0.0094, 0.0122, 0.0094, 0.0454], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0133, 0.0097, 0.0108, 0.0093, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 07:26:20,413 INFO [train.py:904] (5/8) Epoch 21, batch 6850, loss[loss=0.2181, simple_loss=0.3179, pruned_loss=0.05916, over 16861.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2915, pruned_loss=0.0601, over 3112030.65 frames. ], batch size: 116, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:26:29,620 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:40,670 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 07:26:42,203 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:21,614 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.648e+02 3.249e+02 3.781e+02 8.526e+02, threshold=6.499e+02, percent-clipped=1.0 2023-05-01 07:27:34,660 INFO [train.py:904] (5/8) Epoch 21, batch 6900, loss[loss=0.2505, simple_loss=0.3139, pruned_loss=0.0935, over 11607.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2936, pruned_loss=0.05978, over 3116422.03 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:27:41,304 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:48,395 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:28:42,541 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0293, 4.1102, 3.9083, 3.6764, 3.6688, 4.0152, 3.7315, 3.7988], device='cuda:5'), covar=tensor([0.0583, 0.0564, 0.0290, 0.0284, 0.0722, 0.0497, 0.1004, 0.0590], device='cuda:5'), in_proj_covar=tensor([0.0283, 0.0411, 0.0331, 0.0328, 0.0342, 0.0382, 0.0228, 0.0399], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:28:51,084 INFO [train.py:904] (5/8) Epoch 21, batch 6950, loss[loss=0.2038, simple_loss=0.2992, pruned_loss=0.05421, over 17146.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2952, pruned_loss=0.06116, over 3093167.82 frames. ], batch size: 46, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:28:54,459 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:02,108 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209959.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:06,368 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209962.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:22,404 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 07:29:51,934 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.035e+02 3.251e+02 3.766e+02 4.564e+02 9.580e+02, threshold=7.532e+02, percent-clipped=9.0 2023-05-01 07:30:02,745 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 07:30:09,085 INFO [train.py:904] (5/8) Epoch 21, batch 7000, loss[loss=0.1982, simple_loss=0.2963, pruned_loss=0.05004, over 16230.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2946, pruned_loss=0.06039, over 3081066.13 frames. ], batch size: 165, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:30:09,333 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210002.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:31:11,199 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:31:22,079 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6263, 2.4988, 2.3275, 3.7339, 2.7452, 3.8162, 1.4967, 2.7549], device='cuda:5'), covar=tensor([0.1367, 0.0802, 0.1314, 0.0210, 0.0249, 0.0448, 0.1693, 0.0860], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0189, 0.0207, 0.0215, 0.0202, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 07:31:22,810 INFO [train.py:904] (5/8) Epoch 21, batch 7050, loss[loss=0.2086, simple_loss=0.2969, pruned_loss=0.06018, over 16728.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2953, pruned_loss=0.05973, over 3101059.44 frames. ], batch size: 134, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:31:30,578 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:32:24,161 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.868e+02 2.885e+02 3.523e+02 4.096e+02 8.338e+02, threshold=7.047e+02, percent-clipped=1.0 2023-05-01 07:32:37,545 INFO [train.py:904] (5/8) Epoch 21, batch 7100, loss[loss=0.2303, simple_loss=0.2956, pruned_loss=0.08254, over 11189.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2937, pruned_loss=0.05939, over 3100836.80 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:32:50,877 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3769, 5.4032, 5.2246, 4.8846, 4.8575, 5.2935, 5.1502, 4.9549], device='cuda:5'), covar=tensor([0.0526, 0.0375, 0.0266, 0.0287, 0.1012, 0.0370, 0.0295, 0.0602], device='cuda:5'), in_proj_covar=tensor([0.0280, 0.0405, 0.0327, 0.0323, 0.0338, 0.0377, 0.0225, 0.0394], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:32:58,906 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2317, 4.3086, 4.4664, 4.2810, 4.3775, 4.8326, 4.4079, 4.1431], device='cuda:5'), covar=tensor([0.1704, 0.2179, 0.2341, 0.2103, 0.2586, 0.1187, 0.1584, 0.2503], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0589, 0.0649, 0.0491, 0.0650, 0.0680, 0.0507, 0.0659], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 07:33:03,278 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:09,965 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:55,445 INFO [train.py:904] (5/8) Epoch 21, batch 7150, loss[loss=0.2421, simple_loss=0.3058, pruned_loss=0.08915, over 11197.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2923, pruned_loss=0.0593, over 3122576.32 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:34:16,493 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:21,275 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210170.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:53,947 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.076e+02 2.584e+02 3.186e+02 4.207e+02 8.537e+02, threshold=6.371e+02, percent-clipped=2.0 2023-05-01 07:35:07,917 INFO [train.py:904] (5/8) Epoch 21, batch 7200, loss[loss=0.1713, simple_loss=0.2645, pruned_loss=0.03908, over 16461.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2901, pruned_loss=0.05744, over 3125957.84 frames. ], batch size: 68, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:35:15,057 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4013, 5.7154, 5.4752, 5.5163, 5.1170, 5.1484, 5.1218, 5.8380], device='cuda:5'), covar=tensor([0.1171, 0.0787, 0.0931, 0.0847, 0.0835, 0.0743, 0.1104, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0662, 0.0801, 0.0670, 0.0610, 0.0510, 0.0521, 0.0675, 0.0630], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:35:27,063 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:35:27,157 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3384, 3.4984, 3.6319, 3.6062, 3.6104, 3.4409, 3.4644, 3.4871], device='cuda:5'), covar=tensor([0.0394, 0.0599, 0.0417, 0.0423, 0.0519, 0.0488, 0.0786, 0.0534], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0444, 0.0432, 0.0402, 0.0479, 0.0454, 0.0540, 0.0363], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 07:36:28,286 INFO [train.py:904] (5/8) Epoch 21, batch 7250, loss[loss=0.1908, simple_loss=0.2703, pruned_loss=0.05563, over 16444.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2871, pruned_loss=0.05587, over 3131660.15 frames. ], batch size: 35, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:36:43,820 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210262.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:37:31,781 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 2.676e+02 3.072e+02 3.848e+02 7.717e+02, threshold=6.145e+02, percent-clipped=2.0 2023-05-01 07:37:44,655 INFO [train.py:904] (5/8) Epoch 21, batch 7300, loss[loss=0.2384, simple_loss=0.3026, pruned_loss=0.08706, over 11493.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2873, pruned_loss=0.05646, over 3127380.64 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:37:57,775 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8880, 4.8826, 4.6446, 3.9084, 4.7940, 1.7900, 4.5432, 4.3406], device='cuda:5'), covar=tensor([0.0072, 0.0064, 0.0190, 0.0373, 0.0074, 0.2839, 0.0128, 0.0248], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0150, 0.0193, 0.0174, 0.0171, 0.0203, 0.0182, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:37:58,862 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:48,842 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210344.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:59,112 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 07:38:59,617 INFO [train.py:904] (5/8) Epoch 21, batch 7350, loss[loss=0.2286, simple_loss=0.3182, pruned_loss=0.06955, over 15419.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2882, pruned_loss=0.05755, over 3104729.31 frames. ], batch size: 191, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:39:25,039 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0771, 5.3852, 5.1227, 5.1405, 4.8493, 4.8055, 4.7886, 5.4735], device='cuda:5'), covar=tensor([0.1145, 0.0765, 0.0881, 0.0836, 0.0752, 0.0959, 0.1133, 0.0753], device='cuda:5'), in_proj_covar=tensor([0.0653, 0.0793, 0.0663, 0.0603, 0.0505, 0.0515, 0.0668, 0.0622], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:40:00,161 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:40:02,085 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.648e+02 3.204e+02 3.841e+02 6.410e+02, threshold=6.409e+02, percent-clipped=2.0 2023-05-01 07:40:14,863 INFO [train.py:904] (5/8) Epoch 21, batch 7400, loss[loss=0.2234, simple_loss=0.2928, pruned_loss=0.07706, over 11289.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2891, pruned_loss=0.05794, over 3103933.32 frames. ], batch size: 247, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:40:32,312 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210413.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:41:08,468 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-01 07:41:21,587 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5443, 1.6404, 2.1222, 2.4323, 2.4888, 2.7103, 1.6868, 2.7414], device='cuda:5'), covar=tensor([0.0246, 0.0545, 0.0369, 0.0350, 0.0318, 0.0207, 0.0627, 0.0153], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0191, 0.0176, 0.0182, 0.0194, 0.0150, 0.0193, 0.0148], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:41:29,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 07:41:32,234 INFO [train.py:904] (5/8) Epoch 21, batch 7450, loss[loss=0.2586, simple_loss=0.3159, pruned_loss=0.1006, over 11530.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2907, pruned_loss=0.05982, over 3078250.09 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:42:11,635 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9765, 2.2921, 2.3138, 2.9985, 1.8121, 3.1968, 1.8156, 2.7353], device='cuda:5'), covar=tensor([0.1209, 0.0658, 0.1177, 0.0254, 0.0109, 0.0401, 0.1496, 0.0704], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0174, 0.0195, 0.0189, 0.0206, 0.0215, 0.0201, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 07:42:42,645 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.331e+02 3.072e+02 3.553e+02 4.443e+02 7.195e+02, threshold=7.106e+02, percent-clipped=1.0 2023-05-01 07:42:53,266 INFO [train.py:904] (5/8) Epoch 21, batch 7500, loss[loss=0.1894, simple_loss=0.2712, pruned_loss=0.05385, over 17031.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2909, pruned_loss=0.0589, over 3079061.52 frames. ], batch size: 55, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:44:09,162 INFO [train.py:904] (5/8) Epoch 21, batch 7550, loss[loss=0.1871, simple_loss=0.2712, pruned_loss=0.0515, over 16985.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2904, pruned_loss=0.05966, over 3064036.86 frames. ], batch size: 41, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:44:19,871 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6611, 3.9585, 3.0684, 2.3159, 2.7144, 2.4806, 4.2709, 3.5328], device='cuda:5'), covar=tensor([0.2867, 0.0549, 0.1648, 0.2546, 0.2403, 0.2041, 0.0390, 0.1171], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0271, 0.0305, 0.0312, 0.0298, 0.0258, 0.0295, 0.0336], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 07:45:11,355 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.775e+02 3.392e+02 4.108e+02 6.854e+02, threshold=6.785e+02, percent-clipped=0.0 2023-05-01 07:45:23,228 INFO [train.py:904] (5/8) Epoch 21, batch 7600, loss[loss=0.2029, simple_loss=0.2887, pruned_loss=0.05857, over 16324.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2894, pruned_loss=0.05932, over 3075280.24 frames. ], batch size: 146, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:46:08,233 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210632.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:46:37,240 INFO [train.py:904] (5/8) Epoch 21, batch 7650, loss[loss=0.201, simple_loss=0.2925, pruned_loss=0.05474, over 16928.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2904, pruned_loss=0.0602, over 3079817.18 frames. ], batch size: 90, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:46:59,512 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210666.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:25,972 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,164 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,763 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 3.052e+02 3.513e+02 4.198e+02 7.732e+02, threshold=7.025e+02, percent-clipped=2.0 2023-05-01 07:47:50,164 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8861, 3.9684, 2.4730, 4.8650, 3.2049, 4.7525, 2.5657, 3.1965], device='cuda:5'), covar=tensor([0.0289, 0.0395, 0.1687, 0.0167, 0.0763, 0.0489, 0.1485, 0.0772], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0175, 0.0193, 0.0159, 0.0176, 0.0215, 0.0201, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 07:47:55,358 INFO [train.py:904] (5/8) Epoch 21, batch 7700, loss[loss=0.234, simple_loss=0.3008, pruned_loss=0.08361, over 11441.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2896, pruned_loss=0.05971, over 3106906.58 frames. ], batch size: 247, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:47:58,090 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 07:48:12,365 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:34,115 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:56,535 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5379, 4.3354, 4.4541, 4.7295, 4.8874, 4.4885, 4.9559, 4.9030], device='cuda:5'), covar=tensor([0.1942, 0.1459, 0.2079, 0.0909, 0.0787, 0.1060, 0.0732, 0.0831], device='cuda:5'), in_proj_covar=tensor([0.0620, 0.0762, 0.0885, 0.0777, 0.0584, 0.0614, 0.0633, 0.0732], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:48:59,583 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:12,539 INFO [train.py:904] (5/8) Epoch 21, batch 7750, loss[loss=0.1736, simple_loss=0.2731, pruned_loss=0.03705, over 16790.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.29, pruned_loss=0.05961, over 3113732.75 frames. ], batch size: 89, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:49:27,500 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210761.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:33,023 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210764.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:50:07,470 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-01 07:50:11,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3880, 3.2506, 3.5692, 1.8347, 3.7623, 3.7860, 2.8158, 2.8055], device='cuda:5'), covar=tensor([0.0836, 0.0274, 0.0195, 0.1191, 0.0074, 0.0185, 0.0479, 0.0474], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0107, 0.0096, 0.0137, 0.0079, 0.0123, 0.0127, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 07:50:17,833 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8337, 5.2053, 5.4607, 5.2210, 5.3238, 5.8119, 5.2696, 5.0435], device='cuda:5'), covar=tensor([0.1029, 0.1956, 0.2068, 0.1938, 0.2308, 0.0883, 0.1627, 0.2458], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0592, 0.0650, 0.0495, 0.0650, 0.0684, 0.0511, 0.0664], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 07:50:18,620 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.795e+02 3.327e+02 4.093e+02 7.547e+02, threshold=6.654e+02, percent-clipped=2.0 2023-05-01 07:50:31,016 INFO [train.py:904] (5/8) Epoch 21, batch 7800, loss[loss=0.178, simple_loss=0.2709, pruned_loss=0.04254, over 16681.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2906, pruned_loss=0.06071, over 3075766.50 frames. ], batch size: 62, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:51:07,556 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210825.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:51:28,686 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6903, 3.9975, 3.1047, 2.3087, 2.7599, 2.5325, 4.3311, 3.5488], device='cuda:5'), covar=tensor([0.2822, 0.0613, 0.1625, 0.2694, 0.2477, 0.1952, 0.0393, 0.1167], device='cuda:5'), in_proj_covar=tensor([0.0326, 0.0268, 0.0302, 0.0309, 0.0294, 0.0256, 0.0292, 0.0331], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 07:51:48,928 INFO [train.py:904] (5/8) Epoch 21, batch 7850, loss[loss=0.193, simple_loss=0.2875, pruned_loss=0.0492, over 17105.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2917, pruned_loss=0.06091, over 3067110.44 frames. ], batch size: 48, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:52:01,344 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-01 07:52:32,063 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-05-01 07:52:54,104 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 2.722e+02 3.190e+02 3.938e+02 6.813e+02, threshold=6.379e+02, percent-clipped=1.0 2023-05-01 07:53:05,684 INFO [train.py:904] (5/8) Epoch 21, batch 7900, loss[loss=0.2051, simple_loss=0.2958, pruned_loss=0.0572, over 16718.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2903, pruned_loss=0.05978, over 3078131.76 frames. ], batch size: 76, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:53:31,819 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 07:54:24,399 INFO [train.py:904] (5/8) Epoch 21, batch 7950, loss[loss=0.2138, simple_loss=0.2966, pruned_loss=0.06544, over 16393.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2904, pruned_loss=0.05997, over 3095805.86 frames. ], batch size: 146, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:55:14,499 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:20,871 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:29,114 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.961e+02 3.312e+02 4.005e+02 9.237e+02, threshold=6.624e+02, percent-clipped=4.0 2023-05-01 07:55:35,578 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4276, 4.4160, 4.2630, 3.5184, 4.3346, 1.7865, 4.0805, 4.0056], device='cuda:5'), covar=tensor([0.0129, 0.0105, 0.0204, 0.0386, 0.0105, 0.2807, 0.0156, 0.0240], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0151, 0.0195, 0.0174, 0.0172, 0.0204, 0.0182, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:55:41,325 INFO [train.py:904] (5/8) Epoch 21, batch 8000, loss[loss=0.1897, simple_loss=0.2874, pruned_loss=0.04601, over 16900.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2906, pruned_loss=0.05964, over 3117042.61 frames. ], batch size: 96, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:55:53,760 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3372, 3.1188, 3.3899, 1.8481, 3.5898, 3.6256, 2.7609, 2.7143], device='cuda:5'), covar=tensor([0.0791, 0.0251, 0.0180, 0.1113, 0.0070, 0.0176, 0.0459, 0.0440], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0106, 0.0096, 0.0136, 0.0079, 0.0122, 0.0127, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 07:56:12,605 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:35,958 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:45,511 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:56,546 INFO [train.py:904] (5/8) Epoch 21, batch 8050, loss[loss=0.2012, simple_loss=0.2923, pruned_loss=0.05503, over 17040.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2912, pruned_loss=0.06023, over 3094771.69 frames. ], batch size: 53, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:57:59,280 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.685e+02 2.757e+02 3.257e+02 3.943e+02 6.625e+02, threshold=6.515e+02, percent-clipped=2.0 2023-05-01 07:58:10,498 INFO [train.py:904] (5/8) Epoch 21, batch 8100, loss[loss=0.2572, simple_loss=0.322, pruned_loss=0.0962, over 11573.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2904, pruned_loss=0.05942, over 3102107.33 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:58:38,347 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211120.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:59:00,516 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2732, 2.3645, 2.3247, 4.0531, 2.2585, 2.6898, 2.4091, 2.5094], device='cuda:5'), covar=tensor([0.1299, 0.3445, 0.2876, 0.0503, 0.4048, 0.2460, 0.3339, 0.3169], device='cuda:5'), in_proj_covar=tensor([0.0396, 0.0444, 0.0363, 0.0324, 0.0435, 0.0509, 0.0413, 0.0517], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 07:59:22,862 INFO [train.py:904] (5/8) Epoch 21, batch 8150, loss[loss=0.1817, simple_loss=0.2592, pruned_loss=0.05212, over 16523.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2883, pruned_loss=0.05874, over 3111072.56 frames. ], batch size: 62, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:00:07,094 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 08:00:14,049 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5394, 1.8087, 2.1205, 2.4671, 2.5245, 2.8136, 1.9038, 2.7401], device='cuda:5'), covar=tensor([0.0236, 0.0514, 0.0354, 0.0356, 0.0323, 0.0191, 0.0517, 0.0139], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0191, 0.0176, 0.0182, 0.0194, 0.0151, 0.0193, 0.0148], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:00:27,470 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.753e+02 3.326e+02 4.060e+02 8.278e+02, threshold=6.652e+02, percent-clipped=2.0 2023-05-01 08:00:40,754 INFO [train.py:904] (5/8) Epoch 21, batch 8200, loss[loss=0.1968, simple_loss=0.2864, pruned_loss=0.05354, over 16442.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2861, pruned_loss=0.05816, over 3104429.89 frames. ], batch size: 68, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:01:13,050 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-05-01 08:01:59,677 INFO [train.py:904] (5/8) Epoch 21, batch 8250, loss[loss=0.1936, simple_loss=0.2866, pruned_loss=0.05031, over 16634.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2847, pruned_loss=0.05543, over 3092015.35 frames. ], batch size: 134, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:02:56,892 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211288.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:03:06,708 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.346e+02 2.848e+02 3.670e+02 8.296e+02, threshold=5.695e+02, percent-clipped=3.0 2023-05-01 08:03:18,682 INFO [train.py:904] (5/8) Epoch 21, batch 8300, loss[loss=0.1983, simple_loss=0.2921, pruned_loss=0.05225, over 16653.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2822, pruned_loss=0.05254, over 3084791.31 frames. ], batch size: 134, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:03:26,171 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2201, 4.3782, 4.1196, 3.8642, 3.6790, 4.2741, 3.9542, 3.8882], device='cuda:5'), covar=tensor([0.0662, 0.0707, 0.0419, 0.0400, 0.1095, 0.0557, 0.0843, 0.0781], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0411, 0.0330, 0.0325, 0.0339, 0.0379, 0.0228, 0.0395], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:03:51,801 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211322.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:12,348 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211336.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:16,103 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:18,496 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:36,937 INFO [train.py:904] (5/8) Epoch 21, batch 8350, loss[loss=0.2247, simple_loss=0.3002, pruned_loss=0.07463, over 12486.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2818, pruned_loss=0.05079, over 3083622.69 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:05:05,460 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,342 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,456 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:43,492 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.192e+02 2.619e+02 3.085e+02 8.186e+02, threshold=5.238e+02, percent-clipped=3.0 2023-05-01 08:05:55,513 INFO [train.py:904] (5/8) Epoch 21, batch 8400, loss[loss=0.1631, simple_loss=0.2495, pruned_loss=0.03837, over 12243.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.279, pruned_loss=0.04876, over 3073537.71 frames. ], batch size: 246, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:06:07,769 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3609, 2.2072, 2.1399, 4.0796, 2.1460, 2.5507, 2.3128, 2.3760], device='cuda:5'), covar=tensor([0.1144, 0.3840, 0.3195, 0.0467, 0.4486, 0.2622, 0.3738, 0.3525], device='cuda:5'), in_proj_covar=tensor([0.0389, 0.0436, 0.0357, 0.0316, 0.0428, 0.0499, 0.0406, 0.0509], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:06:10,925 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:06:22,508 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:04,467 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:10,982 INFO [train.py:904] (5/8) Epoch 21, batch 8450, loss[loss=0.1576, simple_loss=0.2566, pruned_loss=0.02933, over 16212.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2761, pruned_loss=0.04668, over 3053309.34 frames. ], batch size: 165, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:07:36,212 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:45,085 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211473.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:51,643 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8178, 5.1412, 4.9104, 4.9190, 4.6776, 4.6730, 4.6018, 5.2090], device='cuda:5'), covar=tensor([0.1365, 0.0920, 0.1055, 0.0905, 0.0848, 0.1044, 0.1091, 0.0912], device='cuda:5'), in_proj_covar=tensor([0.0653, 0.0788, 0.0658, 0.0603, 0.0502, 0.0514, 0.0663, 0.0621], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:07:58,473 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5960, 3.5484, 3.5205, 2.7293, 3.4145, 2.0106, 3.1672, 2.8698], device='cuda:5'), covar=tensor([0.0143, 0.0134, 0.0194, 0.0236, 0.0110, 0.2481, 0.0137, 0.0255], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0149, 0.0192, 0.0172, 0.0169, 0.0201, 0.0179, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:08:18,592 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.143e+02 2.501e+02 3.093e+02 5.428e+02, threshold=5.001e+02, percent-clipped=1.0 2023-05-01 08:08:30,002 INFO [train.py:904] (5/8) Epoch 21, batch 8500, loss[loss=0.1708, simple_loss=0.2606, pruned_loss=0.04055, over 15275.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2727, pruned_loss=0.04433, over 3072482.16 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:09:54,387 INFO [train.py:904] (5/8) Epoch 21, batch 8550, loss[loss=0.1608, simple_loss=0.2479, pruned_loss=0.03685, over 12057.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2702, pruned_loss=0.04314, over 3062484.89 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:11:18,420 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.272e+02 2.648e+02 3.114e+02 4.500e+02, threshold=5.297e+02, percent-clipped=0.0 2023-05-01 08:11:25,869 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0868, 4.0269, 4.4005, 4.3995, 4.4148, 4.1615, 4.1558, 4.1195], device='cuda:5'), covar=tensor([0.0371, 0.0665, 0.0491, 0.0452, 0.0487, 0.0405, 0.0911, 0.0502], device='cuda:5'), in_proj_covar=tensor([0.0394, 0.0439, 0.0424, 0.0394, 0.0472, 0.0444, 0.0530, 0.0356], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 08:11:32,299 INFO [train.py:904] (5/8) Epoch 21, batch 8600, loss[loss=0.1855, simple_loss=0.2686, pruned_loss=0.05119, over 12489.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2705, pruned_loss=0.04193, over 3071246.15 frames. ], batch size: 250, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:12:10,870 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 08:12:48,791 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:13:09,696 INFO [train.py:904] (5/8) Epoch 21, batch 8650, loss[loss=0.1596, simple_loss=0.2527, pruned_loss=0.03327, over 12296.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2688, pruned_loss=0.04052, over 3053066.66 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:14:30,914 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:14:43,488 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.069e+02 2.535e+02 3.155e+02 6.979e+02, threshold=5.069e+02, percent-clipped=2.0 2023-05-01 08:14:57,324 INFO [train.py:904] (5/8) Epoch 21, batch 8700, loss[loss=0.1734, simple_loss=0.2743, pruned_loss=0.03623, over 16415.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2665, pruned_loss=0.03957, over 3067970.61 frames. ], batch size: 75, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:16:10,368 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:16:13,638 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8774, 4.3722, 4.4925, 3.2628, 3.7651, 4.3847, 3.9459, 2.8815], device='cuda:5'), covar=tensor([0.0461, 0.0040, 0.0035, 0.0315, 0.0107, 0.0087, 0.0071, 0.0368], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0080, 0.0081, 0.0132, 0.0095, 0.0105, 0.0092, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 08:16:31,425 INFO [train.py:904] (5/8) Epoch 21, batch 8750, loss[loss=0.1532, simple_loss=0.241, pruned_loss=0.03266, over 12363.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2663, pruned_loss=0.03933, over 3052027.51 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:16:48,500 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7381, 1.3218, 1.6603, 1.6315, 1.8087, 1.8863, 1.6508, 1.8743], device='cuda:5'), covar=tensor([0.0273, 0.0445, 0.0243, 0.0334, 0.0318, 0.0220, 0.0442, 0.0143], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0189, 0.0173, 0.0178, 0.0192, 0.0148, 0.0190, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:17:13,755 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:17:26,766 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 08:18:09,825 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.163e+02 2.564e+02 3.220e+02 5.959e+02, threshold=5.128e+02, percent-clipped=1.0 2023-05-01 08:18:23,766 INFO [train.py:904] (5/8) Epoch 21, batch 8800, loss[loss=0.1503, simple_loss=0.2464, pruned_loss=0.02706, over 17123.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2646, pruned_loss=0.03833, over 3060914.71 frames. ], batch size: 47, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:18:27,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9479, 5.2377, 5.0241, 5.0076, 4.7642, 4.7724, 4.6184, 5.3044], device='cuda:5'), covar=tensor([0.1166, 0.0836, 0.0917, 0.0823, 0.0788, 0.0907, 0.1157, 0.0828], device='cuda:5'), in_proj_covar=tensor([0.0645, 0.0779, 0.0652, 0.0596, 0.0496, 0.0508, 0.0657, 0.0615], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:20:07,844 INFO [train.py:904] (5/8) Epoch 21, batch 8850, loss[loss=0.1757, simple_loss=0.2771, pruned_loss=0.03715, over 15462.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2664, pruned_loss=0.03772, over 3037144.66 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:20:59,463 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6467, 3.4286, 3.8190, 1.8682, 3.9902, 4.0773, 3.1239, 3.1379], device='cuda:5'), covar=tensor([0.0676, 0.0237, 0.0166, 0.1132, 0.0071, 0.0107, 0.0365, 0.0370], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0103, 0.0092, 0.0133, 0.0076, 0.0118, 0.0124, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 08:21:18,250 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6097, 3.6696, 3.4310, 3.1092, 3.2364, 3.5737, 3.3523, 3.4033], device='cuda:5'), covar=tensor([0.0573, 0.0654, 0.0292, 0.0284, 0.0514, 0.0565, 0.1207, 0.0480], device='cuda:5'), in_proj_covar=tensor([0.0277, 0.0402, 0.0323, 0.0320, 0.0331, 0.0372, 0.0225, 0.0387], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:21:38,909 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.129e+02 2.519e+02 2.912e+02 9.426e+02, threshold=5.038e+02, percent-clipped=1.0 2023-05-01 08:21:54,332 INFO [train.py:904] (5/8) Epoch 21, batch 8900, loss[loss=0.1808, simple_loss=0.2692, pruned_loss=0.0462, over 12640.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2673, pruned_loss=0.03721, over 3058014.03 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:23:57,824 INFO [train.py:904] (5/8) Epoch 21, batch 8950, loss[loss=0.157, simple_loss=0.2548, pruned_loss=0.0296, over 16248.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2668, pruned_loss=0.03738, over 3077817.24 frames. ], batch size: 166, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:25:29,299 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 2.092e+02 2.341e+02 2.927e+02 4.703e+02, threshold=4.682e+02, percent-clipped=0.0 2023-05-01 08:25:46,911 INFO [train.py:904] (5/8) Epoch 21, batch 9000, loss[loss=0.141, simple_loss=0.2287, pruned_loss=0.02669, over 17074.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2632, pruned_loss=0.03569, over 3107357.14 frames. ], batch size: 53, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:25:46,912 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 08:25:57,434 INFO [train.py:938] (5/8) Epoch 21, validation: loss=0.1457, simple_loss=0.2498, pruned_loss=0.02077, over 944034.00 frames. 2023-05-01 08:25:57,436 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 08:27:20,398 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212042.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:27:39,218 INFO [train.py:904] (5/8) Epoch 21, batch 9050, loss[loss=0.1492, simple_loss=0.241, pruned_loss=0.02868, over 16838.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2642, pruned_loss=0.03627, over 3112161.71 frames. ], batch size: 102, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:28:14,819 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:28:58,994 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212090.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:29:10,427 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.192e+02 2.458e+02 2.901e+02 5.041e+02, threshold=4.916e+02, percent-clipped=1.0 2023-05-01 08:29:26,264 INFO [train.py:904] (5/8) Epoch 21, batch 9100, loss[loss=0.172, simple_loss=0.2595, pruned_loss=0.04224, over 12608.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2641, pruned_loss=0.03721, over 3097088.53 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:29:30,988 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212104.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:29:53,536 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:31:22,748 INFO [train.py:904] (5/8) Epoch 21, batch 9150, loss[loss=0.1673, simple_loss=0.2578, pruned_loss=0.03838, over 11981.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2645, pruned_loss=0.03712, over 3076548.28 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:31:26,615 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7990, 3.5854, 3.9348, 1.9163, 4.0530, 4.1863, 3.1456, 3.2052], device='cuda:5'), covar=tensor([0.0680, 0.0257, 0.0184, 0.1204, 0.0084, 0.0116, 0.0388, 0.0394], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0104, 0.0092, 0.0134, 0.0076, 0.0118, 0.0124, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 08:31:52,682 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212165.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:32:13,981 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 08:32:26,923 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9070, 2.7052, 2.7183, 1.9859, 2.5586, 2.8055, 2.6401, 1.8571], device='cuda:5'), covar=tensor([0.0434, 0.0076, 0.0070, 0.0383, 0.0140, 0.0093, 0.0101, 0.0459], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0080, 0.0081, 0.0132, 0.0095, 0.0104, 0.0091, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 08:32:57,285 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.079e+02 2.542e+02 3.144e+02 5.396e+02, threshold=5.084e+02, percent-clipped=1.0 2023-05-01 08:33:09,947 INFO [train.py:904] (5/8) Epoch 21, batch 9200, loss[loss=0.1735, simple_loss=0.2642, pruned_loss=0.04144, over 15374.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2605, pruned_loss=0.03644, over 3080130.56 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:37,257 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5946, 3.5705, 3.5265, 2.7940, 3.4655, 2.0186, 3.2122, 2.9042], device='cuda:5'), covar=tensor([0.0151, 0.0167, 0.0197, 0.0240, 0.0122, 0.2423, 0.0163, 0.0291], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0147, 0.0188, 0.0167, 0.0167, 0.0200, 0.0176, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:34:48,829 INFO [train.py:904] (5/8) Epoch 21, batch 9250, loss[loss=0.1518, simple_loss=0.2455, pruned_loss=0.02907, over 11721.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2609, pruned_loss=0.03689, over 3084000.93 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:51,569 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:34:52,159 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-05-01 08:36:25,640 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 2.162e+02 2.598e+02 3.159e+02 6.088e+02, threshold=5.195e+02, percent-clipped=3.0 2023-05-01 08:36:39,764 INFO [train.py:904] (5/8) Epoch 21, batch 9300, loss[loss=0.15, simple_loss=0.2317, pruned_loss=0.03415, over 12252.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2594, pruned_loss=0.03635, over 3061007.75 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:36:53,413 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-01 08:37:01,683 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0959, 4.1424, 4.4298, 4.4092, 4.4056, 4.1853, 4.1341, 4.1759], device='cuda:5'), covar=tensor([0.0400, 0.0864, 0.0550, 0.0586, 0.0628, 0.0513, 0.0921, 0.0490], device='cuda:5'), in_proj_covar=tensor([0.0386, 0.0430, 0.0419, 0.0388, 0.0465, 0.0437, 0.0518, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 08:37:07,748 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212314.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:38:22,842 INFO [train.py:904] (5/8) Epoch 21, batch 9350, loss[loss=0.1901, simple_loss=0.2806, pruned_loss=0.04979, over 15395.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2596, pruned_loss=0.03625, over 3078320.55 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:38:44,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1710, 2.3602, 2.0544, 2.2199, 2.7458, 2.4300, 2.6571, 2.9369], device='cuda:5'), covar=tensor([0.0148, 0.0438, 0.0534, 0.0470, 0.0296, 0.0416, 0.0210, 0.0252], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0226, 0.0218, 0.0218, 0.0226, 0.0225, 0.0222, 0.0219], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:39:26,822 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5890, 3.6660, 2.7632, 2.1679, 2.3337, 2.4463, 3.9100, 3.1548], device='cuda:5'), covar=tensor([0.2830, 0.0634, 0.1842, 0.2920, 0.2827, 0.1971, 0.0395, 0.1342], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0260, 0.0297, 0.0303, 0.0284, 0.0252, 0.0284, 0.0325], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 08:39:47,853 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 2.212e+02 2.571e+02 3.078e+02 5.703e+02, threshold=5.142e+02, percent-clipped=1.0 2023-05-01 08:40:02,997 INFO [train.py:904] (5/8) Epoch 21, batch 9400, loss[loss=0.1538, simple_loss=0.24, pruned_loss=0.03377, over 12554.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2596, pruned_loss=0.03604, over 3076133.83 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:05,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0220, 3.2878, 3.7455, 1.9919, 3.0086, 2.2735, 3.4473, 3.3847], device='cuda:5'), covar=tensor([0.0258, 0.0843, 0.0409, 0.2139, 0.0783, 0.0957, 0.0698, 0.1088], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0155, 0.0161, 0.0148, 0.0140, 0.0125, 0.0138, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 08:41:10,821 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-05-01 08:41:18,570 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:41:34,232 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7224, 2.6599, 1.9028, 2.8493, 2.1202, 2.8410, 2.1638, 2.4199], device='cuda:5'), covar=tensor([0.0329, 0.0378, 0.1415, 0.0351, 0.0759, 0.0583, 0.1288, 0.0685], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0153, 0.0172, 0.0208, 0.0196, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 08:41:42,929 INFO [train.py:904] (5/8) Epoch 21, batch 9450, loss[loss=0.1517, simple_loss=0.2398, pruned_loss=0.03178, over 16687.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2608, pruned_loss=0.03615, over 3045155.64 frames. ], batch size: 62, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:49,107 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6046, 2.9945, 3.3429, 1.9425, 2.7949, 2.1489, 3.1375, 3.1307], device='cuda:5'), covar=tensor([0.0280, 0.0830, 0.0466, 0.2132, 0.0860, 0.1070, 0.0691, 0.0959], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0155, 0.0161, 0.0148, 0.0140, 0.0125, 0.0138, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 08:41:58,681 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212460.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:43:11,247 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 1.974e+02 2.463e+02 3.006e+02 5.675e+02, threshold=4.927e+02, percent-clipped=1.0 2023-05-01 08:43:21,220 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:43:25,080 INFO [train.py:904] (5/8) Epoch 21, batch 9500, loss[loss=0.1777, simple_loss=0.2715, pruned_loss=0.04193, over 15208.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2596, pruned_loss=0.03551, over 3049950.14 frames. ], batch size: 191, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:43:53,865 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-01 08:44:01,552 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 08:45:02,486 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6315, 2.6103, 2.4407, 4.0535, 2.6202, 3.9939, 1.4902, 2.9563], device='cuda:5'), covar=tensor([0.1439, 0.0811, 0.1255, 0.0184, 0.0152, 0.0363, 0.1713, 0.0762], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0180, 0.0195, 0.0208, 0.0197, 0.0188], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 08:45:12,079 INFO [train.py:904] (5/8) Epoch 21, batch 9550, loss[loss=0.1659, simple_loss=0.2573, pruned_loss=0.03726, over 12169.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2602, pruned_loss=0.03604, over 3038106.82 frames. ], batch size: 250, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:45:39,005 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5104, 4.4795, 4.3229, 3.8192, 4.4108, 1.7145, 4.1441, 4.0574], device='cuda:5'), covar=tensor([0.0102, 0.0093, 0.0203, 0.0311, 0.0113, 0.2694, 0.0143, 0.0270], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0145, 0.0185, 0.0164, 0.0164, 0.0197, 0.0173, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:45:45,562 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 08:46:39,907 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 2.109e+02 2.597e+02 3.170e+02 5.717e+02, threshold=5.194e+02, percent-clipped=4.0 2023-05-01 08:46:50,965 INFO [train.py:904] (5/8) Epoch 21, batch 9600, loss[loss=0.2019, simple_loss=0.2958, pruned_loss=0.05397, over 15325.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2622, pruned_loss=0.03681, over 3064466.00 frames. ], batch size: 191, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:47:06,248 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212609.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:47:36,606 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0269, 4.1007, 3.9217, 3.6546, 3.6699, 4.0434, 3.6844, 3.8161], device='cuda:5'), covar=tensor([0.0550, 0.0579, 0.0315, 0.0287, 0.0673, 0.0516, 0.0939, 0.0605], device='cuda:5'), in_proj_covar=tensor([0.0275, 0.0397, 0.0322, 0.0316, 0.0327, 0.0368, 0.0223, 0.0384], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 08:48:37,661 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4515, 3.5220, 2.0836, 3.9452, 2.5602, 3.8889, 2.2756, 2.8351], device='cuda:5'), covar=tensor([0.0295, 0.0365, 0.1796, 0.0229, 0.0972, 0.0507, 0.1573, 0.0780], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0153, 0.0172, 0.0207, 0.0196, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 08:48:38,610 INFO [train.py:904] (5/8) Epoch 21, batch 9650, loss[loss=0.1724, simple_loss=0.2683, pruned_loss=0.03826, over 17010.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2637, pruned_loss=0.03721, over 3062231.65 frames. ], batch size: 109, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:50:11,919 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.547e+02 2.239e+02 2.609e+02 3.201e+02 8.250e+02, threshold=5.217e+02, percent-clipped=3.0 2023-05-01 08:50:20,730 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212698.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:50:28,040 INFO [train.py:904] (5/8) Epoch 21, batch 9700, loss[loss=0.17, simple_loss=0.2639, pruned_loss=0.03806, over 16805.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2624, pruned_loss=0.03702, over 3058227.66 frames. ], batch size: 124, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:50:40,287 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 08:50:46,027 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:51:36,324 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5160, 3.0332, 2.8909, 1.7651, 2.6267, 1.9721, 3.0463, 3.2147], device='cuda:5'), covar=tensor([0.0311, 0.0839, 0.0800, 0.2557, 0.1102, 0.1336, 0.0777, 0.0899], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0154, 0.0160, 0.0147, 0.0139, 0.0124, 0.0137, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 08:51:40,414 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5309, 3.6572, 2.6030, 2.1980, 2.2665, 2.2253, 3.7639, 3.0825], device='cuda:5'), covar=tensor([0.3038, 0.0673, 0.2042, 0.2935, 0.2946, 0.2371, 0.0486, 0.1304], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0260, 0.0298, 0.0303, 0.0284, 0.0251, 0.0284, 0.0324], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 08:52:10,279 INFO [train.py:904] (5/8) Epoch 21, batch 9750, loss[loss=0.1608, simple_loss=0.2563, pruned_loss=0.03265, over 16907.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2602, pruned_loss=0.03671, over 3061759.34 frames. ], batch size: 116, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:52:25,434 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:27,232 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212760.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:52:49,228 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:53,753 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212774.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:53:36,935 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 2.010e+02 2.340e+02 2.880e+02 4.703e+02, threshold=4.680e+02, percent-clipped=0.0 2023-05-01 08:53:37,578 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:53:48,986 INFO [train.py:904] (5/8) Epoch 21, batch 9800, loss[loss=0.1688, simple_loss=0.2728, pruned_loss=0.03238, over 16673.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2611, pruned_loss=0.03578, over 3067485.31 frames. ], batch size: 134, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:53:55,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2457, 4.2338, 4.0588, 3.5190, 4.1802, 1.6082, 3.9486, 3.7557], device='cuda:5'), covar=tensor([0.0095, 0.0091, 0.0169, 0.0252, 0.0090, 0.2956, 0.0126, 0.0267], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0145, 0.0185, 0.0163, 0.0164, 0.0197, 0.0174, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:54:01,431 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212808.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:54:31,961 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 08:54:51,611 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212835.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:55:31,554 INFO [train.py:904] (5/8) Epoch 21, batch 9850, loss[loss=0.1509, simple_loss=0.2533, pruned_loss=0.02424, over 16712.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2622, pruned_loss=0.03536, over 3076354.98 frames. ], batch size: 89, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:55:47,397 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1294, 2.3130, 2.0643, 2.2322, 2.6499, 2.3486, 2.6159, 2.8351], device='cuda:5'), covar=tensor([0.0140, 0.0444, 0.0517, 0.0433, 0.0296, 0.0443, 0.0196, 0.0281], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0225, 0.0218, 0.0217, 0.0225, 0.0224, 0.0221, 0.0218], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:56:18,584 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8177, 3.7574, 3.9139, 4.0054, 4.0795, 3.6954, 4.0860, 4.1322], device='cuda:5'), covar=tensor([0.1659, 0.1194, 0.1252, 0.0741, 0.0573, 0.1759, 0.0681, 0.0664], device='cuda:5'), in_proj_covar=tensor([0.0596, 0.0730, 0.0853, 0.0747, 0.0566, 0.0593, 0.0611, 0.0708], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 08:56:23,813 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5734, 2.5612, 2.4200, 3.6889, 2.1888, 3.7335, 1.5256, 2.7299], device='cuda:5'), covar=tensor([0.1488, 0.0800, 0.1223, 0.0151, 0.0098, 0.0372, 0.1683, 0.0841], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0170, 0.0191, 0.0181, 0.0196, 0.0209, 0.0198, 0.0189], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 08:56:30,504 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3987, 3.3384, 2.7052, 2.0458, 2.1669, 2.3025, 3.4641, 3.0050], device='cuda:5'), covar=tensor([0.2969, 0.0617, 0.1713, 0.3162, 0.2786, 0.2205, 0.0447, 0.1323], device='cuda:5'), in_proj_covar=tensor([0.0316, 0.0259, 0.0296, 0.0301, 0.0282, 0.0249, 0.0282, 0.0320], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 08:56:31,830 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1271, 2.5623, 2.6463, 1.9128, 2.8138, 2.8897, 2.4995, 2.5061], device='cuda:5'), covar=tensor([0.0654, 0.0241, 0.0214, 0.1004, 0.0111, 0.0222, 0.0452, 0.0391], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0103, 0.0090, 0.0133, 0.0076, 0.0117, 0.0123, 0.0123], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 08:56:33,836 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-05-01 08:57:07,914 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.176e+02 2.812e+02 3.349e+02 5.684e+02, threshold=5.624e+02, percent-clipped=7.0 2023-05-01 08:57:21,379 INFO [train.py:904] (5/8) Epoch 21, batch 9900, loss[loss=0.175, simple_loss=0.2897, pruned_loss=0.03015, over 16863.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2626, pruned_loss=0.0357, over 3048714.34 frames. ], batch size: 90, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:57:38,741 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:58:58,966 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-05-01 08:59:17,588 INFO [train.py:904] (5/8) Epoch 21, batch 9950, loss[loss=0.1797, simple_loss=0.2869, pruned_loss=0.03627, over 16867.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.265, pruned_loss=0.03624, over 3058468.96 frames. ], batch size: 124, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:59:29,579 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212957.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:00:58,654 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 2.203e+02 2.698e+02 3.794e+02 1.048e+03, threshold=5.396e+02, percent-clipped=2.0 2023-05-01 09:01:02,694 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2054, 5.2071, 5.0369, 4.5637, 4.7342, 5.1355, 5.0140, 4.6839], device='cuda:5'), covar=tensor([0.0557, 0.0541, 0.0331, 0.0322, 0.1001, 0.0459, 0.0272, 0.0711], device='cuda:5'), in_proj_covar=tensor([0.0273, 0.0396, 0.0320, 0.0314, 0.0325, 0.0366, 0.0221, 0.0381], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 09:01:17,081 INFO [train.py:904] (5/8) Epoch 21, batch 10000, loss[loss=0.1737, simple_loss=0.2736, pruned_loss=0.03691, over 16267.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2639, pruned_loss=0.036, over 3081168.85 frames. ], batch size: 166, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:01:22,049 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 09:02:58,012 INFO [train.py:904] (5/8) Epoch 21, batch 10050, loss[loss=0.171, simple_loss=0.2655, pruned_loss=0.03826, over 12172.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.264, pruned_loss=0.03591, over 3068554.95 frames. ], batch size: 249, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:03:02,987 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213054.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:03:20,727 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5686, 4.5813, 4.3490, 3.9204, 4.4659, 1.6005, 4.2260, 4.1719], device='cuda:5'), covar=tensor([0.0100, 0.0114, 0.0214, 0.0267, 0.0112, 0.2701, 0.0159, 0.0234], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0145, 0.0185, 0.0162, 0.0164, 0.0197, 0.0173, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:03:28,740 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:03:49,611 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 09:04:00,078 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.09 vs. limit=5.0 2023-05-01 09:04:21,051 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 2.063e+02 2.495e+02 3.076e+02 5.792e+02, threshold=4.990e+02, percent-clipped=2.0 2023-05-01 09:04:21,678 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0539, 4.0231, 3.9194, 3.2135, 3.8546, 1.7583, 3.6325, 3.5472], device='cuda:5'), covar=tensor([0.0135, 0.0163, 0.0177, 0.0280, 0.0130, 0.2594, 0.0149, 0.0264], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0145, 0.0184, 0.0162, 0.0164, 0.0196, 0.0173, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:04:21,683 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213095.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:04:32,159 INFO [train.py:904] (5/8) Epoch 21, batch 10100, loss[loss=0.1785, simple_loss=0.2589, pruned_loss=0.04903, over 12333.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2643, pruned_loss=0.03639, over 3076750.18 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:05:01,848 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5642, 4.8772, 4.6810, 4.6670, 4.3949, 4.3302, 4.2633, 4.9083], device='cuda:5'), covar=tensor([0.1199, 0.0831, 0.0967, 0.0805, 0.0768, 0.1340, 0.1177, 0.0923], device='cuda:5'), in_proj_covar=tensor([0.0635, 0.0772, 0.0639, 0.0585, 0.0491, 0.0502, 0.0646, 0.0604], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:05:27,306 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213130.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:31,796 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9503, 3.2194, 3.5531, 2.0505, 2.8949, 2.1718, 3.3806, 3.3799], device='cuda:5'), covar=tensor([0.0280, 0.0877, 0.0509, 0.2129, 0.0848, 0.1108, 0.0739, 0.1146], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0154, 0.0161, 0.0148, 0.0140, 0.0125, 0.0138, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:05:42,670 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:44,967 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7701, 3.8260, 2.4169, 4.4326, 2.8337, 4.3303, 2.6227, 3.0965], device='cuda:5'), covar=tensor([0.0302, 0.0364, 0.1632, 0.0174, 0.0965, 0.0420, 0.1470, 0.0798], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0168, 0.0187, 0.0151, 0.0170, 0.0205, 0.0195, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 09:05:51,230 INFO [train.py:904] (5/8) Epoch 21, batch 10150, loss[loss=0.1562, simple_loss=0.2442, pruned_loss=0.03408, over 12564.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2631, pruned_loss=0.03648, over 3054169.39 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:06:16,181 INFO [train.py:904] (5/8) Epoch 22, batch 0, loss[loss=0.2482, simple_loss=0.3122, pruned_loss=0.09214, over 16891.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3122, pruned_loss=0.09214, over 16891.00 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:06:16,181 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 09:06:23,635 INFO [train.py:938] (5/8) Epoch 22, validation: loss=0.1457, simple_loss=0.2489, pruned_loss=0.0212, over 944034.00 frames. 2023-05-01 09:06:23,636 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 09:07:26,344 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 2.320e+02 2.870e+02 3.515e+02 6.456e+02, threshold=5.740e+02, percent-clipped=5.0 2023-05-01 09:07:34,026 INFO [train.py:904] (5/8) Epoch 22, batch 50, loss[loss=0.1809, simple_loss=0.2824, pruned_loss=0.03967, over 17067.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2745, pruned_loss=0.05094, over 745515.97 frames. ], batch size: 50, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:08:41,854 INFO [train.py:904] (5/8) Epoch 22, batch 100, loss[loss=0.2234, simple_loss=0.3043, pruned_loss=0.07127, over 11717.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.267, pruned_loss=0.04634, over 1311438.22 frames. ], batch size: 246, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:09:44,157 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4401, 2.3082, 2.3013, 4.3260, 2.3544, 2.7379, 2.4075, 2.4715], device='cuda:5'), covar=tensor([0.1240, 0.3725, 0.3145, 0.0483, 0.4072, 0.2631, 0.3639, 0.3597], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0438, 0.0361, 0.0320, 0.0431, 0.0499, 0.0409, 0.0511], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:09:44,708 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.194e+02 2.643e+02 3.119e+02 6.629e+02, threshold=5.286e+02, percent-clipped=1.0 2023-05-01 09:09:51,952 INFO [train.py:904] (5/8) Epoch 22, batch 150, loss[loss=0.1849, simple_loss=0.2603, pruned_loss=0.05477, over 16279.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2658, pruned_loss=0.0459, over 1758406.04 frames. ], batch size: 165, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:00,680 INFO [train.py:904] (5/8) Epoch 22, batch 200, loss[loss=0.1851, simple_loss=0.2716, pruned_loss=0.04926, over 16308.00 frames. ], tot_loss[loss=0.177, simple_loss=0.264, pruned_loss=0.04502, over 2105726.86 frames. ], batch size: 165, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:02,349 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:11:20,467 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213367.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:00,919 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.291e+02 2.770e+02 3.341e+02 7.349e+02, threshold=5.539e+02, percent-clipped=1.0 2023-05-01 09:12:07,194 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213402.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:08,120 INFO [train.py:904] (5/8) Epoch 22, batch 250, loss[loss=0.1598, simple_loss=0.2484, pruned_loss=0.03557, over 17220.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2627, pruned_loss=0.04462, over 2379499.33 frames. ], batch size: 44, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:12:24,678 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8126, 4.1332, 3.0528, 2.3259, 2.6511, 2.5654, 4.3566, 3.5354], device='cuda:5'), covar=tensor([0.2783, 0.0554, 0.1797, 0.3003, 0.2743, 0.2067, 0.0443, 0.1442], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0264, 0.0301, 0.0307, 0.0288, 0.0255, 0.0288, 0.0328], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 09:12:25,570 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213415.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:46,388 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:48,241 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7003, 2.5272, 1.8358, 2.6925, 2.0790, 2.8149, 2.0899, 2.3203], device='cuda:5'), covar=tensor([0.0337, 0.0378, 0.1336, 0.0273, 0.0740, 0.0420, 0.1240, 0.0647], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0159, 0.0177, 0.0214, 0.0202, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:12:55,211 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6908, 3.7725, 2.2668, 4.3234, 2.9622, 4.2515, 2.5025, 3.0190], device='cuda:5'), covar=tensor([0.0322, 0.0413, 0.1693, 0.0281, 0.0825, 0.0566, 0.1572, 0.0853], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0159, 0.0177, 0.0214, 0.0202, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:13:17,490 INFO [train.py:904] (5/8) Epoch 22, batch 300, loss[loss=0.1768, simple_loss=0.249, pruned_loss=0.05234, over 16796.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2594, pruned_loss=0.04298, over 2583117.46 frames. ], batch size: 102, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:13:52,066 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:14:19,527 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.218e+02 2.593e+02 2.976e+02 4.668e+02, threshold=5.187e+02, percent-clipped=0.0 2023-05-01 09:14:25,248 INFO [train.py:904] (5/8) Epoch 22, batch 350, loss[loss=0.1781, simple_loss=0.2489, pruned_loss=0.05365, over 16540.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2577, pruned_loss=0.04259, over 2741096.99 frames. ], batch size: 68, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:14:30,346 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6178, 6.0517, 5.7598, 5.8188, 5.3449, 5.3947, 5.3866, 6.1881], device='cuda:5'), covar=tensor([0.1524, 0.0927, 0.1099, 0.0799, 0.0992, 0.0723, 0.1244, 0.0904], device='cuda:5'), in_proj_covar=tensor([0.0658, 0.0804, 0.0663, 0.0610, 0.0511, 0.0520, 0.0672, 0.0628], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:14:43,995 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-01 09:15:16,964 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8538, 2.8488, 2.6928, 4.8895, 3.9690, 4.3316, 1.7229, 3.3197], device='cuda:5'), covar=tensor([0.1317, 0.0788, 0.1192, 0.0176, 0.0182, 0.0378, 0.1553, 0.0704], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0172, 0.0192, 0.0185, 0.0198, 0.0211, 0.0200, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:15:34,186 INFO [train.py:904] (5/8) Epoch 22, batch 400, loss[loss=0.1585, simple_loss=0.2384, pruned_loss=0.03931, over 15886.00 frames. ], tot_loss[loss=0.17, simple_loss=0.256, pruned_loss=0.04201, over 2868037.67 frames. ], batch size: 35, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:15:40,969 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9998, 5.5960, 5.6510, 5.4143, 5.4634, 6.0713, 5.5214, 5.2370], device='cuda:5'), covar=tensor([0.1015, 0.2041, 0.2402, 0.2406, 0.2993, 0.1093, 0.1640, 0.2523], device='cuda:5'), in_proj_covar=tensor([0.0406, 0.0591, 0.0654, 0.0493, 0.0652, 0.0688, 0.0514, 0.0658], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 09:15:52,135 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:16:15,083 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8724, 3.9321, 4.1388, 4.1144, 4.1515, 3.9259, 3.9556, 3.9495], device='cuda:5'), covar=tensor([0.0366, 0.0704, 0.0422, 0.0448, 0.0469, 0.0477, 0.0759, 0.0503], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0443, 0.0427, 0.0401, 0.0475, 0.0450, 0.0531, 0.0363], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 09:16:36,632 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.265e+02 2.651e+02 3.318e+02 1.993e+03, threshold=5.302e+02, percent-clipped=3.0 2023-05-01 09:16:38,930 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6570, 2.7389, 2.5888, 4.9291, 3.9848, 4.3112, 1.5361, 3.2361], device='cuda:5'), covar=tensor([0.1468, 0.0860, 0.1352, 0.0232, 0.0213, 0.0437, 0.1733, 0.0782], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0172, 0.0192, 0.0185, 0.0199, 0.0212, 0.0200, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:16:43,118 INFO [train.py:904] (5/8) Epoch 22, batch 450, loss[loss=0.1595, simple_loss=0.2412, pruned_loss=0.03892, over 16788.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2547, pruned_loss=0.04213, over 2971627.05 frames. ], batch size: 102, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:17:16,850 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213627.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:17:53,040 INFO [train.py:904] (5/8) Epoch 22, batch 500, loss[loss=0.1621, simple_loss=0.2596, pruned_loss=0.03231, over 17147.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2536, pruned_loss=0.04125, over 3035509.95 frames. ], batch size: 48, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:18:54,887 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.187e+02 2.448e+02 2.877e+02 6.308e+02, threshold=4.896e+02, percent-clipped=1.0 2023-05-01 09:19:01,586 INFO [train.py:904] (5/8) Epoch 22, batch 550, loss[loss=0.1709, simple_loss=0.2498, pruned_loss=0.04595, over 16715.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2528, pruned_loss=0.04081, over 3103183.20 frames. ], batch size: 89, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:20:10,675 INFO [train.py:904] (5/8) Epoch 22, batch 600, loss[loss=0.1479, simple_loss=0.2233, pruned_loss=0.03627, over 16823.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2522, pruned_loss=0.04065, over 3158587.97 frames. ], batch size: 102, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:20:47,768 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 09:21:13,522 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.232e+02 2.565e+02 3.000e+02 5.415e+02, threshold=5.129e+02, percent-clipped=2.0 2023-05-01 09:21:21,585 INFO [train.py:904] (5/8) Epoch 22, batch 650, loss[loss=0.1663, simple_loss=0.2454, pruned_loss=0.04354, over 16767.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2507, pruned_loss=0.04001, over 3184992.07 frames. ], batch size: 134, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:21:54,593 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 09:22:30,260 INFO [train.py:904] (5/8) Epoch 22, batch 700, loss[loss=0.1697, simple_loss=0.2613, pruned_loss=0.03911, over 17033.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2501, pruned_loss=0.03962, over 3204367.36 frames. ], batch size: 50, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:23:15,778 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7470, 4.9992, 5.1022, 4.9513, 4.9393, 5.5813, 5.0555, 4.7764], device='cuda:5'), covar=tensor([0.1323, 0.2026, 0.2740, 0.2119, 0.2794, 0.1146, 0.1866, 0.2442], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0595, 0.0658, 0.0495, 0.0655, 0.0691, 0.0517, 0.0658], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 09:23:35,483 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.030e+02 2.447e+02 3.057e+02 6.587e+02, threshold=4.894e+02, percent-clipped=3.0 2023-05-01 09:23:41,797 INFO [train.py:904] (5/8) Epoch 22, batch 750, loss[loss=0.1627, simple_loss=0.2463, pruned_loss=0.0395, over 15656.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2511, pruned_loss=0.04014, over 3227843.68 frames. ], batch size: 191, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:23:55,202 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2374, 3.3334, 3.6216, 2.2055, 3.0386, 2.4389, 3.6827, 3.5831], device='cuda:5'), covar=tensor([0.0258, 0.0859, 0.0564, 0.1907, 0.0856, 0.0937, 0.0568, 0.1027], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0161, 0.0166, 0.0153, 0.0144, 0.0129, 0.0143, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:23:58,541 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0579, 5.6018, 5.7176, 5.4278, 5.5389, 6.1495, 5.5663, 5.3284], device='cuda:5'), covar=tensor([0.0906, 0.2050, 0.2281, 0.2178, 0.2916, 0.1127, 0.1603, 0.2224], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0596, 0.0658, 0.0496, 0.0656, 0.0692, 0.0518, 0.0658], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 09:24:08,403 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:24:26,174 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6994, 3.7017, 2.3791, 3.9858, 2.9721, 3.9210, 2.3809, 3.0028], device='cuda:5'), covar=tensor([0.0270, 0.0414, 0.1452, 0.0395, 0.0727, 0.0881, 0.1482, 0.0700], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0177, 0.0196, 0.0163, 0.0178, 0.0217, 0.0203, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:24:53,401 INFO [train.py:904] (5/8) Epoch 22, batch 800, loss[loss=0.1429, simple_loss=0.2298, pruned_loss=0.02799, over 16811.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.251, pruned_loss=0.03994, over 3237247.95 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:25:31,599 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1177, 5.0997, 4.9443, 4.4411, 4.6185, 5.0102, 4.9430, 4.6121], device='cuda:5'), covar=tensor([0.0594, 0.0614, 0.0342, 0.0385, 0.1146, 0.0514, 0.0365, 0.0804], device='cuda:5'), in_proj_covar=tensor([0.0291, 0.0422, 0.0339, 0.0337, 0.0347, 0.0392, 0.0233, 0.0407], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:25:56,459 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 09:25:56,895 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.094e+02 2.441e+02 2.760e+02 5.264e+02, threshold=4.882e+02, percent-clipped=1.0 2023-05-01 09:26:06,627 INFO [train.py:904] (5/8) Epoch 22, batch 850, loss[loss=0.1698, simple_loss=0.2661, pruned_loss=0.03675, over 16624.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2501, pruned_loss=0.03942, over 3257309.65 frames. ], batch size: 62, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:26:17,290 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7764, 2.3782, 2.2985, 3.3432, 2.6217, 3.5925, 1.5968, 2.7040], device='cuda:5'), covar=tensor([0.1307, 0.0810, 0.1218, 0.0222, 0.0162, 0.0401, 0.1552, 0.0845], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0173, 0.0193, 0.0188, 0.0202, 0.0215, 0.0202, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:27:17,280 INFO [train.py:904] (5/8) Epoch 22, batch 900, loss[loss=0.1944, simple_loss=0.2652, pruned_loss=0.06184, over 16846.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2491, pruned_loss=0.03924, over 3269767.14 frames. ], batch size: 116, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:28:19,802 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 1.995e+02 2.382e+02 2.694e+02 6.707e+02, threshold=4.763e+02, percent-clipped=1.0 2023-05-01 09:28:24,218 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9121, 5.2917, 5.0590, 5.0685, 4.7998, 4.7688, 4.7591, 5.3745], device='cuda:5'), covar=tensor([0.1373, 0.0926, 0.1019, 0.0888, 0.0912, 0.1102, 0.1161, 0.0942], device='cuda:5'), in_proj_covar=tensor([0.0676, 0.0821, 0.0679, 0.0626, 0.0525, 0.0532, 0.0690, 0.0646], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:28:27,536 INFO [train.py:904] (5/8) Epoch 22, batch 950, loss[loss=0.1508, simple_loss=0.232, pruned_loss=0.03482, over 15397.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2492, pruned_loss=0.03911, over 3286212.52 frames. ], batch size: 190, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:29:03,355 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214129.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:29:35,598 INFO [train.py:904] (5/8) Epoch 22, batch 1000, loss[loss=0.1577, simple_loss=0.2367, pruned_loss=0.03939, over 16862.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2489, pruned_loss=0.03924, over 3304997.26 frames. ], batch size: 102, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:30:15,713 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7063, 1.7842, 1.6175, 1.4958, 1.8860, 1.5949, 1.6301, 1.9039], device='cuda:5'), covar=tensor([0.0230, 0.0341, 0.0461, 0.0402, 0.0231, 0.0319, 0.0197, 0.0226], device='cuda:5'), in_proj_covar=tensor([0.0212, 0.0238, 0.0228, 0.0229, 0.0239, 0.0237, 0.0238, 0.0233], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:30:20,596 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7512, 4.7243, 4.6500, 4.1034, 4.7058, 1.8360, 4.4159, 4.3310], device='cuda:5'), covar=tensor([0.0142, 0.0125, 0.0193, 0.0369, 0.0112, 0.2724, 0.0186, 0.0246], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0155, 0.0197, 0.0174, 0.0175, 0.0208, 0.0186, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:30:29,153 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214190.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:30:39,708 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.141e+02 2.430e+02 2.890e+02 5.407e+02, threshold=4.860e+02, percent-clipped=2.0 2023-05-01 09:30:46,647 INFO [train.py:904] (5/8) Epoch 22, batch 1050, loss[loss=0.1723, simple_loss=0.2457, pruned_loss=0.04938, over 16669.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2489, pruned_loss=0.03928, over 3301952.89 frames. ], batch size: 134, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:30:51,615 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3119, 3.4968, 3.7470, 2.1125, 3.0317, 2.4546, 3.7485, 3.6937], device='cuda:5'), covar=tensor([0.0273, 0.0951, 0.0531, 0.2015, 0.0868, 0.0991, 0.0610, 0.1079], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0154, 0.0145, 0.0130, 0.0144, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:31:07,830 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:13,787 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:56,098 INFO [train.py:904] (5/8) Epoch 22, batch 1100, loss[loss=0.1711, simple_loss=0.2701, pruned_loss=0.03602, over 17255.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2481, pruned_loss=0.03872, over 3311701.07 frames. ], batch size: 52, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:32:19,742 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214270.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:31,293 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214279.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:43,279 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6567, 6.0768, 5.7935, 5.8850, 5.4460, 5.4978, 5.4569, 6.2037], device='cuda:5'), covar=tensor([0.1445, 0.0976, 0.1085, 0.0847, 0.0935, 0.0640, 0.1132, 0.0893], device='cuda:5'), in_proj_covar=tensor([0.0677, 0.0828, 0.0682, 0.0629, 0.0527, 0.0535, 0.0694, 0.0651], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:32:57,755 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.112e+02 2.466e+02 2.921e+02 8.908e+02, threshold=4.932e+02, percent-clipped=7.0 2023-05-01 09:33:03,846 INFO [train.py:904] (5/8) Epoch 22, batch 1150, loss[loss=0.1595, simple_loss=0.241, pruned_loss=0.03896, over 17023.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2477, pruned_loss=0.03849, over 3313693.39 frames. ], batch size: 41, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:34:15,638 INFO [train.py:904] (5/8) Epoch 22, batch 1200, loss[loss=0.1728, simple_loss=0.253, pruned_loss=0.04624, over 16741.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2465, pruned_loss=0.03839, over 3320274.03 frames. ], batch size: 124, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:34:23,490 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7044, 2.6225, 2.2972, 2.5966, 2.9249, 2.7390, 3.2168, 3.1434], device='cuda:5'), covar=tensor([0.0143, 0.0442, 0.0546, 0.0462, 0.0318, 0.0406, 0.0260, 0.0295], device='cuda:5'), in_proj_covar=tensor([0.0214, 0.0239, 0.0229, 0.0230, 0.0240, 0.0239, 0.0240, 0.0234], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:34:40,592 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5886, 3.9656, 4.1561, 2.8752, 3.5980, 4.1198, 3.7216, 2.3997], device='cuda:5'), covar=tensor([0.0514, 0.0322, 0.0053, 0.0380, 0.0133, 0.0113, 0.0111, 0.0487], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0083, 0.0084, 0.0134, 0.0098, 0.0108, 0.0094, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 09:35:18,110 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.085e+02 2.414e+02 2.922e+02 4.636e+02, threshold=4.828e+02, percent-clipped=0.0 2023-05-01 09:35:20,809 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 09:35:25,091 INFO [train.py:904] (5/8) Epoch 22, batch 1250, loss[loss=0.1794, simple_loss=0.2525, pruned_loss=0.05318, over 16885.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.247, pruned_loss=0.03893, over 3315533.23 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:35:38,861 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-01 09:36:35,043 INFO [train.py:904] (5/8) Epoch 22, batch 1300, loss[loss=0.1417, simple_loss=0.2281, pruned_loss=0.02767, over 16821.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2469, pruned_loss=0.03932, over 3318680.69 frames. ], batch size: 42, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:37:18,497 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214485.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:37:34,614 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.092e+02 2.483e+02 2.893e+02 4.900e+02, threshold=4.965e+02, percent-clipped=1.0 2023-05-01 09:37:42,346 INFO [train.py:904] (5/8) Epoch 22, batch 1350, loss[loss=0.1657, simple_loss=0.2448, pruned_loss=0.04329, over 16741.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2479, pruned_loss=0.03938, over 3318352.91 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:37:55,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8087, 4.0985, 3.1062, 2.3835, 2.7337, 2.6932, 4.3841, 3.5319], device='cuda:5'), covar=tensor([0.2747, 0.0576, 0.1688, 0.2612, 0.2519, 0.1980, 0.0377, 0.1214], device='cuda:5'), in_proj_covar=tensor([0.0326, 0.0271, 0.0307, 0.0313, 0.0297, 0.0261, 0.0295, 0.0338], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 09:38:48,218 INFO [train.py:904] (5/8) Epoch 22, batch 1400, loss[loss=0.1499, simple_loss=0.2398, pruned_loss=0.02997, over 17158.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2477, pruned_loss=0.03932, over 3327963.09 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:39:17,612 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214574.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:39:50,092 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.086e+02 2.334e+02 2.820e+02 4.887e+02, threshold=4.669e+02, percent-clipped=0.0 2023-05-01 09:39:57,061 INFO [train.py:904] (5/8) Epoch 22, batch 1450, loss[loss=0.1641, simple_loss=0.2569, pruned_loss=0.03567, over 16836.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2467, pruned_loss=0.03921, over 3327308.58 frames. ], batch size: 57, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:41:07,198 INFO [train.py:904] (5/8) Epoch 22, batch 1500, loss[loss=0.183, simple_loss=0.2777, pruned_loss=0.04419, over 17004.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2464, pruned_loss=0.03932, over 3316452.10 frames. ], batch size: 55, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:42:10,192 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.149e+02 2.472e+02 3.063e+02 5.864e+02, threshold=4.945e+02, percent-clipped=3.0 2023-05-01 09:42:16,367 INFO [train.py:904] (5/8) Epoch 22, batch 1550, loss[loss=0.1767, simple_loss=0.2576, pruned_loss=0.04795, over 16825.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2476, pruned_loss=0.04013, over 3319528.03 frames. ], batch size: 96, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:42:25,045 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 09:42:59,690 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8398, 4.0815, 2.2982, 4.6211, 3.1373, 4.6114, 2.4344, 3.3916], device='cuda:5'), covar=tensor([0.0385, 0.0337, 0.2000, 0.0292, 0.0858, 0.0421, 0.2002, 0.0723], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0181, 0.0199, 0.0168, 0.0181, 0.0223, 0.0207, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:43:06,125 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2757, 2.2415, 2.3550, 3.9953, 2.3071, 2.6324, 2.3106, 2.4351], device='cuda:5'), covar=tensor([0.1444, 0.3735, 0.2987, 0.0617, 0.3643, 0.2500, 0.3871, 0.2930], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0452, 0.0372, 0.0332, 0.0442, 0.0518, 0.0423, 0.0530], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:43:28,012 INFO [train.py:904] (5/8) Epoch 22, batch 1600, loss[loss=0.1913, simple_loss=0.2701, pruned_loss=0.05624, over 16737.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2495, pruned_loss=0.04127, over 3310379.95 frames. ], batch size: 134, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:44:12,582 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:44:32,066 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.192e+02 2.616e+02 3.002e+02 5.148e+02, threshold=5.232e+02, percent-clipped=2.0 2023-05-01 09:44:37,887 INFO [train.py:904] (5/8) Epoch 22, batch 1650, loss[loss=0.1734, simple_loss=0.2476, pruned_loss=0.04963, over 16788.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2503, pruned_loss=0.04134, over 3313364.30 frames. ], batch size: 124, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:45:02,883 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-01 09:45:20,795 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:45:39,447 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-01 09:45:47,643 INFO [train.py:904] (5/8) Epoch 22, batch 1700, loss[loss=0.1661, simple_loss=0.2437, pruned_loss=0.04424, over 16746.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.253, pruned_loss=0.04204, over 3298138.02 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:46:18,480 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:46:53,105 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.472e+02 2.965e+02 3.692e+02 1.718e+03, threshold=5.930e+02, percent-clipped=4.0 2023-05-01 09:46:58,562 INFO [train.py:904] (5/8) Epoch 22, batch 1750, loss[loss=0.1489, simple_loss=0.2436, pruned_loss=0.02717, over 17197.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2537, pruned_loss=0.0416, over 3303514.54 frames. ], batch size: 44, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:47:06,229 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0092, 3.0748, 1.9577, 3.2530, 2.3999, 3.2858, 2.1045, 2.5670], device='cuda:5'), covar=tensor([0.0364, 0.0447, 0.1585, 0.0366, 0.0848, 0.0712, 0.1455, 0.0799], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0181, 0.0198, 0.0168, 0.0181, 0.0223, 0.0206, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:47:25,707 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:48:06,729 INFO [train.py:904] (5/8) Epoch 22, batch 1800, loss[loss=0.1818, simple_loss=0.266, pruned_loss=0.04879, over 16896.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2557, pruned_loss=0.04191, over 3302269.84 frames. ], batch size: 90, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:48:31,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1478, 4.5810, 4.5896, 3.2988, 3.8790, 4.5331, 4.1077, 2.9695], device='cuda:5'), covar=tensor([0.0417, 0.0055, 0.0037, 0.0348, 0.0116, 0.0073, 0.0083, 0.0397], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0083, 0.0084, 0.0134, 0.0098, 0.0109, 0.0095, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 09:49:13,316 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.476e+02 2.061e+02 2.528e+02 2.863e+02 4.992e+02, threshold=5.055e+02, percent-clipped=0.0 2023-05-01 09:49:18,213 INFO [train.py:904] (5/8) Epoch 22, batch 1850, loss[loss=0.154, simple_loss=0.2543, pruned_loss=0.02684, over 17040.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2556, pruned_loss=0.04144, over 3316582.28 frames. ], batch size: 50, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:49:31,237 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4207, 2.3665, 2.3617, 4.2034, 2.3251, 2.7384, 2.3940, 2.5256], device='cuda:5'), covar=tensor([0.1308, 0.3734, 0.3180, 0.0541, 0.4272, 0.2624, 0.3734, 0.3734], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0453, 0.0373, 0.0332, 0.0442, 0.0519, 0.0424, 0.0530], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:49:33,379 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4137, 3.5034, 3.7255, 2.5182, 3.3694, 3.7418, 3.5253, 2.1617], device='cuda:5'), covar=tensor([0.0505, 0.0211, 0.0062, 0.0429, 0.0127, 0.0113, 0.0106, 0.0489], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0135, 0.0099, 0.0109, 0.0095, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 09:49:47,893 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8281, 2.7590, 2.5227, 2.6170, 3.0970, 2.8969, 3.4464, 3.3206], device='cuda:5'), covar=tensor([0.0166, 0.0429, 0.0493, 0.0471, 0.0293, 0.0394, 0.0253, 0.0287], device='cuda:5'), in_proj_covar=tensor([0.0218, 0.0241, 0.0232, 0.0232, 0.0242, 0.0241, 0.0244, 0.0237], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:50:27,233 INFO [train.py:904] (5/8) Epoch 22, batch 1900, loss[loss=0.1558, simple_loss=0.2501, pruned_loss=0.03078, over 17141.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2549, pruned_loss=0.04056, over 3316511.00 frames. ], batch size: 48, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:51:31,796 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.086e+02 2.418e+02 2.993e+02 8.338e+02, threshold=4.836e+02, percent-clipped=2.0 2023-05-01 09:51:36,124 INFO [train.py:904] (5/8) Epoch 22, batch 1950, loss[loss=0.1519, simple_loss=0.2475, pruned_loss=0.02815, over 17172.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2553, pruned_loss=0.04047, over 3310807.87 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:51:42,419 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6256, 4.4335, 4.6451, 4.8388, 4.9810, 4.4664, 4.9071, 4.9586], device='cuda:5'), covar=tensor([0.1635, 0.1274, 0.1525, 0.0722, 0.0553, 0.1118, 0.1223, 0.0783], device='cuda:5'), in_proj_covar=tensor([0.0658, 0.0810, 0.0947, 0.0822, 0.0619, 0.0651, 0.0673, 0.0782], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:52:44,790 INFO [train.py:904] (5/8) Epoch 22, batch 2000, loss[loss=0.1697, simple_loss=0.2651, pruned_loss=0.03713, over 12278.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2545, pruned_loss=0.04012, over 3300023.95 frames. ], batch size: 245, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:53:28,786 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3248, 4.1632, 4.3881, 4.5430, 4.6190, 4.2007, 4.4037, 4.6183], device='cuda:5'), covar=tensor([0.1528, 0.1177, 0.1230, 0.0630, 0.0604, 0.1277, 0.2262, 0.0728], device='cuda:5'), in_proj_covar=tensor([0.0659, 0.0813, 0.0951, 0.0823, 0.0620, 0.0654, 0.0675, 0.0784], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:53:48,737 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.173e+02 2.667e+02 3.241e+02 6.066e+02, threshold=5.334e+02, percent-clipped=3.0 2023-05-01 09:53:53,930 INFO [train.py:904] (5/8) Epoch 22, batch 2050, loss[loss=0.1574, simple_loss=0.2366, pruned_loss=0.03908, over 16775.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2548, pruned_loss=0.04071, over 3295724.88 frames. ], batch size: 89, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:54:31,517 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215230.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:55:01,200 INFO [train.py:904] (5/8) Epoch 22, batch 2100, loss[loss=0.1842, simple_loss=0.282, pruned_loss=0.04322, over 17076.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2555, pruned_loss=0.04109, over 3293753.27 frames. ], batch size: 53, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:55:54,371 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215291.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:56:04,631 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.033e+02 2.346e+02 2.893e+02 7.245e+02, threshold=4.692e+02, percent-clipped=2.0 2023-05-01 09:56:09,033 INFO [train.py:904] (5/8) Epoch 22, batch 2150, loss[loss=0.1773, simple_loss=0.2674, pruned_loss=0.04364, over 16571.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2573, pruned_loss=0.04228, over 3300260.04 frames. ], batch size: 68, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:56:39,688 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215325.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:56:51,000 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:57:18,610 INFO [train.py:904] (5/8) Epoch 22, batch 2200, loss[loss=0.1616, simple_loss=0.2482, pruned_loss=0.0375, over 16773.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2571, pruned_loss=0.04245, over 3291622.76 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:57:52,439 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9493, 2.9148, 2.5921, 4.7786, 3.8291, 4.2048, 1.8737, 3.0657], device='cuda:5'), covar=tensor([0.1165, 0.0727, 0.1161, 0.0216, 0.0231, 0.0469, 0.1381, 0.0794], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0191, 0.0204, 0.0217, 0.0202, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:58:02,284 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215386.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 09:58:14,345 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215394.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 09:58:20,110 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4504, 5.4390, 5.1969, 4.7263, 5.2433, 2.2106, 5.0051, 5.1793], device='cuda:5'), covar=tensor([0.0065, 0.0067, 0.0181, 0.0362, 0.0099, 0.2389, 0.0122, 0.0163], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0157, 0.0200, 0.0178, 0.0178, 0.0209, 0.0189, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 09:58:20,775 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.171e+02 2.606e+02 3.111e+02 9.974e+02, threshold=5.211e+02, percent-clipped=4.0 2023-05-01 09:58:22,749 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5652, 3.3040, 3.7635, 1.9492, 3.8438, 3.9232, 3.0857, 2.8243], device='cuda:5'), covar=tensor([0.0825, 0.0278, 0.0181, 0.1220, 0.0103, 0.0210, 0.0409, 0.0477], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0140, 0.0081, 0.0128, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 09:58:24,674 INFO [train.py:904] (5/8) Epoch 22, batch 2250, loss[loss=0.1436, simple_loss=0.2274, pruned_loss=0.02993, over 17005.00 frames. ], tot_loss[loss=0.172, simple_loss=0.258, pruned_loss=0.04297, over 3287497.96 frames. ], batch size: 41, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:59:27,052 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 09:59:36,626 INFO [train.py:904] (5/8) Epoch 22, batch 2300, loss[loss=0.1534, simple_loss=0.2494, pruned_loss=0.02863, over 17200.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2583, pruned_loss=0.04303, over 3298661.35 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:00:42,883 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.161e+02 2.569e+02 3.134e+02 6.590e+02, threshold=5.137e+02, percent-clipped=2.0 2023-05-01 10:00:46,514 INFO [train.py:904] (5/8) Epoch 22, batch 2350, loss[loss=0.1592, simple_loss=0.2481, pruned_loss=0.03519, over 17180.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.258, pruned_loss=0.04322, over 3293077.12 frames. ], batch size: 45, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:00:46,946 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:00:55,720 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4233, 5.3846, 5.1230, 4.6511, 5.1909, 2.2270, 4.9272, 5.0705], device='cuda:5'), covar=tensor([0.0083, 0.0074, 0.0222, 0.0363, 0.0101, 0.2474, 0.0131, 0.0192], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0158, 0.0200, 0.0178, 0.0179, 0.0209, 0.0190, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:01:03,316 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4481, 2.4528, 2.4359, 4.4514, 2.3300, 2.7508, 2.4294, 2.6181], device='cuda:5'), covar=tensor([0.1276, 0.3591, 0.3116, 0.0456, 0.4200, 0.2646, 0.3847, 0.3636], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0453, 0.0373, 0.0332, 0.0440, 0.0520, 0.0423, 0.0530], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:01:29,706 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 10:01:48,887 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215548.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:01:55,011 INFO [train.py:904] (5/8) Epoch 22, batch 2400, loss[loss=0.1619, simple_loss=0.255, pruned_loss=0.0344, over 17172.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2591, pruned_loss=0.04309, over 3303017.28 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:02:11,550 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215564.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:12,690 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215565.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:41,712 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215586.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:02:59,710 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 2.111e+02 2.344e+02 3.035e+02 7.498e+02, threshold=4.688e+02, percent-clipped=4.0 2023-05-01 10:03:04,616 INFO [train.py:904] (5/8) Epoch 22, batch 2450, loss[loss=0.1542, simple_loss=0.2395, pruned_loss=0.0344, over 16905.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.26, pruned_loss=0.04321, over 3289186.66 frames. ], batch size: 96, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:03:12,008 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215609.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:03:31,911 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 10:03:35,708 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215626.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:04:11,704 INFO [train.py:904] (5/8) Epoch 22, batch 2500, loss[loss=0.1634, simple_loss=0.245, pruned_loss=0.04096, over 16758.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2584, pruned_loss=0.04217, over 3299060.36 frames. ], batch size: 89, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:04:12,298 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 10:04:51,431 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215681.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:04:52,706 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7654, 2.8211, 2.7989, 5.0148, 3.9236, 4.4442, 1.7303, 3.1585], device='cuda:5'), covar=tensor([0.1428, 0.0864, 0.1152, 0.0243, 0.0242, 0.0381, 0.1654, 0.0822], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0191, 0.0204, 0.0217, 0.0202, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 10:04:55,633 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4990, 4.4533, 4.4232, 3.9595, 4.4273, 1.7729, 4.1856, 4.0057], device='cuda:5'), covar=tensor([0.0123, 0.0102, 0.0167, 0.0279, 0.0100, 0.2763, 0.0137, 0.0221], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0158, 0.0201, 0.0179, 0.0179, 0.0210, 0.0190, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:05:02,745 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215689.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:05:15,512 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.601e+02 2.254e+02 2.666e+02 2.993e+02 6.188e+02, threshold=5.333e+02, percent-clipped=1.0 2023-05-01 10:05:20,575 INFO [train.py:904] (5/8) Epoch 22, batch 2550, loss[loss=0.2246, simple_loss=0.307, pruned_loss=0.07114, over 11807.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2592, pruned_loss=0.04293, over 3287612.84 frames. ], batch size: 246, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:06:09,026 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8645, 3.0486, 3.2541, 2.1258, 2.8238, 2.2411, 3.4028, 3.3480], device='cuda:5'), covar=tensor([0.0251, 0.0993, 0.0583, 0.1951, 0.0870, 0.1032, 0.0588, 0.0936], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0165, 0.0167, 0.0155, 0.0146, 0.0131, 0.0145, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 10:06:30,389 INFO [train.py:904] (5/8) Epoch 22, batch 2600, loss[loss=0.1715, simple_loss=0.2463, pruned_loss=0.04832, over 16840.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2584, pruned_loss=0.04233, over 3304022.57 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:06:34,259 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7334, 3.9404, 2.5690, 4.5411, 2.9701, 4.4794, 2.6541, 3.2242], device='cuda:5'), covar=tensor([0.0339, 0.0412, 0.1553, 0.0269, 0.0940, 0.0573, 0.1486, 0.0754], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0167, 0.0179, 0.0222, 0.0204, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 10:07:36,117 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.223e+02 2.543e+02 2.941e+02 8.287e+02, threshold=5.085e+02, percent-clipped=5.0 2023-05-01 10:07:40,000 INFO [train.py:904] (5/8) Epoch 22, batch 2650, loss[loss=0.1668, simple_loss=0.2588, pruned_loss=0.03734, over 16838.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2596, pruned_loss=0.04234, over 3313035.89 frames. ], batch size: 42, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:08:16,616 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215829.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:08:17,873 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3651, 2.1142, 2.2906, 4.1288, 2.0905, 2.3954, 2.3004, 2.3080], device='cuda:5'), covar=tensor([0.1680, 0.4652, 0.3557, 0.0754, 0.5713, 0.3656, 0.4115, 0.4781], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0452, 0.0373, 0.0332, 0.0440, 0.0521, 0.0423, 0.0531], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:08:48,057 INFO [train.py:904] (5/8) Epoch 22, batch 2700, loss[loss=0.147, simple_loss=0.2341, pruned_loss=0.03, over 16775.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2599, pruned_loss=0.04123, over 3324689.64 frames. ], batch size: 39, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:08:57,517 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:21,872 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0878, 3.8495, 3.8583, 4.2554, 4.3475, 3.9597, 4.1761, 4.3148], device='cuda:5'), covar=tensor([0.1629, 0.1544, 0.2074, 0.0975, 0.0926, 0.1773, 0.2598, 0.1213], device='cuda:5'), in_proj_covar=tensor([0.0660, 0.0819, 0.0954, 0.0827, 0.0627, 0.0659, 0.0677, 0.0786], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:09:34,836 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215886.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:09:39,329 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:53,925 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.142e+02 2.465e+02 2.972e+02 5.160e+02, threshold=4.929e+02, percent-clipped=1.0 2023-05-01 10:09:57,375 INFO [train.py:904] (5/8) Epoch 22, batch 2750, loss[loss=0.1502, simple_loss=0.2438, pruned_loss=0.02827, over 17173.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2592, pruned_loss=0.04072, over 3329935.82 frames. ], batch size: 46, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:09:59,453 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215904.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:10:21,946 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:10:38,794 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=215934.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:11:05,013 INFO [train.py:904] (5/8) Epoch 22, batch 2800, loss[loss=0.1579, simple_loss=0.2479, pruned_loss=0.034, over 17097.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2591, pruned_loss=0.04081, over 3336526.84 frames. ], batch size: 47, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:11:30,762 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8916, 4.6661, 4.9091, 5.1379, 5.3476, 4.6737, 5.2916, 5.3249], device='cuda:5'), covar=tensor([0.1926, 0.1417, 0.1937, 0.0816, 0.0620, 0.1107, 0.0666, 0.0651], device='cuda:5'), in_proj_covar=tensor([0.0663, 0.0824, 0.0959, 0.0831, 0.0629, 0.0661, 0.0680, 0.0790], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:11:42,425 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215981.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:11:53,888 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215989.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:12:07,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.162e+02 2.449e+02 3.092e+02 5.987e+02, threshold=4.899e+02, percent-clipped=3.0 2023-05-01 10:12:11,324 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7322, 3.7676, 2.4540, 4.2231, 2.9142, 4.1527, 2.5344, 3.1245], device='cuda:5'), covar=tensor([0.0291, 0.0400, 0.1446, 0.0297, 0.0802, 0.0606, 0.1353, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0167, 0.0179, 0.0222, 0.0204, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 10:12:14,967 INFO [train.py:904] (5/8) Epoch 22, batch 2850, loss[loss=0.1849, simple_loss=0.27, pruned_loss=0.04989, over 16200.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2588, pruned_loss=0.04084, over 3327574.39 frames. ], batch size: 165, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:12:20,750 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1426, 2.3112, 2.7975, 3.0735, 2.9865, 3.6536, 2.5593, 3.5274], device='cuda:5'), covar=tensor([0.0252, 0.0472, 0.0317, 0.0327, 0.0308, 0.0191, 0.0450, 0.0176], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0196, 0.0182, 0.0187, 0.0199, 0.0157, 0.0198, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:12:51,610 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216028.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:12:52,548 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216029.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:13:03,384 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 10:13:04,676 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:13:26,230 INFO [train.py:904] (5/8) Epoch 22, batch 2900, loss[loss=0.1483, simple_loss=0.2286, pruned_loss=0.034, over 16871.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.257, pruned_loss=0.04039, over 3330125.87 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:13:27,852 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 10:13:55,654 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 10:14:08,132 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-01 10:14:17,786 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:14:31,569 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.264e+02 2.773e+02 3.437e+02 6.159e+02, threshold=5.547e+02, percent-clipped=1.0 2023-05-01 10:14:35,748 INFO [train.py:904] (5/8) Epoch 22, batch 2950, loss[loss=0.1455, simple_loss=0.2296, pruned_loss=0.0307, over 16766.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2565, pruned_loss=0.04064, over 3332415.15 frames. ], batch size: 39, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:14:36,379 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5591, 3.6542, 4.0928, 2.3601, 3.3119, 2.5323, 4.0337, 3.8640], device='cuda:5'), covar=tensor([0.0252, 0.0915, 0.0499, 0.1860, 0.0775, 0.0990, 0.0567, 0.1036], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0165, 0.0167, 0.0154, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 10:15:45,385 INFO [train.py:904] (5/8) Epoch 22, batch 3000, loss[loss=0.1722, simple_loss=0.2477, pruned_loss=0.04829, over 16862.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2571, pruned_loss=0.04132, over 3326323.35 frames. ], batch size: 116, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:15:45,386 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 10:15:52,409 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0950, 2.8679, 3.0173, 2.3197, 2.8855, 3.0891, 3.0154, 2.1139], device='cuda:5'), covar=tensor([0.0410, 0.0101, 0.0071, 0.0329, 0.0132, 0.0104, 0.0088, 0.0401], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0084, 0.0084, 0.0133, 0.0099, 0.0109, 0.0094, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 10:15:54,108 INFO [train.py:938] (5/8) Epoch 22, validation: loss=0.1347, simple_loss=0.2399, pruned_loss=0.0148, over 944034.00 frames. 2023-05-01 10:15:54,109 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 10:16:02,720 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:16:38,787 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:16:59,840 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.146e+02 2.550e+02 3.086e+02 1.158e+03, threshold=5.100e+02, percent-clipped=1.0 2023-05-01 10:17:03,747 INFO [train.py:904] (5/8) Epoch 22, batch 3050, loss[loss=0.1677, simple_loss=0.2494, pruned_loss=0.04296, over 16573.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2571, pruned_loss=0.04134, over 3322317.58 frames. ], batch size: 68, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:17:05,284 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216204.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:17:09,678 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216207.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:13,091 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 10:17:29,315 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216221.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:18:10,828 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216252.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:18:12,233 INFO [train.py:904] (5/8) Epoch 22, batch 3100, loss[loss=0.1928, simple_loss=0.2717, pruned_loss=0.05692, over 15489.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2571, pruned_loss=0.04191, over 3330650.37 frames. ], batch size: 190, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:18:33,928 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216269.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:19:02,549 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-05-01 10:19:16,929 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.355e+02 2.131e+02 2.523e+02 3.440e+02 6.792e+02, threshold=5.047e+02, percent-clipped=4.0 2023-05-01 10:19:21,103 INFO [train.py:904] (5/8) Epoch 22, batch 3150, loss[loss=0.182, simple_loss=0.2541, pruned_loss=0.05495, over 16880.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2561, pruned_loss=0.04187, over 3327489.63 frames. ], batch size: 109, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:19:25,766 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 10:20:29,467 INFO [train.py:904] (5/8) Epoch 22, batch 3200, loss[loss=0.1737, simple_loss=0.2691, pruned_loss=0.03919, over 17119.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2558, pruned_loss=0.04169, over 3326662.15 frames. ], batch size: 53, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:21:13,934 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216384.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:21:36,412 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.182e+02 2.419e+02 2.918e+02 5.281e+02, threshold=4.837e+02, percent-clipped=1.0 2023-05-01 10:21:40,330 INFO [train.py:904] (5/8) Epoch 22, batch 3250, loss[loss=0.2093, simple_loss=0.2863, pruned_loss=0.06612, over 16864.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2562, pruned_loss=0.04254, over 3314178.22 frames. ], batch size: 116, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:22:52,230 INFO [train.py:904] (5/8) Epoch 22, batch 3300, loss[loss=0.1496, simple_loss=0.2355, pruned_loss=0.03185, over 16707.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2572, pruned_loss=0.04273, over 3323246.55 frames. ], batch size: 37, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:23:36,594 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216485.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:23:54,280 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2023-05-01 10:23:56,628 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 2.110e+02 2.633e+02 3.149e+02 8.538e+02, threshold=5.267e+02, percent-clipped=2.0 2023-05-01 10:24:00,078 INFO [train.py:904] (5/8) Epoch 22, batch 3350, loss[loss=0.1628, simple_loss=0.2588, pruned_loss=0.0334, over 17115.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.258, pruned_loss=0.04256, over 3323320.23 frames. ], batch size: 48, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:24:05,850 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0576, 5.4467, 5.2433, 5.2262, 4.9441, 4.8182, 4.8587, 5.5679], device='cuda:5'), covar=tensor([0.1363, 0.0915, 0.0920, 0.0853, 0.0911, 0.0955, 0.1215, 0.0885], device='cuda:5'), in_proj_covar=tensor([0.0694, 0.0851, 0.0701, 0.0645, 0.0538, 0.0544, 0.0712, 0.0666], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:24:42,589 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:25:10,650 INFO [train.py:904] (5/8) Epoch 22, batch 3400, loss[loss=0.1358, simple_loss=0.2205, pruned_loss=0.02553, over 16962.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2583, pruned_loss=0.04279, over 3319630.74 frames. ], batch size: 41, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:26:02,832 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:26:15,742 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.115e+02 2.508e+02 2.941e+02 4.033e+02, threshold=5.017e+02, percent-clipped=0.0 2023-05-01 10:26:19,665 INFO [train.py:904] (5/8) Epoch 22, batch 3450, loss[loss=0.1476, simple_loss=0.2261, pruned_loss=0.0346, over 16883.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2563, pruned_loss=0.04208, over 3329377.98 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:26:27,145 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6720, 3.8635, 2.3555, 4.3883, 2.9053, 4.3514, 2.5923, 3.1583], device='cuda:5'), covar=tensor([0.0338, 0.0384, 0.1682, 0.0287, 0.0902, 0.0543, 0.1491, 0.0817], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0181, 0.0197, 0.0170, 0.0180, 0.0225, 0.0205, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 10:27:17,027 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 10:27:29,271 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216652.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:27:29,984 INFO [train.py:904] (5/8) Epoch 22, batch 3500, loss[loss=0.1922, simple_loss=0.2702, pruned_loss=0.05706, over 11693.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2555, pruned_loss=0.04195, over 3318419.20 frames. ], batch size: 246, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:27:32,791 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216655.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:28:12,653 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216684.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:17,063 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3616, 5.2483, 5.1688, 3.9821, 5.2293, 1.9660, 4.7863, 4.9661], device='cuda:5'), covar=tensor([0.0188, 0.0150, 0.0270, 0.0741, 0.0148, 0.3460, 0.0217, 0.0299], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0161, 0.0205, 0.0183, 0.0183, 0.0213, 0.0194, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:28:35,766 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 1.992e+02 2.337e+02 2.852e+02 6.760e+02, threshold=4.673e+02, percent-clipped=2.0 2023-05-01 10:28:39,288 INFO [train.py:904] (5/8) Epoch 22, batch 3550, loss[loss=0.1604, simple_loss=0.2375, pruned_loss=0.04163, over 16440.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2544, pruned_loss=0.04157, over 3317324.44 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:28:42,504 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:57,373 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216716.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:29:20,540 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216732.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:29:32,798 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9668, 4.2634, 4.2954, 3.2095, 3.6619, 4.2244, 3.8168, 2.5488], device='cuda:5'), covar=tensor([0.0463, 0.0065, 0.0051, 0.0345, 0.0133, 0.0121, 0.0103, 0.0470], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0133, 0.0099, 0.0110, 0.0095, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 10:29:49,496 INFO [train.py:904] (5/8) Epoch 22, batch 3600, loss[loss=0.1471, simple_loss=0.2315, pruned_loss=0.03134, over 16842.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2537, pruned_loss=0.04122, over 3319335.14 frames. ], batch size: 42, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:30:08,382 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216766.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:30:25,234 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5422, 3.5000, 3.7466, 2.6835, 3.4617, 3.8316, 3.5028, 2.1596], device='cuda:5'), covar=tensor([0.0465, 0.0185, 0.0061, 0.0371, 0.0112, 0.0101, 0.0103, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0133, 0.0099, 0.0110, 0.0095, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 10:30:56,337 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2503, 5.7495, 5.8755, 5.6153, 5.6554, 6.2364, 5.7797, 5.4792], device='cuda:5'), covar=tensor([0.0890, 0.1893, 0.2344, 0.2178, 0.2906, 0.1006, 0.1343, 0.2355], device='cuda:5'), in_proj_covar=tensor([0.0422, 0.0611, 0.0671, 0.0507, 0.0679, 0.0707, 0.0527, 0.0678], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 10:31:00,617 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.119e+02 2.478e+02 2.922e+02 5.214e+02, threshold=4.955e+02, percent-clipped=3.0 2023-05-01 10:31:03,555 INFO [train.py:904] (5/8) Epoch 22, batch 3650, loss[loss=0.1563, simple_loss=0.2553, pruned_loss=0.02867, over 17265.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2526, pruned_loss=0.04151, over 3301728.45 frames. ], batch size: 52, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:31:32,863 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1429, 2.1735, 2.2965, 3.7477, 2.2168, 2.4401, 2.2560, 2.3328], device='cuda:5'), covar=tensor([0.1536, 0.3602, 0.3012, 0.0652, 0.3832, 0.2862, 0.3902, 0.3284], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0453, 0.0374, 0.0335, 0.0442, 0.0524, 0.0426, 0.0533], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:32:08,764 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7809, 2.9965, 3.2365, 2.0339, 2.8086, 2.1436, 3.4401, 3.3736], device='cuda:5'), covar=tensor([0.0242, 0.0877, 0.0613, 0.1943, 0.0863, 0.1055, 0.0507, 0.0816], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0166, 0.0167, 0.0155, 0.0146, 0.0131, 0.0145, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 10:32:18,427 INFO [train.py:904] (5/8) Epoch 22, batch 3700, loss[loss=0.1745, simple_loss=0.2573, pruned_loss=0.04588, over 15393.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2511, pruned_loss=0.04277, over 3289098.62 frames. ], batch size: 190, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:32:31,704 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 10:33:31,446 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.238e+02 2.459e+02 3.071e+02 7.166e+02, threshold=4.918e+02, percent-clipped=1.0 2023-05-01 10:33:32,676 INFO [train.py:904] (5/8) Epoch 22, batch 3750, loss[loss=0.193, simple_loss=0.2777, pruned_loss=0.05412, over 16613.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2517, pruned_loss=0.04417, over 3271610.30 frames. ], batch size: 57, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:34:36,647 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216947.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:34:41,902 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3712, 4.6827, 4.4666, 4.5071, 4.2271, 4.1330, 4.1879, 4.7362], device='cuda:5'), covar=tensor([0.1277, 0.0848, 0.1060, 0.0863, 0.0813, 0.1594, 0.1175, 0.0834], device='cuda:5'), in_proj_covar=tensor([0.0698, 0.0854, 0.0703, 0.0647, 0.0541, 0.0548, 0.0716, 0.0668], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:34:44,494 INFO [train.py:904] (5/8) Epoch 22, batch 3800, loss[loss=0.1725, simple_loss=0.2578, pruned_loss=0.04363, over 15596.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2529, pruned_loss=0.04504, over 3263889.85 frames. ], batch size: 190, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:35:03,035 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:35:28,105 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.98 vs. limit=5.0 2023-05-01 10:35:55,034 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.325e+02 2.685e+02 3.161e+02 5.857e+02, threshold=5.370e+02, percent-clipped=3.0 2023-05-01 10:35:56,829 INFO [train.py:904] (5/8) Epoch 22, batch 3850, loss[loss=0.1619, simple_loss=0.2404, pruned_loss=0.04168, over 16769.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2528, pruned_loss=0.04559, over 3265412.93 frames. ], batch size: 83, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:36:08,462 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217011.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:36:29,883 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:37:09,502 INFO [train.py:904] (5/8) Epoch 22, batch 3900, loss[loss=0.1665, simple_loss=0.2437, pruned_loss=0.04471, over 16454.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.252, pruned_loss=0.04578, over 3265859.43 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:37:22,081 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217061.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:37:28,060 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 10:38:21,567 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.294e+02 2.626e+02 3.091e+02 8.707e+02, threshold=5.252e+02, percent-clipped=1.0 2023-05-01 10:38:22,862 INFO [train.py:904] (5/8) Epoch 22, batch 3950, loss[loss=0.1625, simple_loss=0.2394, pruned_loss=0.04283, over 16459.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2516, pruned_loss=0.04603, over 3268830.11 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:39:17,922 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-01 10:39:35,760 INFO [train.py:904] (5/8) Epoch 22, batch 4000, loss[loss=0.2098, simple_loss=0.2899, pruned_loss=0.06487, over 12218.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2521, pruned_loss=0.04666, over 3266141.19 frames. ], batch size: 246, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:40:48,054 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 1.995e+02 2.386e+02 2.991e+02 7.237e+02, threshold=4.771e+02, percent-clipped=2.0 2023-05-01 10:40:49,970 INFO [train.py:904] (5/8) Epoch 22, batch 4050, loss[loss=0.1741, simple_loss=0.2588, pruned_loss=0.04469, over 16350.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2533, pruned_loss=0.04639, over 3245211.12 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:41:51,965 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3288, 1.7072, 2.1514, 2.2747, 2.4802, 2.6959, 1.8653, 2.5438], device='cuda:5'), covar=tensor([0.0254, 0.0525, 0.0330, 0.0399, 0.0340, 0.0202, 0.0572, 0.0141], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0196, 0.0182, 0.0188, 0.0200, 0.0157, 0.0199, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:41:55,379 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:42:04,924 INFO [train.py:904] (5/8) Epoch 22, batch 4100, loss[loss=0.191, simple_loss=0.2787, pruned_loss=0.05166, over 16636.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2554, pruned_loss=0.04608, over 3258176.46 frames. ], batch size: 57, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:43:11,028 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217295.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:43:21,263 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 2.082e+02 2.531e+02 2.908e+02 7.878e+02, threshold=5.062e+02, percent-clipped=6.0 2023-05-01 10:43:23,210 INFO [train.py:904] (5/8) Epoch 22, batch 4150, loss[loss=0.2137, simple_loss=0.3052, pruned_loss=0.06109, over 16291.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2624, pruned_loss=0.0481, over 3258133.73 frames. ], batch size: 165, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:43:36,156 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217311.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:43:52,135 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:44:38,788 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4914, 5.7120, 5.4682, 5.5621, 5.1989, 5.0282, 5.1770, 5.8594], device='cuda:5'), covar=tensor([0.1042, 0.0788, 0.0987, 0.0857, 0.0813, 0.0764, 0.1087, 0.0844], device='cuda:5'), in_proj_covar=tensor([0.0685, 0.0842, 0.0692, 0.0640, 0.0533, 0.0542, 0.0704, 0.0658], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:44:39,666 INFO [train.py:904] (5/8) Epoch 22, batch 4200, loss[loss=0.1856, simple_loss=0.2779, pruned_loss=0.04671, over 16597.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2692, pruned_loss=0.04987, over 3216234.91 frames. ], batch size: 57, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:44:50,072 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217359.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:44:52,709 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:04,366 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-01 10:45:32,832 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217388.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:53,908 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 2.187e+02 2.459e+02 3.018e+02 5.838e+02, threshold=4.919e+02, percent-clipped=1.0 2023-05-01 10:45:55,232 INFO [train.py:904] (5/8) Epoch 22, batch 4250, loss[loss=0.1627, simple_loss=0.2628, pruned_loss=0.03132, over 16646.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2723, pruned_loss=0.04992, over 3176675.33 frames. ], batch size: 89, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:46:04,663 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217409.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:47:04,208 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217449.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:47:09,162 INFO [train.py:904] (5/8) Epoch 22, batch 4300, loss[loss=0.1879, simple_loss=0.2812, pruned_loss=0.04734, over 16769.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2734, pruned_loss=0.04891, over 3179266.40 frames. ], batch size: 83, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:48:07,890 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217492.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:48:23,054 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.190e+02 2.513e+02 2.861e+02 5.953e+02, threshold=5.026e+02, percent-clipped=1.0 2023-05-01 10:48:24,301 INFO [train.py:904] (5/8) Epoch 22, batch 4350, loss[loss=0.197, simple_loss=0.2877, pruned_loss=0.0531, over 16889.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2769, pruned_loss=0.05005, over 3183318.74 frames. ], batch size: 116, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:48:40,558 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-05-01 10:49:39,993 INFO [train.py:904] (5/8) Epoch 22, batch 4400, loss[loss=0.1763, simple_loss=0.2631, pruned_loss=0.04479, over 16969.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2787, pruned_loss=0.05084, over 3194864.24 frames. ], batch size: 41, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:41,128 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:50:37,674 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6215, 2.7352, 2.5707, 4.7428, 3.4862, 4.1350, 1.6954, 2.8748], device='cuda:5'), covar=tensor([0.1501, 0.0926, 0.1336, 0.0167, 0.0355, 0.0369, 0.1846, 0.0906], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0173, 0.0194, 0.0192, 0.0205, 0.0215, 0.0201, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 10:50:52,394 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 2.074e+02 2.354e+02 2.786e+02 4.691e+02, threshold=4.709e+02, percent-clipped=0.0 2023-05-01 10:50:53,498 INFO [train.py:904] (5/8) Epoch 22, batch 4450, loss[loss=0.2299, simple_loss=0.3159, pruned_loss=0.07193, over 16393.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2824, pruned_loss=0.05215, over 3207100.82 frames. ], batch size: 146, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:50:59,554 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0972, 2.3176, 2.3519, 3.9261, 2.1562, 2.6891, 2.3785, 2.4690], device='cuda:5'), covar=tensor([0.1413, 0.3270, 0.2734, 0.0499, 0.3988, 0.2345, 0.3264, 0.3207], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0451, 0.0369, 0.0330, 0.0439, 0.0520, 0.0422, 0.0528], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:51:20,457 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:51:41,132 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 10:52:08,357 INFO [train.py:904] (5/8) Epoch 22, batch 4500, loss[loss=0.1824, simple_loss=0.2805, pruned_loss=0.0422, over 16821.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2825, pruned_loss=0.05256, over 3219056.61 frames. ], batch size: 102, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:52:24,110 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:52:32,281 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:53:18,213 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 1.861e+02 2.083e+02 2.379e+02 4.476e+02, threshold=4.167e+02, percent-clipped=0.0 2023-05-01 10:53:19,302 INFO [train.py:904] (5/8) Epoch 22, batch 4550, loss[loss=0.2281, simple_loss=0.3135, pruned_loss=0.07132, over 17012.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2838, pruned_loss=0.05391, over 3227825.92 frames. ], batch size: 50, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:53:53,271 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:54:18,183 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2460, 1.5629, 1.9270, 2.1016, 2.2432, 2.4822, 1.7251, 2.3496], device='cuda:5'), covar=tensor([0.0227, 0.0516, 0.0341, 0.0335, 0.0320, 0.0178, 0.0599, 0.0141], device='cuda:5'), in_proj_covar=tensor([0.0192, 0.0196, 0.0182, 0.0188, 0.0200, 0.0156, 0.0199, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:54:19,812 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217744.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:54:31,798 INFO [train.py:904] (5/8) Epoch 22, batch 4600, loss[loss=0.1974, simple_loss=0.2818, pruned_loss=0.05648, over 17053.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2851, pruned_loss=0.05432, over 3232427.94 frames. ], batch size: 50, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:54:55,229 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8739, 3.7960, 3.9077, 4.0096, 4.0842, 3.7156, 4.0302, 4.1331], device='cuda:5'), covar=tensor([0.1302, 0.0981, 0.1146, 0.0621, 0.0527, 0.1918, 0.0837, 0.0603], device='cuda:5'), in_proj_covar=tensor([0.0638, 0.0791, 0.0919, 0.0802, 0.0608, 0.0636, 0.0658, 0.0761], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:55:41,529 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 1.780e+02 2.045e+02 2.386e+02 4.197e+02, threshold=4.091e+02, percent-clipped=1.0 2023-05-01 10:55:42,811 INFO [train.py:904] (5/8) Epoch 22, batch 4650, loss[loss=0.1855, simple_loss=0.2743, pruned_loss=0.04833, over 15353.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2843, pruned_loss=0.05436, over 3245395.59 frames. ], batch size: 191, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:55:45,531 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-05-01 10:56:26,656 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 10:56:46,919 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217848.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:56:52,992 INFO [train.py:904] (5/8) Epoch 22, batch 4700, loss[loss=0.1866, simple_loss=0.2779, pruned_loss=0.04761, over 16432.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2817, pruned_loss=0.05327, over 3241381.85 frames. ], batch size: 146, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:57:35,563 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9514, 4.7703, 4.9437, 5.1358, 5.3323, 4.7387, 5.2946, 5.3233], device='cuda:5'), covar=tensor([0.1743, 0.1239, 0.1670, 0.0684, 0.0475, 0.0808, 0.0551, 0.0539], device='cuda:5'), in_proj_covar=tensor([0.0635, 0.0788, 0.0916, 0.0799, 0.0605, 0.0632, 0.0655, 0.0758], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 10:57:52,901 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0346, 5.5532, 5.7697, 5.5197, 5.5132, 6.0977, 5.6222, 5.3211], device='cuda:5'), covar=tensor([0.0853, 0.1555, 0.1873, 0.1664, 0.2342, 0.0871, 0.1337, 0.2195], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0591, 0.0645, 0.0489, 0.0653, 0.0684, 0.0509, 0.0658], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 10:57:59,216 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217898.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:58:04,667 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 1.895e+02 2.114e+02 2.418e+02 3.922e+02, threshold=4.227e+02, percent-clipped=0.0 2023-05-01 10:58:05,940 INFO [train.py:904] (5/8) Epoch 22, batch 4750, loss[loss=0.2093, simple_loss=0.2976, pruned_loss=0.06054, over 12000.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2775, pruned_loss=0.05117, over 3232538.29 frames. ], batch size: 247, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:59:17,342 INFO [train.py:904] (5/8) Epoch 22, batch 4800, loss[loss=0.184, simple_loss=0.2786, pruned_loss=0.0447, over 16742.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2732, pruned_loss=0.04902, over 3227175.37 frames. ], batch size: 124, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:59:27,599 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217959.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:00:25,696 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 11:00:36,343 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.772e+02 2.165e+02 2.536e+02 4.360e+02, threshold=4.330e+02, percent-clipped=2.0 2023-05-01 11:00:36,358 INFO [train.py:904] (5/8) Epoch 22, batch 4850, loss[loss=0.1807, simple_loss=0.2706, pruned_loss=0.04543, over 16475.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2732, pruned_loss=0.04806, over 3212701.33 frames. ], batch size: 75, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:00:39,662 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 11:01:01,609 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218020.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:38,671 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:51,575 INFO [train.py:904] (5/8) Epoch 22, batch 4900, loss[loss=0.1859, simple_loss=0.2702, pruned_loss=0.05084, over 11833.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2723, pruned_loss=0.04661, over 3191912.03 frames. ], batch size: 247, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:02:49,303 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218092.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:03:01,674 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218100.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:03:05,448 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.042e+02 2.302e+02 2.921e+02 6.655e+02, threshold=4.605e+02, percent-clipped=4.0 2023-05-01 11:03:05,470 INFO [train.py:904] (5/8) Epoch 22, batch 4950, loss[loss=0.1754, simple_loss=0.2786, pruned_loss=0.03608, over 16867.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2716, pruned_loss=0.04582, over 3188665.20 frames. ], batch size: 116, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:03:49,425 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 11:04:06,373 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 11:04:11,477 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:04:18,893 INFO [train.py:904] (5/8) Epoch 22, batch 5000, loss[loss=0.1944, simple_loss=0.2916, pruned_loss=0.04866, over 16341.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2735, pruned_loss=0.04607, over 3182601.76 frames. ], batch size: 165, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:04:30,062 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:01,556 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218182.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:18,228 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4357, 4.3753, 4.3674, 3.2600, 4.3998, 1.5374, 4.0644, 4.0366], device='cuda:5'), covar=tensor([0.0144, 0.0143, 0.0201, 0.0671, 0.0145, 0.3309, 0.0179, 0.0325], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0156, 0.0199, 0.0179, 0.0177, 0.0208, 0.0189, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:05:21,404 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218196.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:32,362 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.007e+02 2.287e+02 2.605e+02 5.161e+02, threshold=4.575e+02, percent-clipped=1.0 2023-05-01 11:05:32,378 INFO [train.py:904] (5/8) Epoch 22, batch 5050, loss[loss=0.1607, simple_loss=0.2535, pruned_loss=0.03393, over 17146.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2738, pruned_loss=0.04617, over 3199134.11 frames. ], batch size: 47, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:05:42,453 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:43,693 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9357, 1.9698, 2.5358, 2.8246, 2.7645, 3.3873, 2.2075, 3.3297], device='cuda:5'), covar=tensor([0.0228, 0.0547, 0.0358, 0.0333, 0.0344, 0.0149, 0.0536, 0.0137], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0197, 0.0183, 0.0188, 0.0202, 0.0156, 0.0200, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:06:00,226 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4479, 4.2523, 4.3401, 4.6599, 4.8159, 4.4265, 4.8499, 4.8262], device='cuda:5'), covar=tensor([0.1731, 0.1376, 0.1949, 0.0802, 0.0600, 0.0958, 0.0646, 0.0745], device='cuda:5'), in_proj_covar=tensor([0.0628, 0.0778, 0.0902, 0.0789, 0.0594, 0.0626, 0.0645, 0.0748], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:06:30,556 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218243.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:06:35,217 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8917, 3.2240, 3.3708, 2.0866, 2.8707, 2.2510, 3.3982, 3.3790], device='cuda:5'), covar=tensor([0.0237, 0.0730, 0.0571, 0.1868, 0.0833, 0.0901, 0.0587, 0.0793], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0165, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 11:06:44,158 INFO [train.py:904] (5/8) Epoch 22, batch 5100, loss[loss=0.1764, simple_loss=0.2715, pruned_loss=0.04069, over 16230.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2718, pruned_loss=0.04529, over 3208653.58 frames. ], batch size: 165, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:06:45,778 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218254.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:07:10,550 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218271.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:07:57,331 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 1.925e+02 2.150e+02 2.531e+02 3.604e+02, threshold=4.300e+02, percent-clipped=0.0 2023-05-01 11:07:57,346 INFO [train.py:904] (5/8) Epoch 22, batch 5150, loss[loss=0.1752, simple_loss=0.266, pruned_loss=0.04223, over 11635.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2719, pruned_loss=0.04439, over 3196799.86 frames. ], batch size: 248, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:08:07,907 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-01 11:08:23,190 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:08:29,195 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1104, 4.8405, 5.1143, 5.2974, 5.5579, 4.8927, 5.4867, 5.5234], device='cuda:5'), covar=tensor([0.1828, 0.1574, 0.1951, 0.0885, 0.0593, 0.0810, 0.0602, 0.0705], device='cuda:5'), in_proj_covar=tensor([0.0628, 0.0782, 0.0905, 0.0792, 0.0595, 0.0626, 0.0647, 0.0749], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:09:11,260 INFO [train.py:904] (5/8) Epoch 22, batch 5200, loss[loss=0.1768, simple_loss=0.2722, pruned_loss=0.04077, over 16674.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2712, pruned_loss=0.04403, over 3201099.16 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:09:33,120 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218368.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:10:23,968 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 1.955e+02 2.305e+02 2.711e+02 5.108e+02, threshold=4.610e+02, percent-clipped=2.0 2023-05-01 11:10:23,983 INFO [train.py:904] (5/8) Epoch 22, batch 5250, loss[loss=0.1696, simple_loss=0.2609, pruned_loss=0.03918, over 16661.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2687, pruned_loss=0.04397, over 3204521.82 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:11:36,557 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0642, 2.1612, 2.1673, 3.5870, 2.1240, 2.5043, 2.3097, 2.3728], device='cuda:5'), covar=tensor([0.1379, 0.3383, 0.3019, 0.0586, 0.4055, 0.2525, 0.3586, 0.3096], device='cuda:5'), in_proj_covar=tensor([0.0401, 0.0447, 0.0366, 0.0327, 0.0433, 0.0515, 0.0419, 0.0522], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:11:37,150 INFO [train.py:904] (5/8) Epoch 22, batch 5300, loss[loss=0.1665, simple_loss=0.2536, pruned_loss=0.03975, over 16228.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2657, pruned_loss=0.04279, over 3207077.26 frames. ], batch size: 165, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:11:41,574 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218456.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:12:16,012 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 11:12:19,613 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8980, 4.7325, 4.9811, 5.1295, 5.3317, 4.7272, 5.3264, 5.3450], device='cuda:5'), covar=tensor([0.1732, 0.1297, 0.1624, 0.0815, 0.0543, 0.0876, 0.0531, 0.0528], device='cuda:5'), in_proj_covar=tensor([0.0631, 0.0784, 0.0910, 0.0795, 0.0598, 0.0630, 0.0650, 0.0750], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:12:31,134 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8718, 2.7428, 2.8532, 2.0995, 2.6750, 2.1171, 2.7659, 2.9042], device='cuda:5'), covar=tensor([0.0318, 0.0792, 0.0584, 0.1943, 0.0862, 0.1018, 0.0637, 0.0712], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0164, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 11:12:51,258 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 1.955e+02 2.262e+02 2.726e+02 6.747e+02, threshold=4.525e+02, percent-clipped=3.0 2023-05-01 11:12:51,273 INFO [train.py:904] (5/8) Epoch 22, batch 5350, loss[loss=0.1999, simple_loss=0.2978, pruned_loss=0.05099, over 16719.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2644, pruned_loss=0.04227, over 3224532.09 frames. ], batch size: 124, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:13:00,637 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:13:43,241 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:13:58,150 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4317, 5.7431, 5.4359, 5.5680, 5.2436, 5.1933, 5.1593, 5.8601], device='cuda:5'), covar=tensor([0.1294, 0.0816, 0.1034, 0.0766, 0.0761, 0.0709, 0.1066, 0.0819], device='cuda:5'), in_proj_covar=tensor([0.0668, 0.0816, 0.0677, 0.0622, 0.0520, 0.0526, 0.0683, 0.0640], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:14:03,263 INFO [train.py:904] (5/8) Epoch 22, batch 5400, loss[loss=0.1759, simple_loss=0.2609, pruned_loss=0.04551, over 16201.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2668, pruned_loss=0.04286, over 3220511.00 frames. ], batch size: 35, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:14:06,134 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218554.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:14:23,162 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218566.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:14:28,964 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:34,804 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218574.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:52,510 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 11:15:18,766 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218602.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:15:19,493 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 2.152e+02 2.360e+02 2.669e+02 5.554e+02, threshold=4.720e+02, percent-clipped=2.0 2023-05-01 11:15:19,515 INFO [train.py:904] (5/8) Epoch 22, batch 5450, loss[loss=0.1897, simple_loss=0.2751, pruned_loss=0.05217, over 16605.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2699, pruned_loss=0.04445, over 3202115.58 frames. ], batch size: 57, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:16:11,031 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218635.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:16:38,004 INFO [train.py:904] (5/8) Epoch 22, batch 5500, loss[loss=0.2179, simple_loss=0.3034, pruned_loss=0.06619, over 16862.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2769, pruned_loss=0.0486, over 3189930.80 frames. ], batch size: 39, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:17:53,058 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7960, 4.6530, 4.8104, 5.0007, 5.1663, 4.6110, 5.1684, 5.1684], device='cuda:5'), covar=tensor([0.1744, 0.1219, 0.1490, 0.0697, 0.0614, 0.0998, 0.0601, 0.0619], device='cuda:5'), in_proj_covar=tensor([0.0632, 0.0785, 0.0910, 0.0793, 0.0597, 0.0629, 0.0649, 0.0752], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:17:57,933 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.055e+02 2.950e+02 3.517e+02 4.298e+02 7.348e+02, threshold=7.033e+02, percent-clipped=15.0 2023-05-01 11:17:57,955 INFO [train.py:904] (5/8) Epoch 22, batch 5550, loss[loss=0.2148, simple_loss=0.2986, pruned_loss=0.06554, over 16704.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2841, pruned_loss=0.05366, over 3160637.25 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:18:32,096 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0456, 4.8504, 5.0204, 5.2174, 5.4146, 4.7725, 5.4199, 5.4120], device='cuda:5'), covar=tensor([0.1692, 0.1228, 0.1581, 0.0721, 0.0612, 0.0911, 0.0557, 0.0618], device='cuda:5'), in_proj_covar=tensor([0.0631, 0.0784, 0.0909, 0.0792, 0.0597, 0.0628, 0.0649, 0.0752], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:19:20,445 INFO [train.py:904] (5/8) Epoch 22, batch 5600, loss[loss=0.1961, simple_loss=0.2816, pruned_loss=0.05527, over 16658.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2892, pruned_loss=0.05798, over 3107808.73 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:19:25,788 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218756.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:20:08,440 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8828, 3.1061, 3.1396, 2.0268, 2.9652, 3.2330, 2.9904, 1.8091], device='cuda:5'), covar=tensor([0.0580, 0.0080, 0.0083, 0.0506, 0.0130, 0.0132, 0.0117, 0.0512], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0083, 0.0084, 0.0133, 0.0098, 0.0110, 0.0095, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 11:20:41,874 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.037e+02 3.163e+02 3.914e+02 4.827e+02 8.104e+02, threshold=7.827e+02, percent-clipped=4.0 2023-05-01 11:20:41,895 INFO [train.py:904] (5/8) Epoch 22, batch 5650, loss[loss=0.2032, simple_loss=0.2949, pruned_loss=0.05579, over 16858.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.295, pruned_loss=0.06289, over 3050991.86 frames. ], batch size: 96, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:20:44,086 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218804.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:21:03,627 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2540, 5.5353, 5.2804, 5.2928, 4.9870, 4.9292, 5.0139, 5.6442], device='cuda:5'), covar=tensor([0.1145, 0.0900, 0.1057, 0.0867, 0.0812, 0.0834, 0.1186, 0.0896], device='cuda:5'), in_proj_covar=tensor([0.0657, 0.0805, 0.0667, 0.0612, 0.0509, 0.0517, 0.0670, 0.0630], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:21:09,609 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 11:21:35,458 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218838.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:21:56,905 INFO [train.py:904] (5/8) Epoch 22, batch 5700, loss[loss=0.2102, simple_loss=0.3042, pruned_loss=0.05805, over 16436.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2964, pruned_loss=0.06409, over 3049346.72 frames. ], batch size: 146, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:22:14,842 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7367, 3.7456, 3.9074, 3.6471, 3.8392, 4.2159, 3.8689, 3.5510], device='cuda:5'), covar=tensor([0.2149, 0.2251, 0.2119, 0.2439, 0.2464, 0.1764, 0.1764, 0.2785], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0584, 0.0644, 0.0488, 0.0649, 0.0677, 0.0510, 0.0655], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 11:22:16,269 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218865.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:17,591 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:19,652 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-01 11:22:48,657 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218886.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:13,878 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.926e+02 3.484e+02 4.272e+02 7.645e+02, threshold=6.969e+02, percent-clipped=0.0 2023-05-01 11:23:13,900 INFO [train.py:904] (5/8) Epoch 22, batch 5750, loss[loss=0.2421, simple_loss=0.3057, pruned_loss=0.08921, over 11000.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2994, pruned_loss=0.06619, over 3014101.19 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:23:32,118 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:57,595 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:57,817 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8341, 3.2630, 3.3257, 2.0066, 2.9099, 2.2103, 3.3588, 3.4143], device='cuda:5'), covar=tensor([0.0326, 0.0743, 0.0663, 0.2117, 0.0865, 0.1126, 0.0662, 0.0958], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0166, 0.0153, 0.0145, 0.0130, 0.0143, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 11:24:26,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1397, 2.0355, 2.6983, 3.0839, 2.8808, 3.5412, 2.3958, 3.4445], device='cuda:5'), covar=tensor([0.0229, 0.0538, 0.0363, 0.0291, 0.0358, 0.0165, 0.0491, 0.0165], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0194, 0.0179, 0.0185, 0.0197, 0.0152, 0.0196, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:24:34,392 INFO [train.py:904] (5/8) Epoch 22, batch 5800, loss[loss=0.2154, simple_loss=0.2903, pruned_loss=0.07023, over 12005.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2977, pruned_loss=0.06416, over 3025460.84 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:24:54,964 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:15,176 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4032, 2.1978, 2.9359, 3.2302, 3.0573, 3.7513, 2.4462, 3.7083], device='cuda:5'), covar=tensor([0.0204, 0.0492, 0.0296, 0.0296, 0.0295, 0.0140, 0.0508, 0.0125], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0193, 0.0179, 0.0184, 0.0197, 0.0152, 0.0196, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:25:24,845 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:35,407 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1651, 3.5162, 3.6249, 2.3172, 3.3107, 3.7201, 3.3521, 2.0148], device='cuda:5'), covar=tensor([0.0565, 0.0073, 0.0058, 0.0450, 0.0107, 0.0103, 0.0102, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0083, 0.0084, 0.0133, 0.0098, 0.0110, 0.0095, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 11:25:53,554 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 2.954e+02 3.399e+02 3.875e+02 5.875e+02, threshold=6.799e+02, percent-clipped=0.0 2023-05-01 11:25:53,576 INFO [train.py:904] (5/8) Epoch 22, batch 5850, loss[loss=0.1932, simple_loss=0.281, pruned_loss=0.0527, over 16400.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.295, pruned_loss=0.0622, over 3032198.68 frames. ], batch size: 146, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:25:55,601 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0352, 2.3228, 1.8751, 2.1298, 2.7753, 2.3953, 2.6658, 2.9439], device='cuda:5'), covar=tensor([0.0205, 0.0529, 0.0723, 0.0520, 0.0296, 0.0459, 0.0322, 0.0326], device='cuda:5'), in_proj_covar=tensor([0.0208, 0.0230, 0.0223, 0.0222, 0.0232, 0.0231, 0.0232, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:26:21,517 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:26:29,990 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:02,346 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:06,875 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4809, 4.3790, 4.5548, 4.7157, 4.8731, 4.4116, 4.8294, 4.8798], device='cuda:5'), covar=tensor([0.1960, 0.1208, 0.1584, 0.0730, 0.0575, 0.1034, 0.0715, 0.0635], device='cuda:5'), in_proj_covar=tensor([0.0620, 0.0769, 0.0892, 0.0777, 0.0588, 0.0616, 0.0639, 0.0738], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:27:14,941 INFO [train.py:904] (5/8) Epoch 22, batch 5900, loss[loss=0.1922, simple_loss=0.2866, pruned_loss=0.04893, over 16352.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2943, pruned_loss=0.06144, over 3048072.92 frames. ], batch size: 35, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:27:28,079 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 11:27:49,475 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219072.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:04,712 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:35,886 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.529e+02 2.968e+02 3.874e+02 6.567e+02, threshold=5.937e+02, percent-clipped=0.0 2023-05-01 11:28:35,902 INFO [train.py:904] (5/8) Epoch 22, batch 5950, loss[loss=0.2033, simple_loss=0.2893, pruned_loss=0.05867, over 16607.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2947, pruned_loss=0.06029, over 3054678.11 frames. ], batch size: 68, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:29:24,557 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:29:57,668 INFO [train.py:904] (5/8) Epoch 22, batch 6000, loss[loss=0.2234, simple_loss=0.2955, pruned_loss=0.07563, over 11257.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2939, pruned_loss=0.0601, over 3059030.32 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:29:57,668 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 11:30:07,617 INFO [train.py:938] (5/8) Epoch 22, validation: loss=0.1507, simple_loss=0.2632, pruned_loss=0.01907, over 944034.00 frames. 2023-05-01 11:30:07,618 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 11:30:27,740 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219165.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:12,382 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4996, 4.3828, 4.5240, 4.6993, 4.8496, 4.3820, 4.8204, 4.8583], device='cuda:5'), covar=tensor([0.1872, 0.1243, 0.1602, 0.0773, 0.0629, 0.1072, 0.0783, 0.0708], device='cuda:5'), in_proj_covar=tensor([0.0625, 0.0774, 0.0897, 0.0784, 0.0592, 0.0620, 0.0645, 0.0745], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:31:21,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0662, 2.1693, 2.2387, 3.7914, 2.0714, 2.5424, 2.2341, 2.3288], device='cuda:5'), covar=tensor([0.1511, 0.3804, 0.3092, 0.0569, 0.4325, 0.2624, 0.3898, 0.3425], device='cuda:5'), in_proj_covar=tensor([0.0401, 0.0447, 0.0366, 0.0325, 0.0435, 0.0515, 0.0419, 0.0522], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:31:28,923 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.852e+02 3.411e+02 4.220e+02 7.320e+02, threshold=6.821e+02, percent-clipped=6.0 2023-05-01 11:31:28,945 INFO [train.py:904] (5/8) Epoch 22, batch 6050, loss[loss=0.1877, simple_loss=0.3011, pruned_loss=0.03711, over 16918.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2925, pruned_loss=0.05937, over 3068508.69 frames. ], batch size: 96, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:31:29,685 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:44,912 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:52,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3607, 2.1265, 1.7675, 1.9496, 2.4339, 2.1250, 2.1260, 2.5418], device='cuda:5'), covar=tensor([0.0213, 0.0403, 0.0544, 0.0475, 0.0261, 0.0383, 0.0240, 0.0271], device='cuda:5'), in_proj_covar=tensor([0.0208, 0.0230, 0.0223, 0.0223, 0.0232, 0.0231, 0.0232, 0.0227], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:32:11,178 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219230.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:32:24,424 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6246, 4.2827, 4.2504, 2.7965, 3.7951, 4.3746, 3.7575, 2.5088], device='cuda:5'), covar=tensor([0.0503, 0.0051, 0.0057, 0.0399, 0.0103, 0.0100, 0.0103, 0.0411], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0132, 0.0097, 0.0109, 0.0094, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 11:32:46,228 INFO [train.py:904] (5/8) Epoch 22, batch 6100, loss[loss=0.1939, simple_loss=0.2883, pruned_loss=0.04973, over 16760.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2917, pruned_loss=0.05841, over 3088719.79 frames. ], batch size: 83, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:33:05,421 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219264.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:33:26,348 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:34:04,133 INFO [train.py:904] (5/8) Epoch 22, batch 6150, loss[loss=0.1871, simple_loss=0.2748, pruned_loss=0.04969, over 17044.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2891, pruned_loss=0.05746, over 3093533.12 frames. ], batch size: 53, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:34:05,868 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.531e+02 2.938e+02 3.735e+02 5.885e+02, threshold=5.877e+02, percent-clipped=0.0 2023-05-01 11:34:33,729 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:35:03,743 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:35:06,914 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219342.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:35:22,823 INFO [train.py:904] (5/8) Epoch 22, batch 6200, loss[loss=0.1851, simple_loss=0.2712, pruned_loss=0.0495, over 16544.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2874, pruned_loss=0.05725, over 3086815.15 frames. ], batch size: 62, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:02,568 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:36:42,499 INFO [train.py:904] (5/8) Epoch 22, batch 6250, loss[loss=0.1725, simple_loss=0.2664, pruned_loss=0.0393, over 17015.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2875, pruned_loss=0.05718, over 3087200.83 frames. ], batch size: 55, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:43,089 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219403.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:36:43,737 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.827e+02 2.722e+02 3.215e+02 3.900e+02 7.497e+02, threshold=6.430e+02, percent-clipped=6.0 2023-05-01 11:37:20,984 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:37:57,567 INFO [train.py:904] (5/8) Epoch 22, batch 6300, loss[loss=0.2145, simple_loss=0.29, pruned_loss=0.06951, over 11784.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2879, pruned_loss=0.05697, over 3081166.57 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:38:56,031 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219490.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:39:15,309 INFO [train.py:904] (5/8) Epoch 22, batch 6350, loss[loss=0.2176, simple_loss=0.2998, pruned_loss=0.06771, over 15258.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2884, pruned_loss=0.05769, over 3084244.22 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:39:16,415 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 2.840e+02 3.669e+02 4.695e+02 9.321e+02, threshold=7.339e+02, percent-clipped=9.0 2023-05-01 11:39:21,937 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6436, 4.8298, 4.9625, 4.7609, 4.8287, 5.3458, 4.8548, 4.6126], device='cuda:5'), covar=tensor([0.1220, 0.1927, 0.2600, 0.1861, 0.2263, 0.1070, 0.1785, 0.2571], device='cuda:5'), in_proj_covar=tensor([0.0414, 0.0592, 0.0655, 0.0495, 0.0658, 0.0684, 0.0515, 0.0663], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 11:39:34,085 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5514, 4.4960, 4.3464, 2.7616, 3.8480, 4.5004, 3.8257, 2.5048], device='cuda:5'), covar=tensor([0.0551, 0.0036, 0.0045, 0.0406, 0.0110, 0.0081, 0.0094, 0.0424], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0083, 0.0084, 0.0133, 0.0098, 0.0110, 0.0095, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 11:39:39,673 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4425, 1.6092, 2.1307, 2.3507, 2.4296, 2.6225, 1.8082, 2.5495], device='cuda:5'), covar=tensor([0.0233, 0.0558, 0.0329, 0.0366, 0.0316, 0.0196, 0.0543, 0.0134], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0193, 0.0179, 0.0183, 0.0196, 0.0153, 0.0196, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:40:05,447 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5867, 1.7531, 1.6266, 1.4811, 1.9190, 1.6486, 1.5714, 1.8910], device='cuda:5'), covar=tensor([0.0221, 0.0288, 0.0397, 0.0394, 0.0207, 0.0263, 0.0216, 0.0237], device='cuda:5'), in_proj_covar=tensor([0.0206, 0.0229, 0.0222, 0.0222, 0.0231, 0.0230, 0.0230, 0.0225], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:40:08,107 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 11:40:28,717 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:40:31,091 INFO [train.py:904] (5/8) Epoch 22, batch 6400, loss[loss=0.1938, simple_loss=0.2837, pruned_loss=0.05196, over 16450.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2892, pruned_loss=0.05913, over 3084839.39 frames. ], batch size: 146, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:40:40,459 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219559.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:41:45,419 INFO [train.py:904] (5/8) Epoch 22, batch 6450, loss[loss=0.1907, simple_loss=0.2954, pruned_loss=0.043, over 16586.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2893, pruned_loss=0.05859, over 3080570.93 frames. ], batch size: 68, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:41:47,192 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.795e+02 3.528e+02 4.334e+02 7.123e+02, threshold=7.056e+02, percent-clipped=0.0 2023-05-01 11:42:13,307 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:42:45,365 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:42:47,393 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 11:42:59,618 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1581, 2.4221, 2.6310, 2.0168, 2.7019, 2.8012, 2.4374, 2.3916], device='cuda:5'), covar=tensor([0.0693, 0.0253, 0.0232, 0.0913, 0.0122, 0.0304, 0.0495, 0.0465], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0109, 0.0099, 0.0140, 0.0082, 0.0128, 0.0130, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 11:43:04,575 INFO [train.py:904] (5/8) Epoch 22, batch 6500, loss[loss=0.1711, simple_loss=0.2682, pruned_loss=0.037, over 16707.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2871, pruned_loss=0.058, over 3080839.73 frames. ], batch size: 89, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:43:30,558 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:43:43,445 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:44:01,335 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:44:19,771 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219698.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:44:26,659 INFO [train.py:904] (5/8) Epoch 22, batch 6550, loss[loss=0.2037, simple_loss=0.3082, pruned_loss=0.04962, over 16524.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2898, pruned_loss=0.05884, over 3072107.94 frames. ], batch size: 62, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:44:28,428 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.583e+02 3.105e+02 3.738e+02 8.268e+02, threshold=6.210e+02, percent-clipped=2.0 2023-05-01 11:45:03,216 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:08,812 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:39,630 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9609, 3.1353, 3.1345, 2.1576, 2.9451, 3.2126, 3.0198, 1.8860], device='cuda:5'), covar=tensor([0.0554, 0.0075, 0.0074, 0.0442, 0.0127, 0.0128, 0.0121, 0.0503], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0132, 0.0097, 0.0109, 0.0094, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 11:45:46,191 INFO [train.py:904] (5/8) Epoch 22, batch 6600, loss[loss=0.2567, simple_loss=0.322, pruned_loss=0.09567, over 11555.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2917, pruned_loss=0.05909, over 3085395.36 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:46:02,205 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219763.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:46:04,363 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3688, 3.4113, 2.6623, 2.1409, 2.2952, 2.3177, 3.6459, 3.1822], device='cuda:5'), covar=tensor([0.3242, 0.0783, 0.1961, 0.2982, 0.2611, 0.2152, 0.0606, 0.1318], device='cuda:5'), in_proj_covar=tensor([0.0325, 0.0268, 0.0304, 0.0313, 0.0296, 0.0258, 0.0294, 0.0335], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 11:46:20,415 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3337, 5.3825, 5.1900, 4.7874, 4.8272, 5.2609, 5.2013, 4.9097], device='cuda:5'), covar=tensor([0.0610, 0.0515, 0.0281, 0.0327, 0.1035, 0.0467, 0.0271, 0.0679], device='cuda:5'), in_proj_covar=tensor([0.0290, 0.0427, 0.0339, 0.0337, 0.0346, 0.0391, 0.0234, 0.0405], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:46:23,295 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:46:50,187 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2023-05-01 11:47:06,508 INFO [train.py:904] (5/8) Epoch 22, batch 6650, loss[loss=0.1911, simple_loss=0.2794, pruned_loss=0.05145, over 16677.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.292, pruned_loss=0.05952, over 3093042.78 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:47:07,645 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.943e+02 3.397e+02 4.500e+02 8.184e+02, threshold=6.793e+02, percent-clipped=7.0 2023-05-01 11:47:39,845 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219824.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:10,771 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219846.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:21,247 INFO [train.py:904] (5/8) Epoch 22, batch 6700, loss[loss=0.199, simple_loss=0.2883, pruned_loss=0.05484, over 16217.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2906, pruned_loss=0.05937, over 3091576.64 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:48:27,654 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:30,189 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:49:36,059 INFO [train.py:904] (5/8) Epoch 22, batch 6750, loss[loss=0.255, simple_loss=0.3208, pruned_loss=0.09457, over 11621.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2899, pruned_loss=0.05996, over 3090029.39 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:49:37,872 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.924e+02 3.493e+02 4.502e+02 8.889e+02, threshold=6.985e+02, percent-clipped=2.0 2023-05-01 11:49:41,047 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5420, 3.5027, 3.4394, 2.6319, 3.3855, 2.0849, 3.2218, 2.8142], device='cuda:5'), covar=tensor([0.0179, 0.0153, 0.0236, 0.0297, 0.0131, 0.2494, 0.0162, 0.0290], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0157, 0.0200, 0.0180, 0.0176, 0.0209, 0.0188, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:49:43,896 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:49:51,738 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 11:50:00,720 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:54,136 INFO [train.py:904] (5/8) Epoch 22, batch 6800, loss[loss=0.2389, simple_loss=0.3102, pruned_loss=0.08378, over 11778.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2901, pruned_loss=0.05993, over 3079036.59 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:51:04,075 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0332, 4.1164, 3.9310, 3.6651, 3.7013, 4.0541, 3.6576, 3.8452], device='cuda:5'), covar=tensor([0.0568, 0.0570, 0.0280, 0.0263, 0.0702, 0.0456, 0.1052, 0.0562], device='cuda:5'), in_proj_covar=tensor([0.0291, 0.0428, 0.0342, 0.0338, 0.0348, 0.0394, 0.0235, 0.0407], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:52:04,949 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219998.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:52:15,016 INFO [train.py:904] (5/8) Epoch 22, batch 6850, loss[loss=0.2038, simple_loss=0.312, pruned_loss=0.04782, over 16828.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2907, pruned_loss=0.06031, over 3073871.67 frames. ], batch size: 102, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:52:16,806 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.941e+02 2.883e+02 3.357e+02 3.930e+02 6.765e+02, threshold=6.713e+02, percent-clipped=0.0 2023-05-01 11:53:20,650 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220046.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:53:31,122 INFO [train.py:904] (5/8) Epoch 22, batch 6900, loss[loss=0.2263, simple_loss=0.3089, pruned_loss=0.07183, over 16145.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2934, pruned_loss=0.06024, over 3088081.15 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:53:37,337 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1333, 2.1477, 2.2306, 3.6972, 2.1128, 2.5830, 2.2657, 2.3294], device='cuda:5'), covar=tensor([0.1391, 0.3607, 0.2938, 0.0575, 0.4104, 0.2392, 0.3464, 0.3180], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0447, 0.0366, 0.0326, 0.0436, 0.0515, 0.0418, 0.0521], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:54:24,709 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220087.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:41,866 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220098.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:50,003 INFO [train.py:904] (5/8) Epoch 22, batch 6950, loss[loss=0.1936, simple_loss=0.2786, pruned_loss=0.05431, over 16329.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2954, pruned_loss=0.06249, over 3067750.01 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:54:51,081 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 2.861e+02 3.409e+02 4.251e+02 1.328e+03, threshold=6.818e+02, percent-clipped=1.0 2023-05-01 11:55:14,108 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220119.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:20,416 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6885, 1.7011, 1.4580, 1.3798, 1.8081, 1.4726, 1.5863, 1.8799], device='cuda:5'), covar=tensor([0.0213, 0.0380, 0.0547, 0.0445, 0.0283, 0.0331, 0.0181, 0.0266], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0230, 0.0223, 0.0222, 0.0232, 0.0230, 0.0230, 0.0226], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:55:22,694 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8379, 5.2690, 5.4130, 5.1792, 5.2698, 5.7996, 5.2777, 5.0877], device='cuda:5'), covar=tensor([0.1129, 0.1756, 0.2380, 0.1802, 0.2536, 0.0987, 0.1694, 0.2231], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0589, 0.0651, 0.0490, 0.0654, 0.0679, 0.0513, 0.0659], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 11:55:54,670 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:57,597 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:56:03,774 INFO [train.py:904] (5/8) Epoch 22, batch 7000, loss[loss=0.2055, simple_loss=0.2956, pruned_loss=0.05767, over 16696.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2951, pruned_loss=0.06174, over 3058057.90 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 11:56:14,275 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:56:21,629 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 11:56:43,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6625, 2.5848, 1.8837, 2.7357, 2.1368, 2.8170, 2.1149, 2.3984], device='cuda:5'), covar=tensor([0.0343, 0.0386, 0.1275, 0.0267, 0.0655, 0.0551, 0.1174, 0.0574], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0161, 0.0176, 0.0216, 0.0201, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 11:57:05,686 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220194.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:18,850 INFO [train.py:904] (5/8) Epoch 22, batch 7050, loss[loss=0.2131, simple_loss=0.3008, pruned_loss=0.06266, over 16715.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2956, pruned_loss=0.06118, over 3061434.84 frames. ], batch size: 134, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:57:21,957 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 2.577e+02 3.204e+02 3.781e+02 6.030e+02, threshold=6.408e+02, percent-clipped=0.0 2023-05-01 11:57:22,858 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 11:57:35,266 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:59,360 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220229.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:58:23,578 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3256, 2.5612, 2.1480, 2.3264, 2.9707, 2.5586, 2.9466, 3.1296], device='cuda:5'), covar=tensor([0.0143, 0.0406, 0.0522, 0.0447, 0.0246, 0.0393, 0.0222, 0.0250], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0231, 0.0223, 0.0223, 0.0232, 0.0230, 0.0231, 0.0227], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 11:58:36,514 INFO [train.py:904] (5/8) Epoch 22, batch 7100, loss[loss=0.1936, simple_loss=0.2818, pruned_loss=0.05271, over 16443.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2939, pruned_loss=0.06032, over 3070531.40 frames. ], batch size: 75, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:58:45,440 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:58:45,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5688, 3.6126, 2.8303, 2.2451, 2.3859, 2.4051, 3.8241, 3.2076], device='cuda:5'), covar=tensor([0.2867, 0.0626, 0.1745, 0.2636, 0.2563, 0.2041, 0.0422, 0.1331], device='cuda:5'), in_proj_covar=tensor([0.0325, 0.0267, 0.0304, 0.0313, 0.0297, 0.0258, 0.0293, 0.0335], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 11:59:35,826 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:54,486 INFO [train.py:904] (5/8) Epoch 22, batch 7150, loss[loss=0.1943, simple_loss=0.2797, pruned_loss=0.05438, over 17047.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2919, pruned_loss=0.06009, over 3068782.80 frames. ], batch size: 55, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:59:58,137 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.701e+02 3.416e+02 4.420e+02 1.024e+03, threshold=6.833e+02, percent-clipped=4.0 2023-05-01 12:00:18,200 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 2023-05-01 12:00:20,908 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:01:07,961 INFO [train.py:904] (5/8) Epoch 22, batch 7200, loss[loss=0.1743, simple_loss=0.2672, pruned_loss=0.04071, over 16883.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2894, pruned_loss=0.05794, over 3073348.15 frames. ], batch size: 96, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:01:10,339 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6197, 2.1600, 1.7892, 1.9065, 2.4960, 2.1410, 2.3433, 2.6878], device='cuda:5'), covar=tensor([0.0203, 0.0465, 0.0587, 0.0520, 0.0286, 0.0405, 0.0221, 0.0272], device='cuda:5'), in_proj_covar=tensor([0.0208, 0.0233, 0.0225, 0.0224, 0.0233, 0.0232, 0.0232, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:01:13,674 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 12:01:40,345 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5988, 4.6508, 4.4715, 4.1606, 4.1819, 4.5704, 4.3821, 4.2716], device='cuda:5'), covar=tensor([0.0595, 0.0605, 0.0300, 0.0299, 0.0872, 0.0530, 0.0468, 0.0628], device='cuda:5'), in_proj_covar=tensor([0.0286, 0.0423, 0.0337, 0.0334, 0.0342, 0.0387, 0.0232, 0.0401], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:02:28,589 INFO [train.py:904] (5/8) Epoch 22, batch 7250, loss[loss=0.1896, simple_loss=0.2637, pruned_loss=0.05771, over 16568.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2869, pruned_loss=0.05651, over 3098522.17 frames. ], batch size: 62, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:30,893 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.548e+02 3.054e+02 3.716e+02 9.083e+02, threshold=6.108e+02, percent-clipped=2.0 2023-05-01 12:02:53,062 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220419.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:30,265 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220443.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:45,062 INFO [train.py:904] (5/8) Epoch 22, batch 7300, loss[loss=0.1874, simple_loss=0.2823, pruned_loss=0.0462, over 16697.00 frames. ], tot_loss[loss=0.2, simple_loss=0.287, pruned_loss=0.05653, over 3099478.50 frames. ], batch size: 89, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:03:47,418 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220454.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:04:04,325 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9254, 2.2029, 1.8261, 2.0163, 2.5656, 2.1628, 2.5408, 2.7609], device='cuda:5'), covar=tensor([0.0178, 0.0438, 0.0602, 0.0485, 0.0281, 0.0458, 0.0217, 0.0289], device='cuda:5'), in_proj_covar=tensor([0.0207, 0.0232, 0.0224, 0.0223, 0.0232, 0.0231, 0.0231, 0.0227], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:04:07,574 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:05:02,420 INFO [train.py:904] (5/8) Epoch 22, batch 7350, loss[loss=0.199, simple_loss=0.2812, pruned_loss=0.05839, over 16258.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2878, pruned_loss=0.05735, over 3080908.82 frames. ], batch size: 165, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:05:05,571 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 2.713e+02 3.260e+02 3.989e+02 1.125e+03, threshold=6.520e+02, percent-clipped=5.0 2023-05-01 12:05:18,404 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220513.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:20,498 INFO [train.py:904] (5/8) Epoch 22, batch 7400, loss[loss=0.2342, simple_loss=0.301, pruned_loss=0.08369, over 11352.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2884, pruned_loss=0.05771, over 3097911.98 frames. ], batch size: 246, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:06:25,372 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220555.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:31,531 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0635, 2.4335, 2.6029, 1.8876, 2.6638, 2.7812, 2.3835, 2.3438], device='cuda:5'), covar=tensor([0.0720, 0.0242, 0.0255, 0.1015, 0.0138, 0.0300, 0.0452, 0.0438], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0108, 0.0098, 0.0138, 0.0080, 0.0125, 0.0129, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:06:35,407 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:38,279 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1585, 3.3935, 3.5563, 2.0543, 3.1338, 2.3438, 3.6222, 3.6648], device='cuda:5'), covar=tensor([0.0229, 0.0769, 0.0570, 0.2123, 0.0794, 0.0999, 0.0578, 0.0948], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0165, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:07:05,397 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7158, 3.8487, 2.5391, 4.4443, 2.9529, 4.3335, 2.5637, 3.0734], device='cuda:5'), covar=tensor([0.0296, 0.0383, 0.1569, 0.0220, 0.0825, 0.0636, 0.1515, 0.0805], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0160, 0.0174, 0.0215, 0.0201, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:07:13,081 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:07:41,489 INFO [train.py:904] (5/8) Epoch 22, batch 7450, loss[loss=0.2557, simple_loss=0.321, pruned_loss=0.09523, over 11627.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2894, pruned_loss=0.0591, over 3066773.58 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:07:43,940 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.745e+02 3.256e+02 3.859e+02 7.938e+02, threshold=6.512e+02, percent-clipped=1.0 2023-05-01 12:08:01,213 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:08:04,662 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220616.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 12:08:07,809 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8024, 4.8512, 5.2485, 5.1764, 5.2158, 4.8853, 4.8120, 4.6295], device='cuda:5'), covar=tensor([0.0324, 0.0513, 0.0323, 0.0455, 0.0486, 0.0372, 0.0944, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0406, 0.0450, 0.0435, 0.0405, 0.0484, 0.0460, 0.0544, 0.0368], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 12:09:00,769 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220651.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:09:03,152 INFO [train.py:904] (5/8) Epoch 22, batch 7500, loss[loss=0.1941, simple_loss=0.2809, pruned_loss=0.05369, over 16496.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2902, pruned_loss=0.05851, over 3043059.40 frames. ], batch size: 146, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:10,738 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5555, 4.8240, 4.6254, 4.6275, 4.3281, 4.3235, 4.3219, 4.8846], device='cuda:5'), covar=tensor([0.1234, 0.0867, 0.1050, 0.0937, 0.0883, 0.1324, 0.1168, 0.0919], device='cuda:5'), in_proj_covar=tensor([0.0671, 0.0818, 0.0677, 0.0623, 0.0516, 0.0530, 0.0685, 0.0641], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:10:17,011 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1667, 4.2518, 4.0498, 3.7893, 3.7734, 4.1802, 3.8433, 3.9095], device='cuda:5'), covar=tensor([0.0643, 0.0542, 0.0346, 0.0338, 0.0778, 0.0474, 0.0941, 0.0683], device='cuda:5'), in_proj_covar=tensor([0.0287, 0.0423, 0.0337, 0.0333, 0.0342, 0.0387, 0.0232, 0.0402], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:10:21,164 INFO [train.py:904] (5/8) Epoch 22, batch 7550, loss[loss=0.1898, simple_loss=0.2789, pruned_loss=0.05033, over 16792.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.29, pruned_loss=0.05924, over 3033644.12 frames. ], batch size: 89, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:24,493 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.918e+02 3.506e+02 4.599e+02 8.060e+02, threshold=7.013e+02, percent-clipped=6.0 2023-05-01 12:10:36,080 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:11:23,263 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:11:38,543 INFO [train.py:904] (5/8) Epoch 22, batch 7600, loss[loss=0.2045, simple_loss=0.2921, pruned_loss=0.05838, over 15252.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.289, pruned_loss=0.05915, over 3038434.03 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:11:40,732 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220754.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:37,571 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220791.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:43,565 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5802, 4.5563, 4.3878, 3.6338, 4.4623, 1.7177, 4.2131, 4.0586], device='cuda:5'), covar=tensor([0.0105, 0.0100, 0.0197, 0.0363, 0.0105, 0.2948, 0.0140, 0.0270], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0153, 0.0196, 0.0176, 0.0173, 0.0206, 0.0185, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:12:55,480 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220802.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:56,376 INFO [train.py:904] (5/8) Epoch 22, batch 7650, loss[loss=0.1876, simple_loss=0.2751, pruned_loss=0.05006, over 17117.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2889, pruned_loss=0.05961, over 3041753.72 frames. ], batch size: 47, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:12:59,168 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.956e+02 2.818e+02 3.454e+02 4.152e+02 9.180e+02, threshold=6.907e+02, percent-clipped=1.0 2023-05-01 12:13:20,388 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6724, 2.3514, 2.1041, 3.1615, 1.9235, 3.5269, 1.5137, 2.5451], device='cuda:5'), covar=tensor([0.1533, 0.0984, 0.1519, 0.0223, 0.0199, 0.0440, 0.1963, 0.0935], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0176, 0.0197, 0.0191, 0.0208, 0.0217, 0.0204, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:13:27,062 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:13:34,785 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 12:13:34,863 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 12:14:05,576 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3449, 3.4721, 3.6141, 3.5883, 3.6154, 3.4289, 3.4647, 3.5027], device='cuda:5'), covar=tensor([0.0427, 0.0792, 0.0475, 0.0493, 0.0532, 0.0604, 0.0845, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0450, 0.0434, 0.0404, 0.0482, 0.0458, 0.0542, 0.0367], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 12:14:11,878 INFO [train.py:904] (5/8) Epoch 22, batch 7700, loss[loss=0.1703, simple_loss=0.2627, pruned_loss=0.03894, over 16531.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2883, pruned_loss=0.05918, over 3062175.74 frames. ], batch size: 68, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:14:25,451 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3358, 3.3966, 2.0484, 3.7362, 2.5373, 3.7419, 2.1225, 2.7101], device='cuda:5'), covar=tensor([0.0328, 0.0437, 0.1824, 0.0253, 0.0919, 0.0699, 0.1741, 0.0958], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0161, 0.0175, 0.0216, 0.0201, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:14:25,976 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 12:14:58,255 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:01,861 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220885.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:07,365 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 12:15:11,919 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6115, 2.4243, 2.3250, 4.2898, 3.0174, 3.9981, 1.5351, 2.8192], device='cuda:5'), covar=tensor([0.1652, 0.1024, 0.1458, 0.0204, 0.0281, 0.0405, 0.2020, 0.0915], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0176, 0.0197, 0.0192, 0.0208, 0.0217, 0.0205, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:15:29,150 INFO [train.py:904] (5/8) Epoch 22, batch 7750, loss[loss=0.2011, simple_loss=0.2916, pruned_loss=0.05531, over 16967.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2882, pruned_loss=0.05868, over 3074053.96 frames. ], batch size: 41, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:15:30,844 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220904.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:32,047 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.706e+02 3.407e+02 4.136e+02 9.876e+02, threshold=6.815e+02, percent-clipped=5.0 2023-05-01 12:15:40,124 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220911.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 12:15:45,828 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:16:12,244 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0087, 3.4323, 3.4930, 2.2317, 3.2365, 3.4954, 3.2045, 2.0097], device='cuda:5'), covar=tensor([0.0573, 0.0078, 0.0067, 0.0429, 0.0115, 0.0129, 0.0119, 0.0458], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0083, 0.0084, 0.0133, 0.0097, 0.0109, 0.0094, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 12:16:15,125 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:16:37,992 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0817, 2.1575, 2.2652, 3.5921, 2.1005, 2.4920, 2.2750, 2.2829], device='cuda:5'), covar=tensor([0.1395, 0.3546, 0.2917, 0.0603, 0.4191, 0.2485, 0.3614, 0.3224], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0445, 0.0364, 0.0324, 0.0435, 0.0512, 0.0418, 0.0519], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:16:42,228 INFO [train.py:904] (5/8) Epoch 22, batch 7800, loss[loss=0.2481, simple_loss=0.3121, pruned_loss=0.09205, over 11359.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2891, pruned_loss=0.05921, over 3068334.70 frames. ], batch size: 246, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:16:56,583 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220962.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:00,393 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:55,733 INFO [train.py:904] (5/8) Epoch 22, batch 7850, loss[loss=0.1955, simple_loss=0.2888, pruned_loss=0.05114, over 16535.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2898, pruned_loss=0.05871, over 3081240.78 frames. ], batch size: 75, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:17:58,015 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.766e+02 3.554e+02 4.297e+02 6.446e+02, threshold=7.108e+02, percent-clipped=0.0 2023-05-01 12:18:02,918 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:18:53,820 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 12:19:09,460 INFO [train.py:904] (5/8) Epoch 22, batch 7900, loss[loss=0.205, simple_loss=0.29, pruned_loss=0.05997, over 16737.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2893, pruned_loss=0.05837, over 3080205.20 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:19:36,748 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9080, 2.0764, 2.5155, 2.8502, 2.7244, 3.3865, 2.1925, 3.2793], device='cuda:5'), covar=tensor([0.0241, 0.0465, 0.0348, 0.0310, 0.0336, 0.0141, 0.0524, 0.0150], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0189, 0.0176, 0.0181, 0.0193, 0.0150, 0.0193, 0.0149], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:20:12,850 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2900, 2.3984, 2.4132, 3.9315, 2.2466, 2.7053, 2.4333, 2.5011], device='cuda:5'), covar=tensor([0.1295, 0.3327, 0.2613, 0.0517, 0.3901, 0.2321, 0.3239, 0.3106], device='cuda:5'), in_proj_covar=tensor([0.0397, 0.0446, 0.0363, 0.0324, 0.0435, 0.0513, 0.0418, 0.0519], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:20:27,100 INFO [train.py:904] (5/8) Epoch 22, batch 7950, loss[loss=0.2247, simple_loss=0.2896, pruned_loss=0.07991, over 11286.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2894, pruned_loss=0.05856, over 3098519.88 frames. ], batch size: 246, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:20:32,014 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.674e+02 3.039e+02 3.526e+02 5.687e+02, threshold=6.078e+02, percent-clipped=0.0 2023-05-01 12:21:29,017 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 12:21:41,836 INFO [train.py:904] (5/8) Epoch 22, batch 8000, loss[loss=0.2004, simple_loss=0.2866, pruned_loss=0.05709, over 15272.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2904, pruned_loss=0.05938, over 3093271.76 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:22:02,430 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 12:22:19,116 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:22:55,176 INFO [train.py:904] (5/8) Epoch 22, batch 8050, loss[loss=0.197, simple_loss=0.2844, pruned_loss=0.05482, over 16792.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2903, pruned_loss=0.05889, over 3110020.94 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:23:01,257 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.816e+02 3.381e+02 4.131e+02 1.220e+03, threshold=6.763e+02, percent-clipped=7.0 2023-05-01 12:23:01,914 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8003, 2.7271, 2.8697, 2.1798, 2.7145, 2.1308, 2.7149, 2.9426], device='cuda:5'), covar=tensor([0.0280, 0.0780, 0.0490, 0.1656, 0.0756, 0.0931, 0.0554, 0.0710], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0152, 0.0144, 0.0129, 0.0142, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:23:07,976 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:23:09,436 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 12:23:15,156 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0364, 5.7371, 5.8811, 5.5329, 5.6529, 6.1579, 5.5747, 5.3590], device='cuda:5'), covar=tensor([0.0851, 0.1492, 0.1841, 0.1757, 0.2050, 0.0875, 0.1575, 0.2217], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0590, 0.0654, 0.0491, 0.0653, 0.0683, 0.0515, 0.0659], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 12:23:20,857 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6215, 2.4173, 2.1108, 3.6147, 2.1614, 3.7135, 1.4463, 2.5695], device='cuda:5'), covar=tensor([0.1489, 0.0987, 0.1596, 0.0243, 0.0255, 0.0506, 0.1900, 0.1021], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0190, 0.0207, 0.0215, 0.0202, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:24:08,794 INFO [train.py:904] (5/8) Epoch 22, batch 8100, loss[loss=0.1876, simple_loss=0.2842, pruned_loss=0.04547, over 16762.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2896, pruned_loss=0.05835, over 3110686.39 frames. ], batch size: 83, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:24:17,306 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:24:19,981 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221260.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:25:24,347 INFO [train.py:904] (5/8) Epoch 22, batch 8150, loss[loss=0.1966, simple_loss=0.2777, pruned_loss=0.05778, over 16136.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2883, pruned_loss=0.05841, over 3086731.39 frames. ], batch size: 165, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:25:31,012 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.655e+02 3.146e+02 3.866e+02 6.459e+02, threshold=6.292e+02, percent-clipped=0.0 2023-05-01 12:25:32,096 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221307.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:25:35,272 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3899, 2.9095, 2.6583, 2.3052, 2.2419, 2.2690, 2.9986, 2.8255], device='cuda:5'), covar=tensor([0.2618, 0.0723, 0.1684, 0.2575, 0.2534, 0.2180, 0.0569, 0.1377], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0267, 0.0304, 0.0313, 0.0297, 0.0260, 0.0295, 0.0335], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 12:25:49,602 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4969, 2.2131, 1.8182, 2.0503, 2.5017, 2.2248, 2.3170, 2.6488], device='cuda:5'), covar=tensor([0.0219, 0.0409, 0.0555, 0.0414, 0.0245, 0.0379, 0.0241, 0.0267], device='cuda:5'), in_proj_covar=tensor([0.0206, 0.0231, 0.0223, 0.0222, 0.0231, 0.0229, 0.0231, 0.0225], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:26:41,093 INFO [train.py:904] (5/8) Epoch 22, batch 8200, loss[loss=0.1859, simple_loss=0.2787, pruned_loss=0.04655, over 15393.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2849, pruned_loss=0.05694, over 3104890.10 frames. ], batch size: 191, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:26:44,115 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221355.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:27:16,686 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-01 12:27:58,676 INFO [train.py:904] (5/8) Epoch 22, batch 8250, loss[loss=0.1697, simple_loss=0.2663, pruned_loss=0.0366, over 15258.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2841, pruned_loss=0.05498, over 3062109.91 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:28:05,608 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.638e+02 3.060e+02 3.657e+02 7.056e+02, threshold=6.120e+02, percent-clipped=1.0 2023-05-01 12:28:11,648 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-05-01 12:29:10,443 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7085, 3.0909, 3.4402, 2.1008, 2.9426, 2.2272, 3.3273, 3.2459], device='cuda:5'), covar=tensor([0.0248, 0.0800, 0.0492, 0.2007, 0.0780, 0.1010, 0.0533, 0.0860], device='cuda:5'), in_proj_covar=tensor([0.0153, 0.0162, 0.0164, 0.0152, 0.0143, 0.0128, 0.0140, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:29:17,816 INFO [train.py:904] (5/8) Epoch 22, batch 8300, loss[loss=0.1706, simple_loss=0.2495, pruned_loss=0.04588, over 11920.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2812, pruned_loss=0.05169, over 3067460.89 frames. ], batch size: 246, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:29:57,602 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:30:11,742 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5835, 3.6439, 3.4534, 3.0810, 3.2166, 3.5549, 3.3399, 3.3904], device='cuda:5'), covar=tensor([0.0579, 0.0640, 0.0309, 0.0296, 0.0515, 0.0462, 0.1445, 0.0478], device='cuda:5'), in_proj_covar=tensor([0.0287, 0.0424, 0.0337, 0.0332, 0.0342, 0.0388, 0.0232, 0.0402], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:30:38,300 INFO [train.py:904] (5/8) Epoch 22, batch 8350, loss[loss=0.2105, simple_loss=0.313, pruned_loss=0.05397, over 15272.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.282, pruned_loss=0.05065, over 3061883.80 frames. ], batch size: 191, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:30:43,703 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 2.123e+02 2.436e+02 3.235e+02 5.612e+02, threshold=4.873e+02, percent-clipped=0.0 2023-05-01 12:30:44,311 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:15,660 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221526.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:48,103 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9436, 1.8014, 1.5739, 1.4546, 1.8766, 1.5801, 1.6126, 1.9078], device='cuda:5'), covar=tensor([0.0256, 0.0403, 0.0598, 0.0479, 0.0313, 0.0386, 0.0217, 0.0355], device='cuda:5'), in_proj_covar=tensor([0.0204, 0.0229, 0.0222, 0.0221, 0.0230, 0.0228, 0.0228, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:31:56,147 INFO [train.py:904] (5/8) Epoch 22, batch 8400, loss[loss=0.1471, simple_loss=0.2363, pruned_loss=0.02899, over 12131.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2789, pruned_loss=0.0486, over 3063509.86 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:32:08,943 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:32:21,112 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:33:17,380 INFO [train.py:904] (5/8) Epoch 22, batch 8450, loss[loss=0.1678, simple_loss=0.2654, pruned_loss=0.03512, over 16896.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2764, pruned_loss=0.04642, over 3071180.26 frames. ], batch size: 116, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:33:24,324 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.610e+02 2.296e+02 2.720e+02 3.439e+02 5.542e+02, threshold=5.440e+02, percent-clipped=2.0 2023-05-01 12:33:26,284 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221608.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:34:38,841 INFO [train.py:904] (5/8) Epoch 22, batch 8500, loss[loss=0.1861, simple_loss=0.2609, pruned_loss=0.05559, over 11953.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2731, pruned_loss=0.04457, over 3058928.66 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:34:46,745 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3509, 3.2580, 3.2595, 3.4927, 3.5246, 3.2699, 3.4860, 3.5573], device='cuda:5'), covar=tensor([0.1660, 0.1376, 0.1767, 0.0925, 0.1006, 0.3565, 0.1416, 0.1257], device='cuda:5'), in_proj_covar=tensor([0.0622, 0.0768, 0.0891, 0.0775, 0.0592, 0.0620, 0.0644, 0.0748], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:34:47,989 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9523, 2.7795, 2.6410, 1.9974, 2.5273, 2.7779, 2.6399, 1.9778], device='cuda:5'), covar=tensor([0.0399, 0.0085, 0.0074, 0.0312, 0.0135, 0.0109, 0.0111, 0.0443], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0132, 0.0096, 0.0108, 0.0093, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 12:34:48,367 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-01 12:36:02,542 INFO [train.py:904] (5/8) Epoch 22, batch 8550, loss[loss=0.1774, simple_loss=0.2625, pruned_loss=0.04613, over 12373.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2708, pruned_loss=0.04388, over 3015173.99 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:36:10,061 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.073e+02 2.527e+02 2.950e+02 5.682e+02, threshold=5.053e+02, percent-clipped=1.0 2023-05-01 12:37:41,427 INFO [train.py:904] (5/8) Epoch 22, batch 8600, loss[loss=0.1752, simple_loss=0.2601, pruned_loss=0.0451, over 12508.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2709, pruned_loss=0.0426, over 3034840.48 frames. ], batch size: 246, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:37:53,153 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:39:21,624 INFO [train.py:904] (5/8) Epoch 22, batch 8650, loss[loss=0.1454, simple_loss=0.2504, pruned_loss=0.02016, over 16894.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2692, pruned_loss=0.04121, over 3030860.01 frames. ], batch size: 102, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:39:30,561 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.228e+02 2.506e+02 3.024e+02 6.242e+02, threshold=5.012e+02, percent-clipped=2.0 2023-05-01 12:39:58,981 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:40:29,418 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-01 12:41:06,653 INFO [train.py:904] (5/8) Epoch 22, batch 8700, loss[loss=0.1549, simple_loss=0.2457, pruned_loss=0.03204, over 16516.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2673, pruned_loss=0.04034, over 3072214.08 frames. ], batch size: 62, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:41:27,802 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221863.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:42:16,910 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:42:34,631 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9076, 2.1132, 2.3239, 3.1652, 2.1871, 2.3150, 2.3462, 2.2044], device='cuda:5'), covar=tensor([0.1315, 0.3697, 0.2791, 0.0694, 0.4388, 0.2594, 0.3248, 0.3756], device='cuda:5'), in_proj_covar=tensor([0.0392, 0.0439, 0.0360, 0.0318, 0.0429, 0.0505, 0.0411, 0.0512], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:42:42,876 INFO [train.py:904] (5/8) Epoch 22, batch 8750, loss[loss=0.1656, simple_loss=0.2747, pruned_loss=0.02823, over 16889.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2666, pruned_loss=0.03982, over 3046189.00 frames. ], batch size: 102, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:42:53,177 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.058e+02 2.561e+02 3.126e+02 5.404e+02, threshold=5.122e+02, percent-clipped=2.0 2023-05-01 12:43:06,572 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9454, 3.8674, 4.0763, 4.1395, 4.2717, 3.8326, 4.2488, 4.2916], device='cuda:5'), covar=tensor([0.1765, 0.1309, 0.1257, 0.0777, 0.0614, 0.1727, 0.0853, 0.0697], device='cuda:5'), in_proj_covar=tensor([0.0620, 0.0766, 0.0885, 0.0773, 0.0590, 0.0617, 0.0642, 0.0744], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:44:10,303 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2224, 4.3030, 4.1266, 3.8590, 3.8720, 4.2215, 3.9112, 3.9833], device='cuda:5'), covar=tensor([0.0622, 0.0606, 0.0325, 0.0308, 0.0812, 0.0483, 0.0867, 0.0640], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0416, 0.0332, 0.0329, 0.0336, 0.0382, 0.0229, 0.0396], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:44:31,249 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:44:33,694 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3390, 2.2616, 2.2928, 4.0707, 2.1804, 2.6505, 2.3406, 2.4101], device='cuda:5'), covar=tensor([0.1236, 0.3576, 0.3082, 0.0498, 0.4354, 0.2643, 0.3678, 0.3565], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0440, 0.0361, 0.0319, 0.0430, 0.0506, 0.0412, 0.0513], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:44:34,635 INFO [train.py:904] (5/8) Epoch 22, batch 8800, loss[loss=0.1516, simple_loss=0.2523, pruned_loss=0.02548, over 16504.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2648, pruned_loss=0.03848, over 3048496.61 frames. ], batch size: 75, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:45:32,109 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 12:45:52,378 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.61 vs. limit=5.0 2023-05-01 12:46:22,304 INFO [train.py:904] (5/8) Epoch 22, batch 8850, loss[loss=0.1855, simple_loss=0.2868, pruned_loss=0.04207, over 16675.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2669, pruned_loss=0.03763, over 3044118.87 frames. ], batch size: 134, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:46:28,907 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.701e+02 2.177e+02 2.540e+02 3.054e+02 9.241e+02, threshold=5.079e+02, percent-clipped=4.0 2023-05-01 12:46:35,374 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:47:34,907 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 12:48:07,872 INFO [train.py:904] (5/8) Epoch 22, batch 8900, loss[loss=0.168, simple_loss=0.2691, pruned_loss=0.03346, over 16173.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2677, pruned_loss=0.03724, over 3058641.33 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:48:43,955 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222070.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:50:01,986 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7917, 5.0914, 4.8836, 4.9009, 4.6513, 4.6319, 4.5110, 5.1764], device='cuda:5'), covar=tensor([0.1337, 0.0945, 0.1057, 0.0841, 0.0812, 0.0964, 0.1198, 0.0866], device='cuda:5'), in_proj_covar=tensor([0.0655, 0.0796, 0.0659, 0.0604, 0.0502, 0.0516, 0.0668, 0.0624], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:50:11,684 INFO [train.py:904] (5/8) Epoch 22, batch 8950, loss[loss=0.1541, simple_loss=0.2507, pruned_loss=0.02875, over 16837.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2672, pruned_loss=0.0374, over 3093061.17 frames. ], batch size: 76, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:50:23,611 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.055e+02 2.537e+02 3.242e+02 5.018e+02, threshold=5.074e+02, percent-clipped=0.0 2023-05-01 12:50:37,698 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:51:06,756 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-01 12:51:10,087 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7184, 2.7123, 2.4539, 4.4184, 2.9442, 4.1539, 1.4208, 3.0277], device='cuda:5'), covar=tensor([0.1375, 0.0819, 0.1274, 0.0111, 0.0138, 0.0326, 0.1791, 0.0707], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0184, 0.0202, 0.0212, 0.0201, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 12:52:03,809 INFO [train.py:904] (5/8) Epoch 22, batch 9000, loss[loss=0.1409, simple_loss=0.2366, pruned_loss=0.02264, over 16779.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2638, pruned_loss=0.03555, over 3109301.59 frames. ], batch size: 83, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:52:03,810 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 12:52:13,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5340, 3.1721, 3.5514, 5.2692, 3.5553, 3.4293, 3.4867, 3.3841], device='cuda:5'), covar=tensor([0.0765, 0.2713, 0.2198, 0.0245, 0.2881, 0.2131, 0.2745, 0.2913], device='cuda:5'), in_proj_covar=tensor([0.0394, 0.0441, 0.0363, 0.0319, 0.0431, 0.0507, 0.0414, 0.0514], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 12:52:14,711 INFO [train.py:938] (5/8) Epoch 22, validation: loss=0.1453, simple_loss=0.2492, pruned_loss=0.0207, over 944034.00 frames. 2023-05-01 12:52:14,713 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 12:52:36,648 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:52:48,821 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1444, 3.1761, 1.7652, 3.4834, 2.3152, 3.4182, 1.9269, 2.5810], device='cuda:5'), covar=tensor([0.0317, 0.0410, 0.1908, 0.0269, 0.0907, 0.0680, 0.1848, 0.0831], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0171, 0.0188, 0.0156, 0.0171, 0.0209, 0.0198, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 12:53:02,481 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0241, 3.2769, 3.2601, 2.1819, 2.9750, 3.3022, 3.1039, 1.8570], device='cuda:5'), covar=tensor([0.0528, 0.0054, 0.0055, 0.0412, 0.0110, 0.0074, 0.0089, 0.0533], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0081, 0.0081, 0.0131, 0.0095, 0.0106, 0.0092, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 12:53:22,212 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:53:58,640 INFO [train.py:904] (5/8) Epoch 22, batch 9050, loss[loss=0.1593, simple_loss=0.2483, pruned_loss=0.03512, over 16285.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2647, pruned_loss=0.03637, over 3097948.51 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:54:04,300 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222205.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:54:09,045 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.330e+02 2.762e+02 3.325e+02 5.542e+02, threshold=5.524e+02, percent-clipped=4.0 2023-05-01 12:54:17,178 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:54:23,372 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7169, 4.9635, 5.0466, 4.8684, 4.9329, 5.4429, 4.9911, 4.7236], device='cuda:5'), covar=tensor([0.1145, 0.1563, 0.1842, 0.1901, 0.2332, 0.0926, 0.1517, 0.2365], device='cuda:5'), in_proj_covar=tensor([0.0394, 0.0569, 0.0630, 0.0473, 0.0627, 0.0660, 0.0499, 0.0636], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 12:55:31,430 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:55:31,633 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222246.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 12:55:44,667 INFO [train.py:904] (5/8) Epoch 22, batch 9100, loss[loss=0.1645, simple_loss=0.2517, pruned_loss=0.03865, over 11893.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.265, pruned_loss=0.03721, over 3107571.62 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:56:11,363 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222266.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:57:16,558 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-05-01 12:57:42,766 INFO [train.py:904] (5/8) Epoch 22, batch 9150, loss[loss=0.1661, simple_loss=0.2602, pruned_loss=0.036, over 16360.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2654, pruned_loss=0.03703, over 3100534.79 frames. ], batch size: 166, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:57:53,764 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.177e+02 2.613e+02 3.280e+02 4.905e+02, threshold=5.227e+02, percent-clipped=0.0 2023-05-01 12:58:38,733 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-01 12:59:28,943 INFO [train.py:904] (5/8) Epoch 22, batch 9200, loss[loss=0.1462, simple_loss=0.2379, pruned_loss=0.02724, over 12641.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2608, pruned_loss=0.03585, over 3105972.16 frames. ], batch size: 250, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:59:52,289 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:01:05,912 INFO [train.py:904] (5/8) Epoch 22, batch 9250, loss[loss=0.1376, simple_loss=0.2264, pruned_loss=0.02438, over 12317.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.26, pruned_loss=0.03584, over 3094057.24 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:01:16,245 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.130e+02 2.525e+02 3.050e+02 5.911e+02, threshold=5.049e+02, percent-clipped=2.0 2023-05-01 13:01:28,895 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:02:56,980 INFO [train.py:904] (5/8) Epoch 22, batch 9300, loss[loss=0.1409, simple_loss=0.2406, pruned_loss=0.02061, over 16922.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2587, pruned_loss=0.03546, over 3093093.85 frames. ], batch size: 102, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:03:17,567 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:03:40,912 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0352, 4.1350, 3.9415, 3.6592, 3.6887, 4.0762, 3.7006, 3.8364], device='cuda:5'), covar=tensor([0.0576, 0.0567, 0.0341, 0.0310, 0.0716, 0.0522, 0.0934, 0.0579], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0410, 0.0328, 0.0323, 0.0331, 0.0377, 0.0226, 0.0390], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:04:35,084 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1525, 3.5217, 3.5061, 2.3986, 3.1731, 3.5270, 3.3792, 2.1238], device='cuda:5'), covar=tensor([0.0566, 0.0049, 0.0058, 0.0401, 0.0112, 0.0105, 0.0083, 0.0475], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0081, 0.0082, 0.0131, 0.0095, 0.0106, 0.0092, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 13:04:40,774 INFO [train.py:904] (5/8) Epoch 22, batch 9350, loss[loss=0.1806, simple_loss=0.2766, pruned_loss=0.04231, over 15424.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2591, pruned_loss=0.03552, over 3094171.94 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:04:49,917 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 1.994e+02 2.436e+02 3.069e+02 5.740e+02, threshold=4.871e+02, percent-clipped=2.0 2023-05-01 13:05:01,845 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-05-01 13:05:23,329 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:05:58,199 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222541.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:06:02,164 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3432, 2.3455, 2.2918, 4.1872, 2.2192, 2.7115, 2.4276, 2.5301], device='cuda:5'), covar=tensor([0.1170, 0.3528, 0.3114, 0.0453, 0.4172, 0.2681, 0.3544, 0.3289], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0438, 0.0361, 0.0318, 0.0429, 0.0504, 0.0411, 0.0512], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:06:08,647 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:20,192 INFO [train.py:904] (5/8) Epoch 22, batch 9400, loss[loss=0.1723, simple_loss=0.2816, pruned_loss=0.03147, over 16766.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2598, pruned_loss=0.03565, over 3093013.54 frames. ], batch size: 83, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:06:38,158 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:55,343 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:07:24,545 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222584.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:07:43,445 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:07:54,921 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5649, 3.5097, 3.4993, 2.7698, 3.4717, 2.1121, 3.2267, 2.8656], device='cuda:5'), covar=tensor([0.0118, 0.0123, 0.0163, 0.0170, 0.0095, 0.2184, 0.0118, 0.0227], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0151, 0.0191, 0.0169, 0.0170, 0.0203, 0.0181, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:08:00,583 INFO [train.py:904] (5/8) Epoch 22, batch 9450, loss[loss=0.1746, simple_loss=0.2553, pruned_loss=0.04691, over 12212.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2611, pruned_loss=0.03576, over 3076232.58 frames. ], batch size: 250, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:08:08,418 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 2.238e+02 2.730e+02 3.189e+02 9.782e+02, threshold=5.460e+02, percent-clipped=5.0 2023-05-01 13:08:28,505 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8133, 1.3878, 1.7409, 1.7081, 1.8876, 1.9258, 1.7019, 1.8790], device='cuda:5'), covar=tensor([0.0294, 0.0421, 0.0249, 0.0331, 0.0349, 0.0224, 0.0453, 0.0158], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0186, 0.0173, 0.0177, 0.0190, 0.0147, 0.0190, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:08:56,831 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:09:40,759 INFO [train.py:904] (5/8) Epoch 22, batch 9500, loss[loss=0.1623, simple_loss=0.2556, pruned_loss=0.03454, over 15553.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2599, pruned_loss=0.03535, over 3060225.51 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:09:50,502 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2915, 1.7628, 2.0018, 2.2459, 2.3618, 2.5058, 1.8833, 2.4355], device='cuda:5'), covar=tensor([0.0234, 0.0480, 0.0338, 0.0357, 0.0356, 0.0207, 0.0508, 0.0162], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0187, 0.0173, 0.0177, 0.0190, 0.0147, 0.0190, 0.0145], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:10:06,981 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:11:22,406 INFO [train.py:904] (5/8) Epoch 22, batch 9550, loss[loss=0.179, simple_loss=0.2796, pruned_loss=0.03919, over 16216.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2604, pruned_loss=0.03561, over 3066033.47 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:11:34,519 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 2.117e+02 2.369e+02 3.001e+02 4.954e+02, threshold=4.737e+02, percent-clipped=0.0 2023-05-01 13:11:44,579 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:13:00,663 INFO [train.py:904] (5/8) Epoch 22, batch 9600, loss[loss=0.1845, simple_loss=0.2892, pruned_loss=0.03985, over 16181.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2623, pruned_loss=0.03664, over 3067558.36 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:14:03,256 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7947, 3.7831, 3.8859, 3.7410, 3.9069, 4.2647, 3.9396, 3.6689], device='cuda:5'), covar=tensor([0.1998, 0.2258, 0.2286, 0.2573, 0.2580, 0.1690, 0.1515, 0.2476], device='cuda:5'), in_proj_covar=tensor([0.0387, 0.0563, 0.0625, 0.0465, 0.0621, 0.0651, 0.0493, 0.0625], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 13:14:44,338 INFO [train.py:904] (5/8) Epoch 22, batch 9650, loss[loss=0.1826, simple_loss=0.2804, pruned_loss=0.04234, over 17003.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2641, pruned_loss=0.03713, over 3060640.95 frames. ], batch size: 109, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:14:58,749 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.257e+02 2.597e+02 3.466e+02 1.012e+03, threshold=5.195e+02, percent-clipped=6.0 2023-05-01 13:16:03,748 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222841.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:16:27,407 INFO [train.py:904] (5/8) Epoch 22, batch 9700, loss[loss=0.1759, simple_loss=0.2689, pruned_loss=0.04145, over 15310.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2631, pruned_loss=0.03727, over 3048224.84 frames. ], batch size: 190, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:16:43,338 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:17:23,135 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222879.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:17:42,990 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:03,841 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1327, 1.5413, 1.9196, 2.0662, 2.2100, 2.3133, 1.8045, 2.2146], device='cuda:5'), covar=tensor([0.0280, 0.0554, 0.0328, 0.0401, 0.0363, 0.0244, 0.0572, 0.0191], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0187, 0.0175, 0.0178, 0.0192, 0.0148, 0.0191, 0.0146], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:18:08,547 INFO [train.py:904] (5/8) Epoch 22, batch 9750, loss[loss=0.1519, simple_loss=0.2549, pruned_loss=0.02449, over 16886.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2619, pruned_loss=0.03716, over 3054619.69 frames. ], batch size: 102, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:18:18,157 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.037e+02 2.441e+02 3.104e+02 5.362e+02, threshold=4.883e+02, percent-clipped=3.0 2023-05-01 13:18:19,215 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:52,780 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222926.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:06,268 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4312, 4.5344, 4.6548, 4.4486, 4.5555, 5.0379, 4.5871, 4.2902], device='cuda:5'), covar=tensor([0.1331, 0.2071, 0.2542, 0.2028, 0.2438, 0.0951, 0.1603, 0.2399], device='cuda:5'), in_proj_covar=tensor([0.0386, 0.0561, 0.0623, 0.0463, 0.0620, 0.0647, 0.0490, 0.0622], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 13:19:24,112 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:45,658 INFO [train.py:904] (5/8) Epoch 22, batch 9800, loss[loss=0.1514, simple_loss=0.2396, pruned_loss=0.03156, over 12269.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2621, pruned_loss=0.03605, over 3066532.55 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:20:15,204 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:20:40,820 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:20:43,267 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 13:21:23,075 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:21:26,496 INFO [train.py:904] (5/8) Epoch 22, batch 9850, loss[loss=0.1616, simple_loss=0.2592, pruned_loss=0.03199, over 16274.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.263, pruned_loss=0.03584, over 3068485.97 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:21:28,330 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 13:21:37,445 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.181e+02 2.565e+02 3.075e+02 7.070e+02, threshold=5.130e+02, percent-clipped=2.0 2023-05-01 13:22:19,870 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223030.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:23:01,451 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:23:08,625 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 13:23:15,944 INFO [train.py:904] (5/8) Epoch 22, batch 9900, loss[loss=0.1616, simple_loss=0.2687, pruned_loss=0.02721, over 17148.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2629, pruned_loss=0.03566, over 3047363.28 frames. ], batch size: 48, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:23:44,414 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:25:13,519 INFO [train.py:904] (5/8) Epoch 22, batch 9950, loss[loss=0.1744, simple_loss=0.271, pruned_loss=0.03892, over 16447.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.265, pruned_loss=0.03584, over 3058764.81 frames. ], batch size: 146, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:25:27,895 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.209e+02 2.588e+02 3.009e+02 4.332e+02, threshold=5.177e+02, percent-clipped=0.0 2023-05-01 13:25:40,726 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 13:26:07,141 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:26:45,029 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9955, 2.0729, 2.1505, 3.5425, 2.0248, 2.3355, 2.1721, 2.1775], device='cuda:5'), covar=tensor([0.1350, 0.3702, 0.3230, 0.0611, 0.4298, 0.2673, 0.3946, 0.3470], device='cuda:5'), in_proj_covar=tensor([0.0391, 0.0439, 0.0362, 0.0318, 0.0430, 0.0502, 0.0410, 0.0511], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:26:59,815 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8748, 2.8806, 2.7228, 1.9833, 2.6499, 2.8433, 2.7146, 1.9400], device='cuda:5'), covar=tensor([0.0458, 0.0061, 0.0067, 0.0354, 0.0111, 0.0093, 0.0082, 0.0454], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0081, 0.0082, 0.0131, 0.0096, 0.0106, 0.0092, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 13:27:13,211 INFO [train.py:904] (5/8) Epoch 22, batch 10000, loss[loss=0.1816, simple_loss=0.2907, pruned_loss=0.03628, over 15568.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2636, pruned_loss=0.03548, over 3074520.78 frames. ], batch size: 191, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:27:32,110 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5952, 2.1051, 1.8280, 1.9095, 2.4370, 2.1368, 1.9576, 2.5031], device='cuda:5'), covar=tensor([0.0172, 0.0396, 0.0541, 0.0481, 0.0264, 0.0353, 0.0235, 0.0253], device='cuda:5'), in_proj_covar=tensor([0.0201, 0.0230, 0.0222, 0.0222, 0.0230, 0.0228, 0.0224, 0.0220], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:27:48,922 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3447, 4.3133, 4.1716, 3.4249, 4.2550, 1.7879, 4.0255, 3.8084], device='cuda:5'), covar=tensor([0.0095, 0.0086, 0.0181, 0.0264, 0.0096, 0.2784, 0.0136, 0.0297], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0151, 0.0190, 0.0167, 0.0169, 0.0203, 0.0181, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:27:51,779 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223173.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:28:04,915 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223179.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:28:54,099 INFO [train.py:904] (5/8) Epoch 22, batch 10050, loss[loss=0.1591, simple_loss=0.2594, pruned_loss=0.02942, over 16760.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2633, pruned_loss=0.03538, over 3064645.99 frames. ], batch size: 83, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:29:04,329 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.317e+02 2.078e+02 2.413e+02 2.769e+02 4.519e+02, threshold=4.827e+02, percent-clipped=0.0 2023-05-01 13:29:37,107 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8344, 4.8168, 4.5309, 4.1146, 4.6934, 1.7757, 4.4389, 4.4216], device='cuda:5'), covar=tensor([0.0105, 0.0113, 0.0223, 0.0298, 0.0128, 0.2582, 0.0157, 0.0219], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0150, 0.0190, 0.0166, 0.0169, 0.0203, 0.0180, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:29:40,804 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:42,107 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223227.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:54,690 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223234.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:30:27,337 INFO [train.py:904] (5/8) Epoch 22, batch 10100, loss[loss=0.1731, simple_loss=0.2591, pruned_loss=0.04358, over 16852.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2628, pruned_loss=0.03508, over 3072702.43 frames. ], batch size: 116, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:31:10,810 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:31:39,362 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223296.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:32:13,439 INFO [train.py:904] (5/8) Epoch 23, batch 0, loss[loss=0.212, simple_loss=0.303, pruned_loss=0.06049, over 16708.00 frames. ], tot_loss[loss=0.212, simple_loss=0.303, pruned_loss=0.06049, over 16708.00 frames. ], batch size: 62, lr: 2.97e-03, grad_scale: 8.0 2023-05-01 13:32:13,439 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 13:32:20,850 INFO [train.py:938] (5/8) Epoch 23, validation: loss=0.1454, simple_loss=0.2487, pruned_loss=0.02111, over 944034.00 frames. 2023-05-01 13:32:20,851 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 13:32:28,410 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.627e+02 3.041e+02 3.739e+02 7.225e+02, threshold=6.083e+02, percent-clipped=7.0 2023-05-01 13:32:40,827 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8457, 4.9512, 5.3225, 5.2779, 5.3702, 5.0886, 4.9156, 4.8742], device='cuda:5'), covar=tensor([0.0546, 0.0764, 0.0530, 0.0670, 0.0729, 0.0611, 0.1359, 0.0527], device='cuda:5'), in_proj_covar=tensor([0.0392, 0.0436, 0.0426, 0.0393, 0.0467, 0.0445, 0.0524, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 13:32:49,536 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:33:09,285 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:33:26,118 INFO [train.py:904] (5/8) Epoch 23, batch 50, loss[loss=0.1845, simple_loss=0.2677, pruned_loss=0.05065, over 16485.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2694, pruned_loss=0.04838, over 751744.96 frames. ], batch size: 75, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:34:32,077 INFO [train.py:904] (5/8) Epoch 23, batch 100, loss[loss=0.184, simple_loss=0.2806, pruned_loss=0.04374, over 17131.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2666, pruned_loss=0.04737, over 1319105.01 frames. ], batch size: 48, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:34:33,762 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9687, 5.4768, 5.6079, 5.2876, 5.4282, 5.9927, 5.4224, 5.1719], device='cuda:5'), covar=tensor([0.1060, 0.1874, 0.2219, 0.2085, 0.2365, 0.0898, 0.1579, 0.2433], device='cuda:5'), in_proj_covar=tensor([0.0390, 0.0567, 0.0627, 0.0468, 0.0629, 0.0653, 0.0494, 0.0627], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 13:34:42,064 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.242e+02 2.764e+02 3.330e+02 6.883e+02, threshold=5.527e+02, percent-clipped=1.0 2023-05-01 13:34:55,449 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:35:22,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9189, 4.8859, 4.6822, 4.3261, 4.7768, 2.1129, 4.5878, 4.4242], device='cuda:5'), covar=tensor([0.0099, 0.0105, 0.0228, 0.0299, 0.0111, 0.2454, 0.0142, 0.0266], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0152, 0.0192, 0.0168, 0.0170, 0.0205, 0.0182, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:35:38,699 INFO [train.py:904] (5/8) Epoch 23, batch 150, loss[loss=0.18, simple_loss=0.2678, pruned_loss=0.04611, over 15625.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2646, pruned_loss=0.04581, over 1759310.13 frames. ], batch size: 191, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:35:53,402 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1722, 3.2086, 3.4382, 2.2795, 2.9423, 2.2993, 3.6165, 3.6044], device='cuda:5'), covar=tensor([0.0241, 0.0889, 0.0657, 0.1802, 0.0843, 0.1044, 0.0480, 0.0812], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0159, 0.0164, 0.0151, 0.0143, 0.0128, 0.0140, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 13:36:13,685 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5319, 3.5041, 3.4877, 2.8401, 3.3588, 2.0999, 3.1647, 2.7823], device='cuda:5'), covar=tensor([0.0132, 0.0122, 0.0172, 0.0189, 0.0098, 0.2320, 0.0120, 0.0272], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0153, 0.0193, 0.0169, 0.0171, 0.0205, 0.0183, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:36:47,611 INFO [train.py:904] (5/8) Epoch 23, batch 200, loss[loss=0.1925, simple_loss=0.2949, pruned_loss=0.04506, over 16674.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2646, pruned_loss=0.04614, over 2099490.18 frames. ], batch size: 57, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:36:57,911 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 2.279e+02 2.532e+02 3.033e+02 5.752e+02, threshold=5.065e+02, percent-clipped=1.0 2023-05-01 13:37:21,194 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223529.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 13:37:52,719 INFO [train.py:904] (5/8) Epoch 23, batch 250, loss[loss=0.1517, simple_loss=0.231, pruned_loss=0.03626, over 16753.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2634, pruned_loss=0.04652, over 2358230.56 frames. ], batch size: 39, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:38:29,119 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:38:45,616 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 13:38:54,847 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223596.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:04,051 INFO [train.py:904] (5/8) Epoch 23, batch 300, loss[loss=0.1599, simple_loss=0.2614, pruned_loss=0.02921, over 17082.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2607, pruned_loss=0.04488, over 2579428.75 frames. ], batch size: 49, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:39:14,757 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.219e+02 2.672e+02 3.069e+02 6.824e+02, threshold=5.344e+02, percent-clipped=3.0 2023-05-01 13:39:35,540 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223625.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:35,754 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5610, 2.3281, 2.4089, 4.3593, 2.3487, 2.7124, 2.4281, 2.5558], device='cuda:5'), covar=tensor([0.1209, 0.3668, 0.3133, 0.0508, 0.4202, 0.2594, 0.3558, 0.3710], device='cuda:5'), in_proj_covar=tensor([0.0398, 0.0447, 0.0368, 0.0324, 0.0436, 0.0513, 0.0419, 0.0521], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:39:54,163 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:55,278 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:01,753 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223644.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:13,672 INFO [train.py:904] (5/8) Epoch 23, batch 350, loss[loss=0.1739, simple_loss=0.2631, pruned_loss=0.04237, over 17245.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2578, pruned_loss=0.04393, over 2744627.58 frames. ], batch size: 45, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:40:38,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1576, 4.1783, 4.4874, 4.4890, 4.5271, 4.2362, 4.2676, 4.1839], device='cuda:5'), covar=tensor([0.0394, 0.0746, 0.0432, 0.0403, 0.0473, 0.0470, 0.0764, 0.0576], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0451, 0.0439, 0.0405, 0.0482, 0.0460, 0.0540, 0.0368], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 13:40:42,277 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:03,084 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:15,684 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6783, 3.7348, 2.3376, 4.0770, 3.0045, 4.0735, 2.4182, 3.0444], device='cuda:5'), covar=tensor([0.0323, 0.0498, 0.1632, 0.0524, 0.0776, 0.0726, 0.1605, 0.0825], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0177, 0.0193, 0.0163, 0.0177, 0.0215, 0.0203, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 13:41:22,667 INFO [train.py:904] (5/8) Epoch 23, batch 400, loss[loss=0.1438, simple_loss=0.2357, pruned_loss=0.02593, over 17110.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2562, pruned_loss=0.04317, over 2875969.75 frames. ], batch size: 49, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:41:24,786 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-01 13:41:34,908 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.389e+02 2.103e+02 2.400e+02 2.968e+02 1.328e+03, threshold=4.800e+02, percent-clipped=3.0 2023-05-01 13:41:40,102 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1760, 4.8660, 4.8088, 5.2849, 5.5143, 4.9296, 5.4225, 5.4761], device='cuda:5'), covar=tensor([0.1835, 0.1403, 0.2884, 0.1180, 0.0823, 0.0990, 0.1046, 0.0999], device='cuda:5'), in_proj_covar=tensor([0.0634, 0.0779, 0.0903, 0.0789, 0.0598, 0.0625, 0.0653, 0.0754], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:41:46,592 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223720.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:42:32,064 INFO [train.py:904] (5/8) Epoch 23, batch 450, loss[loss=0.1466, simple_loss=0.2316, pruned_loss=0.03084, over 15906.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2547, pruned_loss=0.04257, over 2972636.77 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:42:52,160 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:43:30,136 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 13:43:40,978 INFO [train.py:904] (5/8) Epoch 23, batch 500, loss[loss=0.1558, simple_loss=0.2367, pruned_loss=0.03748, over 16216.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2527, pruned_loss=0.04177, over 3044611.07 frames. ], batch size: 36, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:43:50,504 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-05-01 13:43:52,984 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.054e+02 2.358e+02 2.805e+02 6.007e+02, threshold=4.715e+02, percent-clipped=4.0 2023-05-01 13:44:16,892 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223829.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:44:49,725 INFO [train.py:904] (5/8) Epoch 23, batch 550, loss[loss=0.1903, simple_loss=0.269, pruned_loss=0.05579, over 16289.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2527, pruned_loss=0.04138, over 3103811.24 frames. ], batch size: 164, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:45:13,153 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-05-01 13:45:17,136 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0406, 4.9868, 4.9090, 4.5051, 4.5879, 4.9871, 4.7810, 4.6588], device='cuda:5'), covar=tensor([0.0645, 0.1056, 0.0398, 0.0365, 0.1020, 0.0717, 0.0459, 0.0772], device='cuda:5'), in_proj_covar=tensor([0.0294, 0.0431, 0.0344, 0.0340, 0.0350, 0.0397, 0.0235, 0.0410], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:45:24,265 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223877.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:45:50,329 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0013, 2.8234, 2.7777, 5.0779, 4.0454, 4.4155, 1.8553, 3.2247], device='cuda:5'), covar=tensor([0.1392, 0.0901, 0.1269, 0.0266, 0.0225, 0.0419, 0.1731, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0174, 0.0193, 0.0188, 0.0201, 0.0215, 0.0203, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 13:45:59,108 INFO [train.py:904] (5/8) Epoch 23, batch 600, loss[loss=0.1509, simple_loss=0.234, pruned_loss=0.03385, over 16353.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2521, pruned_loss=0.04161, over 3151552.41 frames. ], batch size: 68, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:46:11,009 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.055e+02 2.460e+02 3.168e+02 5.428e+02, threshold=4.920e+02, percent-clipped=4.0 2023-05-01 13:46:20,877 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7746, 3.8949, 2.7634, 2.3508, 2.6131, 2.3515, 3.9603, 3.3755], device='cuda:5'), covar=tensor([0.2784, 0.0630, 0.1744, 0.2663, 0.2607, 0.2172, 0.0486, 0.1473], device='cuda:5'), in_proj_covar=tensor([0.0326, 0.0267, 0.0304, 0.0313, 0.0294, 0.0260, 0.0295, 0.0337], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 13:46:42,831 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:47:08,256 INFO [train.py:904] (5/8) Epoch 23, batch 650, loss[loss=0.1678, simple_loss=0.2487, pruned_loss=0.0435, over 16509.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2506, pruned_loss=0.04087, over 3188808.74 frames. ], batch size: 75, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:22,015 INFO [train.py:904] (5/8) Epoch 23, batch 700, loss[loss=0.1546, simple_loss=0.2412, pruned_loss=0.03395, over 17209.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2498, pruned_loss=0.04028, over 3215875.70 frames. ], batch size: 45, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:35,478 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.336e+02 2.741e+02 3.391e+02 8.377e+02, threshold=5.482e+02, percent-clipped=4.0 2023-05-01 13:48:38,657 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 13:48:41,612 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-01 13:49:24,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6804, 6.0811, 5.8046, 5.8110, 5.4159, 5.4753, 5.4292, 6.1613], device='cuda:5'), covar=tensor([0.1362, 0.0901, 0.1061, 0.0828, 0.0963, 0.0661, 0.1206, 0.0965], device='cuda:5'), in_proj_covar=tensor([0.0683, 0.0830, 0.0683, 0.0630, 0.0523, 0.0535, 0.0699, 0.0649], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:49:33,367 INFO [train.py:904] (5/8) Epoch 23, batch 750, loss[loss=0.1605, simple_loss=0.2535, pruned_loss=0.03379, over 17119.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2503, pruned_loss=0.04065, over 3236963.97 frames. ], batch size: 49, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:50:41,992 INFO [train.py:904] (5/8) Epoch 23, batch 800, loss[loss=0.1548, simple_loss=0.2505, pruned_loss=0.02957, over 17092.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2498, pruned_loss=0.04089, over 3247933.24 frames. ], batch size: 55, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:50:54,817 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 2.213e+02 2.619e+02 3.114e+02 4.585e+02, threshold=5.238e+02, percent-clipped=0.0 2023-05-01 13:51:51,764 INFO [train.py:904] (5/8) Epoch 23, batch 850, loss[loss=0.1626, simple_loss=0.2395, pruned_loss=0.04287, over 16741.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2493, pruned_loss=0.04048, over 3259605.26 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:52:36,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7695, 2.7490, 2.3504, 2.7407, 3.0648, 2.8329, 3.3382, 3.3062], device='cuda:5'), covar=tensor([0.0162, 0.0458, 0.0565, 0.0470, 0.0314, 0.0448, 0.0303, 0.0275], device='cuda:5'), in_proj_covar=tensor([0.0218, 0.0241, 0.0232, 0.0232, 0.0242, 0.0241, 0.0241, 0.0235], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 13:53:00,737 INFO [train.py:904] (5/8) Epoch 23, batch 900, loss[loss=0.1654, simple_loss=0.2414, pruned_loss=0.04476, over 12491.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2479, pruned_loss=0.03971, over 3268994.35 frames. ], batch size: 246, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:53:14,896 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.103e+02 2.438e+02 2.948e+02 5.610e+02, threshold=4.876e+02, percent-clipped=1.0 2023-05-01 13:53:15,338 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224212.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:53:25,977 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 13:53:46,057 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224234.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:54:10,209 INFO [train.py:904] (5/8) Epoch 23, batch 950, loss[loss=0.1672, simple_loss=0.2591, pruned_loss=0.03759, over 16611.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2483, pruned_loss=0.03973, over 3281989.89 frames. ], batch size: 62, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:54:38,589 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224273.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:54:50,501 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=224282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:55:20,408 INFO [train.py:904] (5/8) Epoch 23, batch 1000, loss[loss=0.1463, simple_loss=0.2252, pruned_loss=0.03371, over 16698.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2478, pruned_loss=0.03931, over 3296619.43 frames. ], batch size: 37, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:55:33,040 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-05-01 13:55:33,533 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 2.242e+02 2.591e+02 2.992e+02 5.709e+02, threshold=5.183e+02, percent-clipped=2.0 2023-05-01 13:56:17,328 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 13:56:31,386 INFO [train.py:904] (5/8) Epoch 23, batch 1050, loss[loss=0.1484, simple_loss=0.2479, pruned_loss=0.02447, over 17084.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2473, pruned_loss=0.03903, over 3307861.08 frames. ], batch size: 55, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:42,174 INFO [train.py:904] (5/8) Epoch 23, batch 1100, loss[loss=0.1838, simple_loss=0.2609, pruned_loss=0.05331, over 16468.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2473, pruned_loss=0.03915, over 3294851.01 frames. ], batch size: 68, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:54,072 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 2.008e+02 2.334e+02 2.683e+02 4.688e+02, threshold=4.667e+02, percent-clipped=0.0 2023-05-01 13:58:51,573 INFO [train.py:904] (5/8) Epoch 23, batch 1150, loss[loss=0.1602, simple_loss=0.2506, pruned_loss=0.0349, over 17125.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2473, pruned_loss=0.03893, over 3294859.42 frames. ], batch size: 47, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:59:32,730 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-05-01 14:00:00,971 INFO [train.py:904] (5/8) Epoch 23, batch 1200, loss[loss=0.1507, simple_loss=0.2313, pruned_loss=0.03505, over 16718.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2462, pruned_loss=0.03845, over 3304723.32 frames. ], batch size: 124, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:00:14,526 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.279e+02 2.643e+02 3.361e+02 1.197e+03, threshold=5.285e+02, percent-clipped=8.0 2023-05-01 14:00:53,754 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2013, 4.2952, 4.3566, 4.2373, 4.2695, 4.8186, 4.3748, 4.0917], device='cuda:5'), covar=tensor([0.1972, 0.2242, 0.2856, 0.2373, 0.3139, 0.1401, 0.1781, 0.2636], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0603, 0.0670, 0.0496, 0.0669, 0.0694, 0.0523, 0.0664], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 14:01:10,501 INFO [train.py:904] (5/8) Epoch 23, batch 1250, loss[loss=0.146, simple_loss=0.2341, pruned_loss=0.0289, over 17260.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2468, pruned_loss=0.03896, over 3305750.59 frames. ], batch size: 43, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:01:14,810 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6127, 4.7248, 4.8501, 4.6647, 4.7154, 5.3181, 4.8057, 4.4536], device='cuda:5'), covar=tensor([0.1554, 0.2154, 0.2713, 0.2349, 0.3094, 0.1212, 0.1877, 0.2664], device='cuda:5'), in_proj_covar=tensor([0.0412, 0.0601, 0.0667, 0.0494, 0.0666, 0.0692, 0.0521, 0.0661], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 14:01:22,918 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224562.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:01:31,868 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:01:52,155 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 14:02:07,544 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 14:02:20,416 INFO [train.py:904] (5/8) Epoch 23, batch 1300, loss[loss=0.1595, simple_loss=0.2396, pruned_loss=0.03974, over 16420.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2465, pruned_loss=0.03899, over 3296949.31 frames. ], batch size: 146, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:02:33,673 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.195e+02 2.523e+02 3.004e+02 8.579e+02, threshold=5.045e+02, percent-clipped=3.0 2023-05-01 14:02:49,065 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224623.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:02:53,168 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7390, 3.8016, 1.9319, 4.3068, 2.6909, 4.2261, 1.9777, 2.9479], device='cuda:5'), covar=tensor([0.0305, 0.0380, 0.2041, 0.0393, 0.0943, 0.0483, 0.2125, 0.0814], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0181, 0.0197, 0.0169, 0.0180, 0.0221, 0.0207, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:03:29,365 INFO [train.py:904] (5/8) Epoch 23, batch 1350, loss[loss=0.1658, simple_loss=0.2612, pruned_loss=0.03521, over 17115.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2468, pruned_loss=0.03867, over 3307509.85 frames. ], batch size: 47, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:03:49,050 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 14:04:40,084 INFO [train.py:904] (5/8) Epoch 23, batch 1400, loss[loss=0.1827, simple_loss=0.2826, pruned_loss=0.04145, over 16749.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.247, pruned_loss=0.03908, over 3298485.72 frames. ], batch size: 57, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:04:52,699 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.431e+02 2.135e+02 2.533e+02 2.912e+02 4.513e+02, threshold=5.066e+02, percent-clipped=0.0 2023-05-01 14:05:13,976 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:33,824 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:50,558 INFO [train.py:904] (5/8) Epoch 23, batch 1450, loss[loss=0.1521, simple_loss=0.2431, pruned_loss=0.03057, over 17161.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2466, pruned_loss=0.03905, over 3292147.06 frames. ], batch size: 46, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:06:01,609 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3444, 5.2828, 5.0793, 4.6003, 5.1037, 2.0212, 4.8856, 5.0313], device='cuda:5'), covar=tensor([0.0070, 0.0071, 0.0202, 0.0369, 0.0101, 0.2709, 0.0142, 0.0195], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0162, 0.0204, 0.0180, 0.0182, 0.0214, 0.0194, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:06:32,620 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4365, 5.3606, 5.2797, 4.7290, 4.8812, 5.3002, 5.2724, 4.9372], device='cuda:5'), covar=tensor([0.0622, 0.0692, 0.0627, 0.0447, 0.1331, 0.0559, 0.0309, 0.0848], device='cuda:5'), in_proj_covar=tensor([0.0303, 0.0445, 0.0354, 0.0351, 0.0359, 0.0412, 0.0240, 0.0422], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:06:39,725 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:06:39,780 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8005, 3.9225, 2.9661, 2.2960, 2.5257, 2.3375, 3.9979, 3.3590], device='cuda:5'), covar=tensor([0.2630, 0.0529, 0.1735, 0.2848, 0.2702, 0.2238, 0.0489, 0.1342], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0270, 0.0309, 0.0317, 0.0299, 0.0264, 0.0299, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 14:06:59,981 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224801.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:07:01,749 INFO [train.py:904] (5/8) Epoch 23, batch 1500, loss[loss=0.1543, simple_loss=0.2465, pruned_loss=0.03099, over 17226.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2466, pruned_loss=0.0388, over 3307131.97 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:07:03,377 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1115, 5.2024, 5.6011, 5.5784, 5.5918, 5.2520, 5.1902, 5.0201], device='cuda:5'), covar=tensor([0.0324, 0.0435, 0.0376, 0.0374, 0.0449, 0.0355, 0.0845, 0.0399], device='cuda:5'), in_proj_covar=tensor([0.0422, 0.0470, 0.0458, 0.0423, 0.0502, 0.0481, 0.0563, 0.0383], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 14:07:16,528 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.087e+02 2.500e+02 3.086e+02 8.479e+02, threshold=5.000e+02, percent-clipped=3.0 2023-05-01 14:07:44,753 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8445, 2.4800, 2.4007, 3.6471, 2.7835, 3.7369, 1.5188, 2.6823], device='cuda:5'), covar=tensor([0.1388, 0.0796, 0.1198, 0.0208, 0.0170, 0.0422, 0.1776, 0.0893], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0176, 0.0196, 0.0192, 0.0205, 0.0217, 0.0205, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:08:14,255 INFO [train.py:904] (5/8) Epoch 23, batch 1550, loss[loss=0.1975, simple_loss=0.2668, pruned_loss=0.06411, over 16405.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2487, pruned_loss=0.04047, over 3312141.19 frames. ], batch size: 146, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 14:08:34,612 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224868.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:09:23,924 INFO [train.py:904] (5/8) Epoch 23, batch 1600, loss[loss=0.1503, simple_loss=0.2472, pruned_loss=0.02671, over 17237.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2505, pruned_loss=0.04057, over 3315705.62 frames. ], batch size: 43, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:09:36,826 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.147e+02 2.697e+02 3.461e+02 9.919e+02, threshold=5.394e+02, percent-clipped=4.0 2023-05-01 14:09:41,292 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=224916.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:09:43,736 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:10:32,936 INFO [train.py:904] (5/8) Epoch 23, batch 1650, loss[loss=0.1762, simple_loss=0.2539, pruned_loss=0.04928, over 16522.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.251, pruned_loss=0.04082, over 3321230.02 frames. ], batch size: 75, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:41,681 INFO [train.py:904] (5/8) Epoch 23, batch 1700, loss[loss=0.1555, simple_loss=0.2353, pruned_loss=0.03788, over 16738.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2522, pruned_loss=0.04068, over 3322617.19 frames. ], batch size: 39, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:56,169 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 2.319e+02 2.651e+02 3.431e+02 9.066e+02, threshold=5.301e+02, percent-clipped=1.0 2023-05-01 14:12:52,530 INFO [train.py:904] (5/8) Epoch 23, batch 1750, loss[loss=0.1517, simple_loss=0.2405, pruned_loss=0.03144, over 16846.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2526, pruned_loss=0.04058, over 3325195.03 frames. ], batch size: 42, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:13:33,064 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:13:34,952 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3489, 5.2692, 5.1900, 4.6179, 4.7660, 5.2033, 5.1389, 4.7988], device='cuda:5'), covar=tensor([0.0587, 0.0554, 0.0357, 0.0396, 0.1286, 0.0539, 0.0308, 0.0881], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0448, 0.0357, 0.0354, 0.0362, 0.0414, 0.0242, 0.0426], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:13:52,632 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:14:01,708 INFO [train.py:904] (5/8) Epoch 23, batch 1800, loss[loss=0.1731, simple_loss=0.2702, pruned_loss=0.038, over 17075.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2541, pruned_loss=0.0402, over 3328792.18 frames. ], batch size: 55, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:14:02,788 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:14:15,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.184e+02 2.489e+02 3.045e+02 6.303e+02, threshold=4.978e+02, percent-clipped=2.0 2023-05-01 14:14:35,764 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6171, 4.4553, 4.7210, 4.8498, 5.0314, 4.5237, 4.9817, 5.0369], device='cuda:5'), covar=tensor([0.1951, 0.1432, 0.1721, 0.0861, 0.0635, 0.1121, 0.1150, 0.0775], device='cuda:5'), in_proj_covar=tensor([0.0674, 0.0832, 0.0964, 0.0842, 0.0636, 0.0668, 0.0692, 0.0803], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:14:57,323 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225142.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:15:12,134 INFO [train.py:904] (5/8) Epoch 23, batch 1850, loss[loss=0.1873, simple_loss=0.2666, pruned_loss=0.05404, over 16788.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2551, pruned_loss=0.04081, over 3318130.36 frames. ], batch size: 124, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:15:27,776 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:16:22,246 INFO [train.py:904] (5/8) Epoch 23, batch 1900, loss[loss=0.1748, simple_loss=0.2632, pruned_loss=0.04322, over 15564.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2549, pruned_loss=0.04037, over 3317343.31 frames. ], batch size: 191, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:16:22,764 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:16:36,682 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 2.136e+02 2.472e+02 2.864e+02 6.129e+02, threshold=4.945e+02, percent-clipped=2.0 2023-05-01 14:16:43,992 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:17:11,952 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 14:17:31,823 INFO [train.py:904] (5/8) Epoch 23, batch 1950, loss[loss=0.1528, simple_loss=0.2387, pruned_loss=0.03343, over 16827.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2546, pruned_loss=0.04003, over 3317357.17 frames. ], batch size: 42, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:17:49,204 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8510, 2.0604, 2.3186, 3.1086, 2.1568, 2.2494, 2.2704, 2.2015], device='cuda:5'), covar=tensor([0.1467, 0.3806, 0.2774, 0.0797, 0.4110, 0.2634, 0.3447, 0.3486], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0456, 0.0374, 0.0332, 0.0441, 0.0526, 0.0428, 0.0532], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:17:50,059 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225266.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:18:24,632 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9095, 4.6726, 4.9302, 5.1053, 5.3450, 4.6420, 5.3042, 5.3256], device='cuda:5'), covar=tensor([0.2178, 0.1495, 0.2074, 0.0938, 0.0608, 0.1123, 0.0608, 0.0653], device='cuda:5'), in_proj_covar=tensor([0.0677, 0.0838, 0.0971, 0.0844, 0.0639, 0.0672, 0.0694, 0.0805], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:18:37,342 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2268, 2.6823, 2.1298, 2.5306, 3.0374, 2.7939, 3.1612, 3.1412], device='cuda:5'), covar=tensor([0.0199, 0.0399, 0.0584, 0.0398, 0.0272, 0.0364, 0.0256, 0.0280], device='cuda:5'), in_proj_covar=tensor([0.0221, 0.0242, 0.0232, 0.0232, 0.0243, 0.0242, 0.0243, 0.0238], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:18:42,389 INFO [train.py:904] (5/8) Epoch 23, batch 2000, loss[loss=0.1719, simple_loss=0.2614, pruned_loss=0.04117, over 15507.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2545, pruned_loss=0.04014, over 3313363.52 frames. ], batch size: 190, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:18:56,052 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.089e+02 2.408e+02 2.851e+02 8.100e+02, threshold=4.816e+02, percent-clipped=1.0 2023-05-01 14:19:52,857 INFO [train.py:904] (5/8) Epoch 23, batch 2050, loss[loss=0.1838, simple_loss=0.2876, pruned_loss=0.04, over 17274.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2549, pruned_loss=0.04064, over 3300086.64 frames. ], batch size: 52, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:20:02,890 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 14:20:35,702 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:20:54,900 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225396.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:04,283 INFO [train.py:904] (5/8) Epoch 23, batch 2100, loss[loss=0.1579, simple_loss=0.241, pruned_loss=0.03742, over 15998.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2565, pruned_loss=0.04169, over 3284324.56 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:21:09,853 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8212, 1.8621, 2.3767, 2.6725, 2.7535, 2.8078, 1.9000, 2.9986], device='cuda:5'), covar=tensor([0.0212, 0.0579, 0.0358, 0.0327, 0.0340, 0.0307, 0.0652, 0.0181], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0195, 0.0183, 0.0188, 0.0200, 0.0158, 0.0199, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:21:18,934 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.181e+02 2.818e+02 3.430e+02 7.396e+02, threshold=5.635e+02, percent-clipped=7.0 2023-05-01 14:21:29,363 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-05-01 14:21:44,081 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:57,076 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:03,554 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225444.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:15,763 INFO [train.py:904] (5/8) Epoch 23, batch 2150, loss[loss=0.1734, simple_loss=0.2579, pruned_loss=0.04443, over 16448.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2581, pruned_loss=0.04249, over 3293414.12 frames. ], batch size: 146, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:22:25,207 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:29,440 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:19,358 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225498.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:22,122 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:26,036 INFO [train.py:904] (5/8) Epoch 23, batch 2200, loss[loss=0.1693, simple_loss=0.2487, pruned_loss=0.04492, over 16501.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2575, pruned_loss=0.04212, over 3307981.56 frames. ], batch size: 146, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:23:29,436 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225505.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:35,974 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9052, 4.9731, 4.8281, 4.4167, 4.3401, 4.9087, 4.7930, 4.4337], device='cuda:5'), covar=tensor([0.0743, 0.0678, 0.0378, 0.0466, 0.1268, 0.0580, 0.0456, 0.0928], device='cuda:5'), in_proj_covar=tensor([0.0312, 0.0456, 0.0363, 0.0362, 0.0368, 0.0422, 0.0246, 0.0434], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 14:23:40,048 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.156e+02 2.513e+02 2.761e+02 5.563e+02, threshold=5.025e+02, percent-clipped=0.0 2023-05-01 14:23:53,524 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225522.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:55,322 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:24:36,371 INFO [train.py:904] (5/8) Epoch 23, batch 2250, loss[loss=0.1492, simple_loss=0.2343, pruned_loss=0.03206, over 16818.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2575, pruned_loss=0.04214, over 3316983.89 frames. ], batch size: 102, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:24:40,128 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9589, 2.1791, 2.6168, 2.8446, 2.8322, 3.4486, 2.4258, 3.4109], device='cuda:5'), covar=tensor([0.0281, 0.0488, 0.0344, 0.0364, 0.0359, 0.0198, 0.0486, 0.0168], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0188, 0.0200, 0.0157, 0.0199, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:24:45,169 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225559.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 14:24:54,933 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:18,821 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:46,975 INFO [train.py:904] (5/8) Epoch 23, batch 2300, loss[loss=0.1698, simple_loss=0.2683, pruned_loss=0.03564, over 17045.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2568, pruned_loss=0.04179, over 3312100.13 frames. ], batch size: 53, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:25:56,973 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 14:26:01,554 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.061e+02 2.466e+02 3.026e+02 4.196e+02, threshold=4.932e+02, percent-clipped=0.0 2023-05-01 14:26:11,630 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225620.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 14:26:58,885 INFO [train.py:904] (5/8) Epoch 23, batch 2350, loss[loss=0.1413, simple_loss=0.2278, pruned_loss=0.0274, over 16279.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2567, pruned_loss=0.04176, over 3304508.21 frames. ], batch size: 36, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:27:34,811 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3544, 2.3813, 2.4857, 4.0398, 2.2907, 2.6877, 2.4461, 2.5804], device='cuda:5'), covar=tensor([0.1466, 0.3644, 0.2933, 0.0723, 0.4161, 0.2531, 0.3590, 0.3061], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0459, 0.0376, 0.0335, 0.0443, 0.0530, 0.0431, 0.0537], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:28:10,320 INFO [train.py:904] (5/8) Epoch 23, batch 2400, loss[loss=0.1632, simple_loss=0.257, pruned_loss=0.03471, over 17131.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2575, pruned_loss=0.04176, over 3306525.46 frames. ], batch size: 49, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:28:23,215 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.701e+02 2.227e+02 2.621e+02 3.138e+02 6.803e+02, threshold=5.242e+02, percent-clipped=3.0 2023-05-01 14:29:17,850 INFO [train.py:904] (5/8) Epoch 23, batch 2450, loss[loss=0.1545, simple_loss=0.2474, pruned_loss=0.03077, over 17176.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2579, pruned_loss=0.04155, over 3318211.70 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:29:26,325 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:10,489 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-05-01 14:30:17,201 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:22,204 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225798.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:29,833 INFO [train.py:904] (5/8) Epoch 23, batch 2500, loss[loss=0.1834, simple_loss=0.283, pruned_loss=0.04191, over 16635.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2581, pruned_loss=0.04164, over 3320405.94 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:30:36,126 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225807.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:43,415 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.105e+02 2.544e+02 3.075e+02 7.140e+02, threshold=5.088e+02, percent-clipped=2.0 2023-05-01 14:30:51,427 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225818.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:29,064 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225846.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:38,364 INFO [train.py:904] (5/8) Epoch 23, batch 2550, loss[loss=0.1846, simple_loss=0.2666, pruned_loss=0.05128, over 16439.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2578, pruned_loss=0.04145, over 3324597.38 frames. ], batch size: 68, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:31:39,533 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9260, 2.0513, 2.4481, 2.7677, 2.8227, 2.9407, 2.2475, 3.1228], device='cuda:5'), covar=tensor([0.0201, 0.0493, 0.0361, 0.0305, 0.0332, 0.0268, 0.0493, 0.0176], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0195, 0.0183, 0.0188, 0.0201, 0.0158, 0.0198, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:31:49,564 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:50,894 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225862.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:14,118 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225878.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:38,046 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225895.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:49,281 INFO [train.py:904] (5/8) Epoch 23, batch 2600, loss[loss=0.1782, simple_loss=0.2681, pruned_loss=0.04414, over 16615.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2582, pruned_loss=0.04157, over 3330764.27 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:32:51,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 14:33:03,048 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.265e+02 2.588e+02 2.921e+02 5.309e+02, threshold=5.176e+02, percent-clipped=1.0 2023-05-01 14:33:06,092 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225915.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:33:07,233 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9891, 5.0524, 5.4552, 5.4506, 5.4872, 5.1115, 5.0656, 4.8099], device='cuda:5'), covar=tensor([0.0361, 0.0619, 0.0394, 0.0431, 0.0528, 0.0445, 0.1024, 0.0514], device='cuda:5'), in_proj_covar=tensor([0.0427, 0.0476, 0.0462, 0.0427, 0.0509, 0.0485, 0.0568, 0.0387], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 14:33:16,614 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225923.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:33:58,803 INFO [train.py:904] (5/8) Epoch 23, batch 2650, loss[loss=0.1757, simple_loss=0.2768, pruned_loss=0.03727, over 16071.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2593, pruned_loss=0.04158, over 3328616.26 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:34:04,236 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:34:51,998 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-05-01 14:35:12,271 INFO [train.py:904] (5/8) Epoch 23, batch 2700, loss[loss=0.1675, simple_loss=0.2679, pruned_loss=0.03349, over 16574.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2593, pruned_loss=0.0408, over 3333081.83 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:35:25,024 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0928, 4.8590, 5.1270, 5.2948, 5.5093, 4.7866, 5.4971, 5.5095], device='cuda:5'), covar=tensor([0.1887, 0.1386, 0.1719, 0.0783, 0.0568, 0.1003, 0.0588, 0.0586], device='cuda:5'), in_proj_covar=tensor([0.0672, 0.0835, 0.0965, 0.0842, 0.0640, 0.0668, 0.0692, 0.0803], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:35:25,735 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.127e+02 2.428e+02 2.889e+02 7.591e+02, threshold=4.856e+02, percent-clipped=2.0 2023-05-01 14:36:06,302 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:36:13,609 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6241, 4.5711, 4.5166, 3.9440, 4.5572, 1.7353, 4.3054, 4.1707], device='cuda:5'), covar=tensor([0.0125, 0.0111, 0.0175, 0.0303, 0.0102, 0.2919, 0.0142, 0.0227], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0165, 0.0207, 0.0184, 0.0185, 0.0216, 0.0197, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:36:23,070 INFO [train.py:904] (5/8) Epoch 23, batch 2750, loss[loss=0.1684, simple_loss=0.255, pruned_loss=0.04096, over 16831.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2593, pruned_loss=0.04056, over 3333630.67 frames. ], batch size: 83, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:21,882 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226095.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:25,061 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6447, 1.6719, 1.5258, 1.3638, 1.7538, 1.5556, 1.5544, 1.9068], device='cuda:5'), covar=tensor([0.0260, 0.0438, 0.0555, 0.0542, 0.0317, 0.0398, 0.0228, 0.0319], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0243, 0.0231, 0.0233, 0.0244, 0.0242, 0.0244, 0.0239], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:37:31,958 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=226102.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:32,648 INFO [train.py:904] (5/8) Epoch 23, batch 2800, loss[loss=0.1759, simple_loss=0.2579, pruned_loss=0.04697, over 16461.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2591, pruned_loss=0.04096, over 3333005.17 frames. ], batch size: 146, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:47,351 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 2.068e+02 2.362e+02 2.950e+02 8.233e+02, threshold=4.725e+02, percent-clipped=4.0 2023-05-01 14:37:54,706 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:59,266 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1409, 2.0004, 2.7197, 3.0215, 3.0367, 3.4897, 2.0801, 3.5908], device='cuda:5'), covar=tensor([0.0247, 0.0603, 0.0328, 0.0322, 0.0294, 0.0206, 0.0721, 0.0140], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0197, 0.0184, 0.0190, 0.0202, 0.0160, 0.0201, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:38:29,330 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:38:42,390 INFO [train.py:904] (5/8) Epoch 23, batch 2850, loss[loss=0.1649, simple_loss=0.2606, pruned_loss=0.03459, over 17136.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2579, pruned_loss=0.04032, over 3336457.85 frames. ], batch size: 48, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:38:54,417 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:00,760 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:18,089 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:51,837 INFO [train.py:904] (5/8) Epoch 23, batch 2900, loss[loss=0.1571, simple_loss=0.2351, pruned_loss=0.03953, over 15986.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2573, pruned_loss=0.0413, over 3318132.52 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:39:59,595 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 14:40:00,217 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226209.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:05,818 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 2.036e+02 2.506e+02 2.935e+02 4.421e+02, threshold=5.013e+02, percent-clipped=0.0 2023-05-01 14:40:09,330 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226215.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 14:40:13,517 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:24,106 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:34,172 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226233.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:58,440 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226251.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:41:00,401 INFO [train.py:904] (5/8) Epoch 23, batch 2950, loss[loss=0.1469, simple_loss=0.237, pruned_loss=0.02835, over 16984.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.257, pruned_loss=0.04142, over 3319576.28 frames. ], batch size: 41, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:41:14,148 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226263.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:41:56,415 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=226294.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:42:08,780 INFO [train.py:904] (5/8) Epoch 23, batch 3000, loss[loss=0.2034, simple_loss=0.2914, pruned_loss=0.05776, over 16604.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2572, pruned_loss=0.04172, over 3321140.14 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:42:08,780 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 14:42:17,868 INFO [train.py:938] (5/8) Epoch 23, validation: loss=0.1344, simple_loss=0.2397, pruned_loss=0.01456, over 944034.00 frames. 2023-05-01 14:42:17,869 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 14:42:30,995 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.313e+02 2.743e+02 3.282e+02 6.136e+02, threshold=5.486e+02, percent-clipped=4.0 2023-05-01 14:42:47,626 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5257, 5.4797, 5.2452, 4.5196, 5.3489, 2.2546, 5.1105, 5.1863], device='cuda:5'), covar=tensor([0.0085, 0.0082, 0.0221, 0.0453, 0.0098, 0.2469, 0.0125, 0.0189], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0164, 0.0205, 0.0183, 0.0184, 0.0213, 0.0195, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:42:54,746 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 14:43:05,553 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 14:43:16,456 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3223, 5.2953, 4.9859, 4.4561, 5.1197, 2.0550, 4.8513, 4.9700], device='cuda:5'), covar=tensor([0.0075, 0.0081, 0.0226, 0.0458, 0.0112, 0.2747, 0.0149, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0164, 0.0206, 0.0183, 0.0184, 0.0213, 0.0196, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:43:20,839 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2164, 5.0128, 5.2696, 5.4294, 5.6334, 4.9761, 5.6055, 5.5991], device='cuda:5'), covar=tensor([0.2060, 0.1374, 0.1759, 0.0790, 0.0601, 0.0906, 0.0616, 0.0683], device='cuda:5'), in_proj_covar=tensor([0.0676, 0.0841, 0.0971, 0.0847, 0.0644, 0.0674, 0.0696, 0.0809], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:43:26,885 INFO [train.py:904] (5/8) Epoch 23, batch 3050, loss[loss=0.1769, simple_loss=0.2708, pruned_loss=0.04148, over 16664.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2578, pruned_loss=0.04223, over 3307001.82 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:43:29,240 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9698, 2.0545, 2.5864, 2.8798, 2.8071, 3.3569, 2.4113, 3.4148], device='cuda:5'), covar=tensor([0.0275, 0.0524, 0.0344, 0.0356, 0.0365, 0.0217, 0.0483, 0.0153], device='cuda:5'), in_proj_covar=tensor([0.0199, 0.0198, 0.0186, 0.0191, 0.0204, 0.0161, 0.0201, 0.0158], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:44:29,384 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:44:31,751 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 14:44:37,979 INFO [train.py:904] (5/8) Epoch 23, batch 3100, loss[loss=0.1705, simple_loss=0.244, pruned_loss=0.04854, over 16904.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2572, pruned_loss=0.04224, over 3313584.03 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:44:47,865 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 14:44:51,601 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.140e+02 2.727e+02 3.303e+02 6.006e+02, threshold=5.454e+02, percent-clipped=2.0 2023-05-01 14:45:21,021 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0879, 2.0811, 2.7375, 2.9995, 2.9279, 3.4556, 2.4584, 3.4688], device='cuda:5'), covar=tensor([0.0281, 0.0557, 0.0341, 0.0324, 0.0358, 0.0237, 0.0494, 0.0210], device='cuda:5'), in_proj_covar=tensor([0.0198, 0.0197, 0.0185, 0.0190, 0.0202, 0.0160, 0.0200, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:45:31,122 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5532, 3.2864, 3.5931, 1.9143, 3.6580, 3.6744, 3.0022, 2.8030], device='cuda:5'), covar=tensor([0.0715, 0.0257, 0.0195, 0.1120, 0.0115, 0.0222, 0.0437, 0.0445], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0140, 0.0082, 0.0128, 0.0129, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:45:47,148 INFO [train.py:904] (5/8) Epoch 23, batch 3150, loss[loss=0.1466, simple_loss=0.243, pruned_loss=0.02511, over 17094.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2557, pruned_loss=0.04143, over 3319467.31 frames. ], batch size: 47, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:46:23,807 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 14:46:54,861 INFO [train.py:904] (5/8) Epoch 23, batch 3200, loss[loss=0.1477, simple_loss=0.236, pruned_loss=0.02971, over 16807.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2546, pruned_loss=0.04101, over 3320387.53 frames. ], batch size: 42, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:47:09,850 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.052e+02 2.422e+02 2.820e+02 4.460e+02, threshold=4.844e+02, percent-clipped=0.0 2023-05-01 14:47:13,895 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-05-01 14:47:15,834 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:47:51,850 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6365, 2.3890, 2.3760, 4.5004, 2.4150, 2.7879, 2.4659, 2.6349], device='cuda:5'), covar=tensor([0.1249, 0.3770, 0.3247, 0.0483, 0.4108, 0.2817, 0.3665, 0.3690], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0457, 0.0374, 0.0335, 0.0441, 0.0527, 0.0428, 0.0535], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:47:56,286 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8276, 4.0421, 2.3574, 4.6230, 3.1574, 4.4996, 2.3067, 3.1465], device='cuda:5'), covar=tensor([0.0307, 0.0351, 0.1794, 0.0256, 0.0847, 0.0445, 0.1983, 0.0790], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0180, 0.0195, 0.0170, 0.0179, 0.0223, 0.0206, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:48:01,535 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:04,026 INFO [train.py:904] (5/8) Epoch 23, batch 3250, loss[loss=0.1646, simple_loss=0.2645, pruned_loss=0.0324, over 17254.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2544, pruned_loss=0.04057, over 3327535.00 frames. ], batch size: 52, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:48:22,283 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:54,456 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226589.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:49:08,850 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226599.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:49:14,412 INFO [train.py:904] (5/8) Epoch 23, batch 3300, loss[loss=0.1646, simple_loss=0.2556, pruned_loss=0.03678, over 16659.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2554, pruned_loss=0.04096, over 3330307.03 frames. ], batch size: 62, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:49:27,931 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.105e+02 2.619e+02 3.003e+02 4.974e+02, threshold=5.238e+02, percent-clipped=1.0 2023-05-01 14:49:35,502 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2811, 4.3274, 4.6419, 4.6137, 4.6647, 4.3671, 4.3775, 4.2702], device='cuda:5'), covar=tensor([0.0362, 0.0616, 0.0400, 0.0406, 0.0490, 0.0415, 0.0776, 0.0680], device='cuda:5'), in_proj_covar=tensor([0.0434, 0.0485, 0.0469, 0.0434, 0.0515, 0.0491, 0.0578, 0.0392], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 14:50:18,551 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4161, 2.8153, 3.0532, 2.0896, 2.7302, 2.1370, 3.1234, 3.0647], device='cuda:5'), covar=tensor([0.0259, 0.0897, 0.0620, 0.1884, 0.0852, 0.0993, 0.0555, 0.0799], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0170, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-01 14:50:23,123 INFO [train.py:904] (5/8) Epoch 23, batch 3350, loss[loss=0.1932, simple_loss=0.2829, pruned_loss=0.05176, over 15558.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2553, pruned_loss=0.04085, over 3324341.57 frames. ], batch size: 191, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:50:35,374 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7088, 3.4529, 3.8763, 1.9806, 3.8917, 3.9182, 3.1768, 2.8126], device='cuda:5'), covar=tensor([0.0718, 0.0249, 0.0173, 0.1186, 0.0106, 0.0222, 0.0388, 0.0493], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0140, 0.0082, 0.0128, 0.0129, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:50:58,942 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-05-01 14:51:25,732 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:51:34,610 INFO [train.py:904] (5/8) Epoch 23, batch 3400, loss[loss=0.1711, simple_loss=0.2496, pruned_loss=0.0463, over 16122.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2553, pruned_loss=0.04083, over 3309290.40 frames. ], batch size: 165, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:51:37,057 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6699, 2.6331, 2.2827, 2.4593, 2.9613, 2.7411, 3.2940, 3.2121], device='cuda:5'), covar=tensor([0.0158, 0.0495, 0.0589, 0.0485, 0.0324, 0.0444, 0.0254, 0.0296], device='cuda:5'), in_proj_covar=tensor([0.0226, 0.0247, 0.0234, 0.0236, 0.0246, 0.0245, 0.0248, 0.0243], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:51:47,775 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.087e+02 2.446e+02 3.167e+02 4.781e+02, threshold=4.893e+02, percent-clipped=0.0 2023-05-01 14:52:33,289 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:52:43,441 INFO [train.py:904] (5/8) Epoch 23, batch 3450, loss[loss=0.1722, simple_loss=0.2535, pruned_loss=0.04551, over 16475.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2536, pruned_loss=0.04015, over 3309843.69 frames. ], batch size: 146, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:52:46,088 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-01 14:52:53,505 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9146, 3.6817, 4.1789, 2.1204, 4.2931, 4.3152, 3.2233, 3.3648], device='cuda:5'), covar=tensor([0.0742, 0.0267, 0.0208, 0.1218, 0.0085, 0.0266, 0.0463, 0.0440], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0110, 0.0100, 0.0141, 0.0083, 0.0129, 0.0130, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:53:35,697 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 14:53:38,552 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8583, 2.7447, 2.5220, 4.0500, 3.2762, 4.0425, 1.5664, 2.9006], device='cuda:5'), covar=tensor([0.1332, 0.0660, 0.1151, 0.0204, 0.0147, 0.0367, 0.1587, 0.0829], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0176, 0.0197, 0.0196, 0.0207, 0.0219, 0.0205, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:53:50,553 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-01 14:53:52,947 INFO [train.py:904] (5/8) Epoch 23, batch 3500, loss[loss=0.1896, simple_loss=0.2681, pruned_loss=0.05553, over 16675.00 frames. ], tot_loss[loss=0.167, simple_loss=0.253, pruned_loss=0.04047, over 3310571.36 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:54:04,290 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8084, 2.8023, 2.5904, 4.2548, 3.4828, 4.1411, 1.6251, 3.0292], device='cuda:5'), covar=tensor([0.1366, 0.0687, 0.1147, 0.0184, 0.0181, 0.0390, 0.1590, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0176, 0.0196, 0.0195, 0.0207, 0.0219, 0.0205, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:54:07,206 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.312e+02 2.101e+02 2.445e+02 3.055e+02 4.723e+02, threshold=4.890e+02, percent-clipped=0.0 2023-05-01 14:54:24,426 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3959, 2.8818, 3.0864, 2.0337, 2.7061, 2.1040, 3.1280, 3.1111], device='cuda:5'), covar=tensor([0.0275, 0.0893, 0.0542, 0.2052, 0.0899, 0.0997, 0.0597, 0.0925], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0170, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-01 14:54:44,807 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 14:55:03,890 INFO [train.py:904] (5/8) Epoch 23, batch 3550, loss[loss=0.1845, simple_loss=0.2568, pruned_loss=0.05616, over 16919.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2527, pruned_loss=0.04033, over 3319901.69 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 16.0 2023-05-01 14:55:21,797 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5790, 2.3607, 2.3984, 4.4712, 2.3637, 2.7597, 2.3988, 2.5656], device='cuda:5'), covar=tensor([0.1318, 0.3732, 0.3156, 0.0456, 0.4104, 0.2656, 0.3804, 0.3566], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0456, 0.0374, 0.0335, 0.0441, 0.0526, 0.0428, 0.0534], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:55:32,245 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9721, 2.8501, 2.9504, 5.1234, 4.1240, 4.5168, 1.6044, 3.2194], device='cuda:5'), covar=tensor([0.1306, 0.0790, 0.1082, 0.0221, 0.0240, 0.0341, 0.1663, 0.0772], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0175, 0.0196, 0.0194, 0.0206, 0.0218, 0.0204, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:55:52,905 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:55:54,797 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2209, 2.2399, 2.4254, 4.0099, 2.2024, 2.5626, 2.2983, 2.3981], device='cuda:5'), covar=tensor([0.1557, 0.3845, 0.2903, 0.0602, 0.4036, 0.2800, 0.3958, 0.3263], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0457, 0.0374, 0.0335, 0.0441, 0.0527, 0.0428, 0.0534], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:56:09,652 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 14:56:12,767 INFO [train.py:904] (5/8) Epoch 23, batch 3600, loss[loss=0.1803, simple_loss=0.2548, pruned_loss=0.05296, over 16905.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.252, pruned_loss=0.04048, over 3310045.96 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:56:28,216 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.070e+02 2.545e+02 2.955e+02 4.911e+02, threshold=5.089e+02, percent-clipped=1.0 2023-05-01 14:57:03,053 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226937.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:57:12,416 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0885, 4.4389, 4.6340, 4.6214, 4.6556, 4.3800, 4.0792, 4.2804], device='cuda:5'), covar=tensor([0.0749, 0.1242, 0.0714, 0.0858, 0.0892, 0.0729, 0.1609, 0.0798], device='cuda:5'), in_proj_covar=tensor([0.0432, 0.0483, 0.0468, 0.0433, 0.0514, 0.0489, 0.0577, 0.0392], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 14:57:24,673 INFO [train.py:904] (5/8) Epoch 23, batch 3650, loss[loss=0.1392, simple_loss=0.2189, pruned_loss=0.02977, over 16893.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2511, pruned_loss=0.04101, over 3301352.55 frames. ], batch size: 96, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:19,630 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:58:32,263 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-01 14:58:40,561 INFO [train.py:904] (5/8) Epoch 23, batch 3700, loss[loss=0.1672, simple_loss=0.2405, pruned_loss=0.04698, over 16835.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2502, pruned_loss=0.04232, over 3282315.63 frames. ], batch size: 90, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:56,579 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.184e+02 2.522e+02 2.938e+02 6.329e+02, threshold=5.043e+02, percent-clipped=1.0 2023-05-01 14:59:19,917 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9595, 2.5982, 2.0476, 2.3420, 3.0023, 2.7525, 2.9792, 3.0463], device='cuda:5'), covar=tensor([0.0186, 0.0426, 0.0596, 0.0463, 0.0219, 0.0321, 0.0203, 0.0287], device='cuda:5'), in_proj_covar=tensor([0.0224, 0.0244, 0.0231, 0.0233, 0.0243, 0.0242, 0.0245, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 14:59:37,411 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8732, 2.7974, 2.6503, 4.3370, 3.5245, 4.1875, 1.7774, 3.0028], device='cuda:5'), covar=tensor([0.1295, 0.0679, 0.1091, 0.0158, 0.0178, 0.0355, 0.1444, 0.0846], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0175, 0.0195, 0.0194, 0.0206, 0.0217, 0.0203, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 14:59:50,706 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227050.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:59:53,692 INFO [train.py:904] (5/8) Epoch 23, batch 3750, loss[loss=0.1818, simple_loss=0.256, pruned_loss=0.05387, over 16922.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2503, pruned_loss=0.04319, over 3276198.70 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:00:17,077 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0322, 3.0595, 2.6975, 4.5580, 3.7230, 4.2325, 1.7691, 3.1445], device='cuda:5'), covar=tensor([0.1320, 0.0645, 0.1128, 0.0179, 0.0223, 0.0377, 0.1575, 0.0823], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0194, 0.0206, 0.0217, 0.0203, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:00:21,006 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 15:01:04,860 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5432, 2.7386, 2.1684, 2.4694, 3.0839, 2.7510, 3.1539, 3.2603], device='cuda:5'), covar=tensor([0.0104, 0.0356, 0.0546, 0.0464, 0.0237, 0.0348, 0.0202, 0.0244], device='cuda:5'), in_proj_covar=tensor([0.0224, 0.0244, 0.0232, 0.0234, 0.0243, 0.0243, 0.0246, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:01:07,863 INFO [train.py:904] (5/8) Epoch 23, batch 3800, loss[loss=0.1677, simple_loss=0.2441, pruned_loss=0.04569, over 16450.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2518, pruned_loss=0.04428, over 3270134.51 frames. ], batch size: 146, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:01:25,260 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.146e+02 2.630e+02 3.388e+02 6.503e+02, threshold=5.260e+02, percent-clipped=3.0 2023-05-01 15:02:21,631 INFO [train.py:904] (5/8) Epoch 23, batch 3850, loss[loss=0.1637, simple_loss=0.2437, pruned_loss=0.04191, over 16793.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.252, pruned_loss=0.04468, over 3268659.19 frames. ], batch size: 83, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:34,976 INFO [train.py:904] (5/8) Epoch 23, batch 3900, loss[loss=0.1755, simple_loss=0.2478, pruned_loss=0.0516, over 16885.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2513, pruned_loss=0.04493, over 3278286.59 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:51,351 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 15:03:51,652 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.230e+02 2.530e+02 3.049e+02 6.179e+02, threshold=5.060e+02, percent-clipped=1.0 2023-05-01 15:04:47,722 INFO [train.py:904] (5/8) Epoch 23, batch 3950, loss[loss=0.1666, simple_loss=0.2425, pruned_loss=0.04539, over 16813.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2515, pruned_loss=0.04565, over 3271263.28 frames. ], batch size: 83, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:00,636 INFO [train.py:904] (5/8) Epoch 23, batch 4000, loss[loss=0.1553, simple_loss=0.2371, pruned_loss=0.03678, over 16899.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2513, pruned_loss=0.04621, over 3276259.58 frames. ], batch size: 90, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:07,468 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6182, 4.6596, 4.8063, 4.6188, 4.6740, 5.2160, 4.7422, 4.4600], device='cuda:5'), covar=tensor([0.1450, 0.1952, 0.2278, 0.2165, 0.2831, 0.1154, 0.1602, 0.2499], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0614, 0.0678, 0.0508, 0.0676, 0.0701, 0.0531, 0.0678], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 15:06:17,148 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.172e+02 2.478e+02 2.977e+02 5.073e+02, threshold=4.957e+02, percent-clipped=1.0 2023-05-01 15:07:01,810 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:07:13,385 INFO [train.py:904] (5/8) Epoch 23, batch 4050, loss[loss=0.1633, simple_loss=0.256, pruned_loss=0.03526, over 16691.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.252, pruned_loss=0.04511, over 3285707.79 frames. ], batch size: 89, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:07:48,463 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7435, 4.7263, 4.5089, 3.6131, 4.6584, 1.5883, 4.4168, 4.0101], device='cuda:5'), covar=tensor([0.0092, 0.0073, 0.0213, 0.0426, 0.0096, 0.3425, 0.0144, 0.0376], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0166, 0.0208, 0.0184, 0.0185, 0.0214, 0.0197, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:08:27,056 INFO [train.py:904] (5/8) Epoch 23, batch 4100, loss[loss=0.1713, simple_loss=0.262, pruned_loss=0.04032, over 17123.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2539, pruned_loss=0.04481, over 3262506.13 frames. ], batch size: 48, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:39,149 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227411.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:08:42,773 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.752e+02 1.921e+02 2.258e+02 5.560e+02, threshold=3.841e+02, percent-clipped=1.0 2023-05-01 15:09:45,384 INFO [train.py:904] (5/8) Epoch 23, batch 4150, loss[loss=0.1998, simple_loss=0.2974, pruned_loss=0.05108, over 15312.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2611, pruned_loss=0.04689, over 3230415.76 frames. ], batch size: 191, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:10:16,193 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227472.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:10:22,910 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1410, 2.2317, 2.3142, 3.7952, 2.1997, 2.5883, 2.3324, 2.3844], device='cuda:5'), covar=tensor([0.1393, 0.3457, 0.2813, 0.0561, 0.4025, 0.2472, 0.3668, 0.3231], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0458, 0.0374, 0.0334, 0.0440, 0.0529, 0.0428, 0.0535], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:11:03,947 INFO [train.py:904] (5/8) Epoch 23, batch 4200, loss[loss=0.204, simple_loss=0.3039, pruned_loss=0.05203, over 16646.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2676, pruned_loss=0.04882, over 3174146.64 frames. ], batch size: 76, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:11:20,463 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.273e+02 2.688e+02 3.076e+02 6.689e+02, threshold=5.376e+02, percent-clipped=9.0 2023-05-01 15:12:16,358 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-05-01 15:12:19,945 INFO [train.py:904] (5/8) Epoch 23, batch 4250, loss[loss=0.1694, simple_loss=0.2677, pruned_loss=0.03558, over 16186.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2708, pruned_loss=0.04839, over 3170446.62 frames. ], batch size: 165, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:13:36,413 INFO [train.py:904] (5/8) Epoch 23, batch 4300, loss[loss=0.1861, simple_loss=0.2796, pruned_loss=0.04636, over 17009.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2712, pruned_loss=0.04741, over 3176902.24 frames. ], batch size: 50, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:13:55,090 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.112e+02 2.458e+02 3.022e+02 6.068e+02, threshold=4.917e+02, percent-clipped=1.0 2023-05-01 15:13:55,788 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6690, 2.8489, 2.6509, 5.0032, 3.9072, 4.1488, 1.7679, 2.9377], device='cuda:5'), covar=tensor([0.1445, 0.0861, 0.1404, 0.0240, 0.0422, 0.0496, 0.1638, 0.1018], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0175, 0.0195, 0.0193, 0.0206, 0.0215, 0.0202, 0.0193], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:14:03,430 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7627, 3.7558, 3.9038, 3.6150, 3.8704, 4.2437, 3.8983, 3.5149], device='cuda:5'), covar=tensor([0.2595, 0.2097, 0.2214, 0.2465, 0.2579, 0.1873, 0.1519, 0.2549], device='cuda:5'), in_proj_covar=tensor([0.0420, 0.0609, 0.0671, 0.0502, 0.0669, 0.0698, 0.0526, 0.0674], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 15:14:42,093 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=227645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:14:53,547 INFO [train.py:904] (5/8) Epoch 23, batch 4350, loss[loss=0.1946, simple_loss=0.2872, pruned_loss=0.05098, over 16719.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2744, pruned_loss=0.04841, over 3186848.17 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:15:55,219 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:15:55,362 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:16:09,826 INFO [train.py:904] (5/8) Epoch 23, batch 4400, loss[loss=0.1916, simple_loss=0.2753, pruned_loss=0.05395, over 16736.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2767, pruned_loss=0.04937, over 3189427.50 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:16:10,466 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6849, 4.0249, 3.1166, 2.4635, 2.8050, 2.7028, 4.4064, 3.6296], device='cuda:5'), covar=tensor([0.2915, 0.0592, 0.1583, 0.2380, 0.2455, 0.1782, 0.0379, 0.1110], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0271, 0.0308, 0.0317, 0.0302, 0.0264, 0.0300, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 15:16:27,162 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.099e+02 2.601e+02 2.882e+02 6.298e+02, threshold=5.202e+02, percent-clipped=2.0 2023-05-01 15:16:36,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2152, 5.4923, 5.2664, 5.3210, 5.0570, 4.8343, 4.9286, 5.6052], device='cuda:5'), covar=tensor([0.1081, 0.0771, 0.0934, 0.0775, 0.0706, 0.0866, 0.1125, 0.0750], device='cuda:5'), in_proj_covar=tensor([0.0688, 0.0833, 0.0689, 0.0636, 0.0527, 0.0534, 0.0700, 0.0654], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:17:22,035 INFO [train.py:904] (5/8) Epoch 23, batch 4450, loss[loss=0.2083, simple_loss=0.2992, pruned_loss=0.05876, over 16555.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2806, pruned_loss=0.05071, over 3203753.05 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:17:23,577 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227754.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:17:26,857 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 15:17:38,075 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3561, 3.5987, 3.7489, 2.1221, 3.0693, 2.3696, 3.5639, 3.8095], device='cuda:5'), covar=tensor([0.0197, 0.0741, 0.0501, 0.2116, 0.0812, 0.1019, 0.0569, 0.0868], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0167, 0.0168, 0.0153, 0.0145, 0.0129, 0.0143, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:17:41,777 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:18:35,088 INFO [train.py:904] (5/8) Epoch 23, batch 4500, loss[loss=0.1918, simple_loss=0.2674, pruned_loss=0.05815, over 11218.00 frames. ], tot_loss[loss=0.192, simple_loss=0.281, pruned_loss=0.05153, over 3204997.06 frames. ], batch size: 246, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:18:52,361 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 1.777e+02 2.085e+02 2.523e+02 4.139e+02, threshold=4.170e+02, percent-clipped=0.0 2023-05-01 15:18:52,845 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9878, 5.0301, 4.8471, 4.5182, 4.5821, 4.9494, 4.6632, 4.6452], device='cuda:5'), covar=tensor([0.0455, 0.0242, 0.0229, 0.0234, 0.0721, 0.0257, 0.0405, 0.0490], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0443, 0.0353, 0.0352, 0.0358, 0.0409, 0.0241, 0.0421], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:19:10,561 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4480, 2.2651, 2.9637, 3.4295, 3.2354, 3.8016, 2.3587, 3.8389], device='cuda:5'), covar=tensor([0.0182, 0.0513, 0.0283, 0.0239, 0.0264, 0.0129, 0.0571, 0.0100], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0194, 0.0181, 0.0187, 0.0200, 0.0157, 0.0198, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:19:11,805 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:19:14,690 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3592, 3.0741, 3.4712, 1.7784, 3.6230, 3.6606, 2.8094, 2.7604], device='cuda:5'), covar=tensor([0.0812, 0.0318, 0.0220, 0.1208, 0.0090, 0.0165, 0.0516, 0.0474], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0139, 0.0082, 0.0127, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:19:48,160 INFO [train.py:904] (5/8) Epoch 23, batch 4550, loss[loss=0.2125, simple_loss=0.3014, pruned_loss=0.06176, over 16753.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2821, pruned_loss=0.05246, over 3218638.43 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:20:39,340 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227888.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:21:00,811 INFO [train.py:904] (5/8) Epoch 23, batch 4600, loss[loss=0.1743, simple_loss=0.266, pruned_loss=0.04134, over 17132.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2829, pruned_loss=0.05275, over 3224909.33 frames. ], batch size: 47, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:21:18,784 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.806e+02 2.082e+02 2.426e+02 3.731e+02, threshold=4.163e+02, percent-clipped=0.0 2023-05-01 15:21:36,793 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1898, 5.2474, 5.5232, 5.5230, 5.6039, 5.2301, 5.2050, 4.8145], device='cuda:5'), covar=tensor([0.0249, 0.0393, 0.0282, 0.0315, 0.0383, 0.0288, 0.0795, 0.0432], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0455, 0.0440, 0.0409, 0.0489, 0.0462, 0.0545, 0.0370], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 15:22:14,206 INFO [train.py:904] (5/8) Epoch 23, batch 4650, loss[loss=0.175, simple_loss=0.2699, pruned_loss=0.04002, over 16918.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2822, pruned_loss=0.05283, over 3230637.13 frames. ], batch size: 96, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:13,233 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8455, 2.6178, 2.8167, 2.0847, 2.6282, 2.0871, 2.7091, 2.7914], device='cuda:5'), covar=tensor([0.0282, 0.0899, 0.0551, 0.1982, 0.0872, 0.0964, 0.0650, 0.0823], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0153, 0.0145, 0.0129, 0.0143, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:23:28,877 INFO [train.py:904] (5/8) Epoch 23, batch 4700, loss[loss=0.2062, simple_loss=0.2798, pruned_loss=0.06629, over 11088.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.279, pruned_loss=0.05164, over 3228744.22 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:34,974 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:23:45,413 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.844e+02 2.189e+02 2.555e+02 4.222e+02, threshold=4.378e+02, percent-clipped=1.0 2023-05-01 15:24:18,867 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0025, 4.0790, 4.2972, 4.2663, 4.2950, 4.0810, 4.0662, 4.0207], device='cuda:5'), covar=tensor([0.0321, 0.0546, 0.0400, 0.0458, 0.0435, 0.0369, 0.0769, 0.0463], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0454, 0.0440, 0.0410, 0.0488, 0.0461, 0.0547, 0.0370], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 15:24:34,876 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228049.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:24:41,825 INFO [train.py:904] (5/8) Epoch 23, batch 4750, loss[loss=0.1606, simple_loss=0.2485, pruned_loss=0.03634, over 16930.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2754, pruned_loss=0.0499, over 3226987.02 frames. ], batch size: 116, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:25:02,236 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:03,382 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:19,551 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228080.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:54,069 INFO [train.py:904] (5/8) Epoch 23, batch 4800, loss[loss=0.1809, simple_loss=0.2813, pruned_loss=0.04024, over 15464.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.272, pruned_loss=0.04815, over 3213860.22 frames. ], batch size: 190, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:26:10,775 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.116e+02 1.852e+02 2.092e+02 2.435e+02 6.400e+02, threshold=4.184e+02, percent-clipped=1.0 2023-05-01 15:26:11,947 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228115.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:26:50,587 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228141.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:27:07,821 INFO [train.py:904] (5/8) Epoch 23, batch 4850, loss[loss=0.219, simple_loss=0.3195, pruned_loss=0.05924, over 15420.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2735, pruned_loss=0.04756, over 3202551.32 frames. ], batch size: 190, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:27:42,709 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0603, 2.1130, 1.8092, 1.7601, 2.3197, 1.9818, 1.8647, 2.3925], device='cuda:5'), covar=tensor([0.0215, 0.0430, 0.0553, 0.0518, 0.0266, 0.0392, 0.0186, 0.0274], device='cuda:5'), in_proj_covar=tensor([0.0217, 0.0237, 0.0227, 0.0229, 0.0237, 0.0237, 0.0237, 0.0235], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:27:54,530 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228183.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:28:23,612 INFO [train.py:904] (5/8) Epoch 23, batch 4900, loss[loss=0.1711, simple_loss=0.2629, pruned_loss=0.03961, over 16786.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2727, pruned_loss=0.04639, over 3202117.79 frames. ], batch size: 124, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:28:42,124 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.282e+02 1.946e+02 2.242e+02 2.632e+02 4.949e+02, threshold=4.484e+02, percent-clipped=2.0 2023-05-01 15:29:36,726 INFO [train.py:904] (5/8) Epoch 23, batch 4950, loss[loss=0.1751, simple_loss=0.2661, pruned_loss=0.04203, over 16547.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2723, pruned_loss=0.04589, over 3196791.84 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:30:51,758 INFO [train.py:904] (5/8) Epoch 23, batch 5000, loss[loss=0.1767, simple_loss=0.2779, pruned_loss=0.03778, over 16664.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2732, pruned_loss=0.04557, over 3196279.54 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:31:10,003 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 1.949e+02 2.305e+02 2.743e+02 3.734e+02, threshold=4.611e+02, percent-clipped=0.0 2023-05-01 15:31:18,721 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5635, 3.5864, 2.2274, 4.2412, 2.6898, 4.1476, 2.3426, 2.9272], device='cuda:5'), covar=tensor([0.0308, 0.0371, 0.1609, 0.0121, 0.0880, 0.0419, 0.1497, 0.0860], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0175, 0.0192, 0.0164, 0.0175, 0.0216, 0.0200, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:31:30,426 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:36,837 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:00,045 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:04,604 INFO [train.py:904] (5/8) Epoch 23, batch 5050, loss[loss=0.1876, simple_loss=0.2854, pruned_loss=0.04485, over 16654.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2733, pruned_loss=0.04521, over 3214333.04 frames. ], batch size: 76, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:32:19,783 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228363.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:39,633 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7421, 3.7632, 2.3132, 4.5210, 2.9264, 4.3562, 2.5313, 3.0688], device='cuda:5'), covar=tensor([0.0255, 0.0367, 0.1623, 0.0118, 0.0840, 0.0400, 0.1383, 0.0773], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0175, 0.0192, 0.0164, 0.0175, 0.0216, 0.0200, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:32:58,953 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228390.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:00,921 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9331, 2.1374, 2.2298, 3.4199, 2.0842, 2.3893, 2.2393, 2.2934], device='cuda:5'), covar=tensor([0.1498, 0.3517, 0.2944, 0.0646, 0.4155, 0.2595, 0.3575, 0.3245], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0455, 0.0371, 0.0330, 0.0436, 0.0523, 0.0424, 0.0530], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:33:04,537 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228394.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:09,566 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:17,240 INFO [train.py:904] (5/8) Epoch 23, batch 5100, loss[loss=0.1669, simple_loss=0.2655, pruned_loss=0.0341, over 16834.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2716, pruned_loss=0.04474, over 3200350.73 frames. ], batch size: 96, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:33:22,462 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4170, 3.3635, 3.4630, 3.5330, 3.5944, 3.3187, 3.5718, 3.6422], device='cuda:5'), covar=tensor([0.1226, 0.0951, 0.1027, 0.0593, 0.0578, 0.2293, 0.0927, 0.0730], device='cuda:5'), in_proj_covar=tensor([0.0642, 0.0796, 0.0918, 0.0802, 0.0611, 0.0637, 0.0658, 0.0766], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:33:34,757 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 1.954e+02 2.359e+02 2.853e+02 6.747e+02, threshold=4.718e+02, percent-clipped=3.0 2023-05-01 15:33:55,469 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228429.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:34:06,446 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228436.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:34:30,927 INFO [train.py:904] (5/8) Epoch 23, batch 5150, loss[loss=0.1889, simple_loss=0.2828, pruned_loss=0.04747, over 16735.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.272, pruned_loss=0.04374, over 3214760.07 frames. ], batch size: 124, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:35:13,106 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7583, 4.7464, 4.5790, 3.0717, 3.9528, 4.6269, 3.9291, 2.6198], device='cuda:5'), covar=tensor([0.0552, 0.0030, 0.0035, 0.0394, 0.0093, 0.0092, 0.0103, 0.0452], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0085, 0.0087, 0.0135, 0.0100, 0.0112, 0.0096, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 15:35:15,346 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:35:25,964 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228490.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 15:35:44,374 INFO [train.py:904] (5/8) Epoch 23, batch 5200, loss[loss=0.1756, simple_loss=0.2646, pruned_loss=0.04331, over 16850.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2704, pruned_loss=0.04326, over 3213360.05 frames. ], batch size: 116, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:35:49,240 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8771, 2.9426, 2.8387, 5.2162, 4.0488, 4.3484, 1.9318, 3.2337], device='cuda:5'), covar=tensor([0.1262, 0.0757, 0.1143, 0.0094, 0.0353, 0.0389, 0.1467, 0.0784], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0176, 0.0196, 0.0192, 0.0205, 0.0215, 0.0204, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:36:01,315 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 2.022e+02 2.336e+02 2.717e+02 6.475e+02, threshold=4.673e+02, percent-clipped=1.0 2023-05-01 15:36:25,657 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228531.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:36:47,517 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8273, 1.4063, 1.6552, 1.7817, 1.8898, 1.9691, 1.6471, 1.8362], device='cuda:5'), covar=tensor([0.0276, 0.0458, 0.0253, 0.0369, 0.0351, 0.0272, 0.0471, 0.0181], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0189, 0.0202, 0.0158, 0.0199, 0.0156], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:36:57,635 INFO [train.py:904] (5/8) Epoch 23, batch 5250, loss[loss=0.1694, simple_loss=0.2604, pruned_loss=0.03923, over 16486.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2681, pruned_loss=0.04288, over 3201911.02 frames. ], batch size: 75, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:37:49,431 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 15:38:10,576 INFO [train.py:904] (5/8) Epoch 23, batch 5300, loss[loss=0.1466, simple_loss=0.2376, pruned_loss=0.02775, over 16928.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2651, pruned_loss=0.04203, over 3203410.82 frames. ], batch size: 90, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:38:23,358 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3707, 1.6996, 2.0682, 2.3831, 2.4852, 2.7140, 1.8503, 2.5834], device='cuda:5'), covar=tensor([0.0252, 0.0596, 0.0355, 0.0432, 0.0357, 0.0257, 0.0562, 0.0174], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0195, 0.0182, 0.0188, 0.0201, 0.0158, 0.0199, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:38:28,440 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 1.965e+02 2.226e+02 2.604e+02 5.338e+02, threshold=4.452e+02, percent-clipped=3.0 2023-05-01 15:39:23,402 INFO [train.py:904] (5/8) Epoch 23, batch 5350, loss[loss=0.1707, simple_loss=0.2735, pruned_loss=0.03389, over 16773.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2629, pruned_loss=0.04127, over 3203988.23 frames. ], batch size: 83, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:39:38,515 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:39:49,004 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 15:40:10,713 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228685.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:16,861 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228689.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:36,802 INFO [train.py:904] (5/8) Epoch 23, batch 5400, loss[loss=0.1952, simple_loss=0.295, pruned_loss=0.04774, over 16744.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.266, pruned_loss=0.04194, over 3196763.04 frames. ], batch size: 124, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:40:48,792 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:54,349 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 1.938e+02 2.190e+02 2.538e+02 4.586e+02, threshold=4.379e+02, percent-clipped=1.0 2023-05-01 15:41:27,339 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:41:54,072 INFO [train.py:904] (5/8) Epoch 23, batch 5450, loss[loss=0.2149, simple_loss=0.3003, pruned_loss=0.06476, over 16894.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2693, pruned_loss=0.04354, over 3193963.76 frames. ], batch size: 116, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:42:41,724 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 15:42:42,359 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228784.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:42:44,380 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228785.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:43:12,596 INFO [train.py:904] (5/8) Epoch 23, batch 5500, loss[loss=0.2719, simple_loss=0.3368, pruned_loss=0.1035, over 11937.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2764, pruned_loss=0.0478, over 3162822.13 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:43:32,428 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.602e+02 3.171e+02 3.709e+02 7.104e+02, threshold=6.343e+02, percent-clipped=12.0 2023-05-01 15:44:31,570 INFO [train.py:904] (5/8) Epoch 23, batch 5550, loss[loss=0.2324, simple_loss=0.3171, pruned_loss=0.07388, over 16288.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2832, pruned_loss=0.05268, over 3144117.77 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:45:26,474 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1024, 5.1092, 4.9780, 4.5918, 4.6119, 4.9883, 4.8847, 4.7630], device='cuda:5'), covar=tensor([0.0675, 0.0669, 0.0286, 0.0351, 0.0998, 0.0525, 0.0434, 0.0650], device='cuda:5'), in_proj_covar=tensor([0.0297, 0.0439, 0.0347, 0.0345, 0.0353, 0.0404, 0.0236, 0.0413], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:45:49,306 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 15:45:52,977 INFO [train.py:904] (5/8) Epoch 23, batch 5600, loss[loss=0.2189, simple_loss=0.2999, pruned_loss=0.06897, over 16198.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2878, pruned_loss=0.05687, over 3099274.45 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:45:55,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9169, 4.1869, 4.0118, 4.0541, 3.7439, 3.7939, 3.9002, 4.1735], device='cuda:5'), covar=tensor([0.1118, 0.0872, 0.1075, 0.0810, 0.0784, 0.1697, 0.0849, 0.0964], device='cuda:5'), in_proj_covar=tensor([0.0672, 0.0816, 0.0674, 0.0621, 0.0517, 0.0525, 0.0688, 0.0642], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:46:10,854 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.072e+02 3.210e+02 3.734e+02 4.915e+02 1.270e+03, threshold=7.469e+02, percent-clipped=7.0 2023-05-01 15:46:20,939 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-05-01 15:46:45,688 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:47:13,523 INFO [train.py:904] (5/8) Epoch 23, batch 5650, loss[loss=0.1994, simple_loss=0.2879, pruned_loss=0.05545, over 16596.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2927, pruned_loss=0.06069, over 3077368.33 frames. ], batch size: 62, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:47:19,121 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 15:48:04,607 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228985.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:11,618 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:22,394 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228996.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:32,730 INFO [train.py:904] (5/8) Epoch 23, batch 5700, loss[loss=0.2207, simple_loss=0.311, pruned_loss=0.06523, over 16635.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2941, pruned_loss=0.06194, over 3066816.65 frames. ], batch size: 57, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:48:50,701 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.137e+02 3.035e+02 3.772e+02 4.443e+02 6.175e+02, threshold=7.544e+02, percent-clipped=0.0 2023-05-01 15:49:18,827 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229033.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:49:25,417 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:49:28,923 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1477, 2.9298, 3.2331, 1.6771, 3.4084, 3.4366, 2.6936, 2.5987], device='cuda:5'), covar=tensor([0.0929, 0.0342, 0.0250, 0.1313, 0.0107, 0.0227, 0.0504, 0.0528], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0139, 0.0082, 0.0127, 0.0128, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 15:49:50,907 INFO [train.py:904] (5/8) Epoch 23, batch 5750, loss[loss=0.2285, simple_loss=0.3016, pruned_loss=0.07764, over 11522.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2963, pruned_loss=0.06333, over 3031903.07 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:50:22,125 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9074, 4.1660, 3.9906, 4.0466, 3.7136, 3.8016, 3.8558, 4.1510], device='cuda:5'), covar=tensor([0.1090, 0.0860, 0.1027, 0.0825, 0.0817, 0.1591, 0.0903, 0.0973], device='cuda:5'), in_proj_covar=tensor([0.0670, 0.0812, 0.0673, 0.0619, 0.0516, 0.0523, 0.0685, 0.0639], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:50:44,869 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229085.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:51:13,993 INFO [train.py:904] (5/8) Epoch 23, batch 5800, loss[loss=0.1828, simple_loss=0.2708, pruned_loss=0.04741, over 16542.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2959, pruned_loss=0.06226, over 3038494.34 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:51:32,215 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.918e+02 3.323e+02 4.052e+02 8.500e+02, threshold=6.645e+02, percent-clipped=1.0 2023-05-01 15:51:43,956 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3913, 2.4756, 2.4785, 4.1819, 2.3665, 2.8702, 2.5308, 2.6582], device='cuda:5'), covar=tensor([0.1269, 0.3328, 0.2726, 0.0485, 0.3760, 0.2261, 0.3192, 0.3100], device='cuda:5'), in_proj_covar=tensor([0.0400, 0.0449, 0.0367, 0.0325, 0.0432, 0.0516, 0.0419, 0.0524], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 15:52:01,489 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229133.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:52:31,397 INFO [train.py:904] (5/8) Epoch 23, batch 5850, loss[loss=0.1975, simple_loss=0.2885, pruned_loss=0.05329, over 16460.00 frames. ], tot_loss[loss=0.208, simple_loss=0.294, pruned_loss=0.06104, over 3037467.36 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:53:53,142 INFO [train.py:904] (5/8) Epoch 23, batch 5900, loss[loss=0.183, simple_loss=0.2764, pruned_loss=0.04485, over 17236.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2931, pruned_loss=0.06024, over 3065779.85 frames. ], batch size: 45, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:54:16,074 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 2.520e+02 3.106e+02 3.673e+02 6.247e+02, threshold=6.213e+02, percent-clipped=0.0 2023-05-01 15:55:16,730 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 15:55:17,141 INFO [train.py:904] (5/8) Epoch 23, batch 5950, loss[loss=0.1691, simple_loss=0.2618, pruned_loss=0.03822, over 16796.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2935, pruned_loss=0.0591, over 3077980.67 frames. ], batch size: 102, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:56:18,286 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229291.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:33,854 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:37,758 INFO [train.py:904] (5/8) Epoch 23, batch 6000, loss[loss=0.2097, simple_loss=0.296, pruned_loss=0.06168, over 16907.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2925, pruned_loss=0.05881, over 3074481.24 frames. ], batch size: 116, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:56:37,759 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 15:56:49,499 INFO [train.py:938] (5/8) Epoch 23, validation: loss=0.1497, simple_loss=0.2623, pruned_loss=0.01859, over 944034.00 frames. 2023-05-01 15:56:49,500 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17672MB 2023-05-01 15:57:07,756 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 2.769e+02 3.211e+02 3.808e+02 9.716e+02, threshold=6.422e+02, percent-clipped=3.0 2023-05-01 15:57:44,729 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5461, 3.7069, 2.8108, 2.2775, 2.4997, 2.3085, 3.9067, 3.3912], device='cuda:5'), covar=tensor([0.3150, 0.0684, 0.1826, 0.2777, 0.2598, 0.2148, 0.0517, 0.1204], device='cuda:5'), in_proj_covar=tensor([0.0328, 0.0270, 0.0306, 0.0315, 0.0298, 0.0261, 0.0299, 0.0339], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 15:58:06,106 INFO [train.py:904] (5/8) Epoch 23, batch 6050, loss[loss=0.1998, simple_loss=0.2951, pruned_loss=0.05227, over 16983.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2914, pruned_loss=0.05845, over 3077949.43 frames. ], batch size: 53, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:58:21,139 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:59:21,998 INFO [train.py:904] (5/8) Epoch 23, batch 6100, loss[loss=0.1984, simple_loss=0.2874, pruned_loss=0.05474, over 17217.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2911, pruned_loss=0.05766, over 3102546.68 frames. ], batch size: 45, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:59:40,510 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.664e+02 3.486e+02 4.048e+02 8.551e+02, threshold=6.973e+02, percent-clipped=3.0 2023-05-01 16:00:22,337 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229443.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:00:27,402 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 16:00:36,345 INFO [train.py:904] (5/8) Epoch 23, batch 6150, loss[loss=0.1808, simple_loss=0.2703, pruned_loss=0.04566, over 17224.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2892, pruned_loss=0.05702, over 3093138.59 frames. ], batch size: 45, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 16:01:15,279 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229477.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:01:53,551 INFO [train.py:904] (5/8) Epoch 23, batch 6200, loss[loss=0.1923, simple_loss=0.2791, pruned_loss=0.05277, over 15443.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.287, pruned_loss=0.05623, over 3098690.57 frames. ], batch size: 190, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:01:55,819 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229504.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:02:13,874 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2891, 3.8799, 3.8348, 2.5007, 3.5096, 3.8952, 3.5446, 2.2184], device='cuda:5'), covar=tensor([0.0575, 0.0053, 0.0055, 0.0446, 0.0098, 0.0111, 0.0099, 0.0456], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0085, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 16:02:14,504 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.788e+02 3.260e+02 3.956e+02 6.763e+02, threshold=6.520e+02, percent-clipped=0.0 2023-05-01 16:02:50,009 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:03:11,726 INFO [train.py:904] (5/8) Epoch 23, batch 6250, loss[loss=0.196, simple_loss=0.2962, pruned_loss=0.04787, over 17109.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2857, pruned_loss=0.05509, over 3115397.58 frames. ], batch size: 48, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:03:44,754 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5375, 2.7267, 2.2845, 2.5559, 3.1379, 2.7557, 3.1630, 3.2780], device='cuda:5'), covar=tensor([0.0131, 0.0433, 0.0563, 0.0449, 0.0244, 0.0393, 0.0246, 0.0267], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0237, 0.0227, 0.0229, 0.0237, 0.0237, 0.0236, 0.0235], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:03:48,949 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:04,852 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-01 16:04:09,830 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:28,569 INFO [train.py:904] (5/8) Epoch 23, batch 6300, loss[loss=0.2186, simple_loss=0.3023, pruned_loss=0.0674, over 15352.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.286, pruned_loss=0.05481, over 3121670.88 frames. ], batch size: 190, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:04:50,640 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.619e+02 3.224e+02 4.167e+02 1.378e+03, threshold=6.449e+02, percent-clipped=8.0 2023-05-01 16:04:52,917 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9512, 4.1966, 4.0216, 4.0685, 3.7997, 3.8495, 3.9071, 4.1824], device='cuda:5'), covar=tensor([0.1085, 0.0862, 0.0972, 0.0827, 0.0737, 0.1622, 0.0871, 0.1024], device='cuda:5'), in_proj_covar=tensor([0.0669, 0.0814, 0.0674, 0.0622, 0.0515, 0.0525, 0.0685, 0.0641], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:05:27,308 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229638.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:28,547 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:48,009 INFO [train.py:904] (5/8) Epoch 23, batch 6350, loss[loss=0.237, simple_loss=0.3025, pruned_loss=0.08573, over 11397.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2867, pruned_loss=0.05604, over 3113528.06 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:05:51,906 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7967, 2.8315, 2.7020, 4.7738, 3.6301, 4.1702, 1.7921, 3.0561], device='cuda:5'), covar=tensor([0.1322, 0.0790, 0.1218, 0.0145, 0.0317, 0.0400, 0.1579, 0.0825], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0176, 0.0197, 0.0193, 0.0207, 0.0217, 0.0205, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:05:53,556 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229656.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:57,382 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229659.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:06:08,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9715, 5.2575, 5.0011, 5.0187, 4.8255, 4.7918, 4.6479, 5.3694], device='cuda:5'), covar=tensor([0.1169, 0.0816, 0.1033, 0.0941, 0.0796, 0.0958, 0.1231, 0.0873], device='cuda:5'), in_proj_covar=tensor([0.0673, 0.0818, 0.0677, 0.0624, 0.0517, 0.0527, 0.0688, 0.0643], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:06:10,752 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5668, 2.4429, 2.2840, 3.5531, 2.5291, 3.7740, 1.4039, 2.6871], device='cuda:5'), covar=tensor([0.1394, 0.0838, 0.1351, 0.0190, 0.0198, 0.0398, 0.1780, 0.0848], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0177, 0.0198, 0.0193, 0.0207, 0.0217, 0.0205, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:06:23,063 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 16:07:04,886 INFO [train.py:904] (5/8) Epoch 23, batch 6400, loss[loss=0.1941, simple_loss=0.2779, pruned_loss=0.05518, over 17027.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2888, pruned_loss=0.0589, over 3066245.73 frames. ], batch size: 53, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:07:24,870 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.907e+02 2.905e+02 3.417e+02 4.252e+02 8.377e+02, threshold=6.835e+02, percent-clipped=2.0 2023-05-01 16:07:32,372 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229720.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:08:21,148 INFO [train.py:904] (5/8) Epoch 23, batch 6450, loss[loss=0.2147, simple_loss=0.3066, pruned_loss=0.06142, over 16858.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2887, pruned_loss=0.05782, over 3082694.35 frames. ], batch size: 42, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:09:07,172 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0447, 2.2807, 2.3617, 2.7523, 2.0738, 3.1985, 1.8655, 2.7373], device='cuda:5'), covar=tensor([0.1153, 0.0628, 0.1077, 0.0160, 0.0123, 0.0454, 0.1497, 0.0717], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0176, 0.0198, 0.0193, 0.0207, 0.0217, 0.0205, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:09:23,110 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-05-01 16:09:34,339 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229799.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:09:40,120 INFO [train.py:904] (5/8) Epoch 23, batch 6500, loss[loss=0.2281, simple_loss=0.2931, pruned_loss=0.0816, over 11878.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2871, pruned_loss=0.05701, over 3095248.70 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:09:59,298 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.807e+02 2.714e+02 3.280e+02 3.954e+02 7.541e+02, threshold=6.561e+02, percent-clipped=1.0 2023-05-01 16:10:01,738 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9505, 3.2516, 3.2001, 2.1281, 3.0128, 3.2367, 3.0400, 1.9538], device='cuda:5'), covar=tensor([0.0608, 0.0064, 0.0079, 0.0463, 0.0123, 0.0121, 0.0121, 0.0497], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0085, 0.0086, 0.0134, 0.0098, 0.0111, 0.0096, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 16:10:25,445 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:10:57,914 INFO [train.py:904] (5/8) Epoch 23, batch 6550, loss[loss=0.2546, simple_loss=0.3162, pruned_loss=0.09652, over 11347.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.29, pruned_loss=0.05791, over 3094120.21 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:02,430 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229894.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:12:15,590 INFO [train.py:904] (5/8) Epoch 23, batch 6600, loss[loss=0.2412, simple_loss=0.3125, pruned_loss=0.08494, over 11851.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2914, pruned_loss=0.05794, over 3088783.74 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:35,466 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.769e+02 2.706e+02 3.377e+02 4.261e+02 8.660e+02, threshold=6.754e+02, percent-clipped=4.0 2023-05-01 16:13:02,431 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:27,880 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8107, 5.1063, 4.8776, 4.8743, 4.6101, 4.5680, 4.5372, 5.1962], device='cuda:5'), covar=tensor([0.1148, 0.0809, 0.0922, 0.0857, 0.0867, 0.1086, 0.1046, 0.0781], device='cuda:5'), in_proj_covar=tensor([0.0674, 0.0814, 0.0675, 0.0624, 0.0518, 0.0525, 0.0686, 0.0642], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:13:33,870 INFO [train.py:904] (5/8) Epoch 23, batch 6650, loss[loss=0.1887, simple_loss=0.2772, pruned_loss=0.05005, over 16350.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2913, pruned_loss=0.05872, over 3081636.68 frames. ], batch size: 35, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:13:37,648 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:38,704 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:26,127 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229987.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:51,668 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0958, 2.3609, 2.3934, 2.7370, 1.9325, 3.1152, 1.9175, 2.7537], device='cuda:5'), covar=tensor([0.1120, 0.0566, 0.1031, 0.0189, 0.0149, 0.0363, 0.1403, 0.0662], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0177, 0.0198, 0.0193, 0.0208, 0.0217, 0.0206, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:14:52,383 INFO [train.py:904] (5/8) Epoch 23, batch 6700, loss[loss=0.2511, simple_loss=0.3152, pruned_loss=0.09353, over 10806.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2898, pruned_loss=0.05902, over 3077229.77 frames. ], batch size: 249, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:14:54,790 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230004.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:55,116 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-01 16:15:11,333 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230015.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:15:12,083 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.536e+02 3.063e+02 3.747e+02 8.042e+02, threshold=6.126e+02, percent-clipped=2.0 2023-05-01 16:16:02,814 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:16:09,700 INFO [train.py:904] (5/8) Epoch 23, batch 6750, loss[loss=0.1936, simple_loss=0.2775, pruned_loss=0.05489, over 16471.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2884, pruned_loss=0.05889, over 3086999.36 frames. ], batch size: 75, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:17:19,763 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230099.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:17:24,166 INFO [train.py:904] (5/8) Epoch 23, batch 6800, loss[loss=0.2438, simple_loss=0.3065, pruned_loss=0.09052, over 11364.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2892, pruned_loss=0.05914, over 3084938.54 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:17:43,641 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 2.682e+02 3.194e+02 4.039e+02 7.212e+02, threshold=6.387e+02, percent-clipped=3.0 2023-05-01 16:18:11,097 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:23,787 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2064, 4.2032, 4.1223, 3.3350, 4.1256, 1.6669, 3.9386, 3.7322], device='cuda:5'), covar=tensor([0.0135, 0.0135, 0.0197, 0.0367, 0.0116, 0.2916, 0.0168, 0.0280], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0176, 0.0207, 0.0188, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:18:31,926 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:36,516 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230150.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:40,529 INFO [train.py:904] (5/8) Epoch 23, batch 6850, loss[loss=0.2601, simple_loss=0.325, pruned_loss=0.09759, over 11520.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2904, pruned_loss=0.05966, over 3071024.67 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:19:19,020 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-05-01 16:19:23,366 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:19:34,543 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4925, 1.7279, 2.1120, 2.3494, 2.4392, 2.6773, 1.7900, 2.6251], device='cuda:5'), covar=tensor([0.0194, 0.0533, 0.0315, 0.0353, 0.0332, 0.0218, 0.0603, 0.0183], device='cuda:5'), in_proj_covar=tensor([0.0190, 0.0193, 0.0179, 0.0185, 0.0199, 0.0155, 0.0198, 0.0154], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:19:56,854 INFO [train.py:904] (5/8) Epoch 23, batch 6900, loss[loss=0.2689, simple_loss=0.3294, pruned_loss=0.1043, over 11213.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2926, pruned_loss=0.05913, over 3090818.67 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:20:11,483 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:20,077 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.571e+02 3.126e+02 3.927e+02 7.395e+02, threshold=6.253e+02, percent-clipped=1.0 2023-05-01 16:20:45,562 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230233.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:59,027 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230241.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:13,090 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230250.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:17,992 INFO [train.py:904] (5/8) Epoch 23, batch 6950, loss[loss=0.1726, simple_loss=0.2677, pruned_loss=0.0387, over 16807.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2933, pruned_loss=0.05971, over 3101162.90 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:22:01,067 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230281.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:34,139 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230302.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:34,887 INFO [train.py:904] (5/8) Epoch 23, batch 7000, loss[loss=0.2156, simple_loss=0.3136, pruned_loss=0.05878, over 16589.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2929, pruned_loss=0.05867, over 3111909.67 frames. ], batch size: 68, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:22:53,536 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230315.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:22:56,098 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.703e+02 3.306e+02 4.193e+02 6.685e+02, threshold=6.612e+02, percent-clipped=2.0 2023-05-01 16:23:18,286 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 16:23:35,807 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230343.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:23:50,825 INFO [train.py:904] (5/8) Epoch 23, batch 7050, loss[loss=0.1989, simple_loss=0.294, pruned_loss=0.05194, over 16381.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2937, pruned_loss=0.05883, over 3093931.50 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:24:05,842 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230363.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:25:07,485 INFO [train.py:904] (5/8) Epoch 23, batch 7100, loss[loss=0.195, simple_loss=0.2816, pruned_loss=0.05417, over 16720.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.292, pruned_loss=0.05868, over 3090051.97 frames. ], batch size: 76, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:25:13,720 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 16:25:30,796 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.925e+02 3.270e+02 3.973e+02 7.550e+02, threshold=6.541e+02, percent-clipped=2.0 2023-05-01 16:25:47,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7714, 3.9696, 2.9474, 2.3709, 2.7382, 2.5840, 4.4008, 3.4900], device='cuda:5'), covar=tensor([0.2904, 0.0642, 0.1910, 0.2562, 0.2667, 0.2027, 0.0440, 0.1340], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0269, 0.0306, 0.0316, 0.0298, 0.0262, 0.0299, 0.0340], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 16:25:50,720 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7973, 3.7524, 3.8758, 3.6812, 3.8485, 4.2287, 3.8881, 3.5951], device='cuda:5'), covar=tensor([0.1986, 0.2357, 0.2771, 0.2386, 0.2541, 0.1819, 0.1744, 0.2425], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0599, 0.0664, 0.0497, 0.0660, 0.0688, 0.0519, 0.0664], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 16:26:24,971 INFO [train.py:904] (5/8) Epoch 23, batch 7150, loss[loss=0.2079, simple_loss=0.3022, pruned_loss=0.05677, over 16854.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2904, pruned_loss=0.05875, over 3092823.99 frames. ], batch size: 96, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:26:47,054 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3865, 3.3592, 3.4121, 3.5104, 3.5296, 3.2989, 3.5174, 3.5912], device='cuda:5'), covar=tensor([0.1269, 0.0954, 0.0995, 0.0587, 0.0667, 0.2162, 0.0969, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0624, 0.0779, 0.0895, 0.0784, 0.0597, 0.0622, 0.0648, 0.0753], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:26:49,365 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:27:42,223 INFO [train.py:904] (5/8) Epoch 23, batch 7200, loss[loss=0.1955, simple_loss=0.2909, pruned_loss=0.05, over 16591.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2894, pruned_loss=0.05778, over 3077508.05 frames. ], batch size: 134, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:27:47,865 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230506.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:03,479 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.554e+02 3.131e+02 3.804e+02 6.997e+02, threshold=6.261e+02, percent-clipped=1.0 2023-05-01 16:28:16,857 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:23,123 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230529.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:28:30,100 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:58,997 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230550.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:29:02,540 INFO [train.py:904] (5/8) Epoch 23, batch 7250, loss[loss=0.1857, simple_loss=0.2737, pruned_loss=0.04881, over 16373.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2872, pruned_loss=0.0567, over 3088738.29 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:29:07,001 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 16:29:15,462 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0283, 3.2345, 3.2285, 2.0739, 2.9728, 3.2310, 3.0526, 1.9245], device='cuda:5'), covar=tensor([0.0572, 0.0070, 0.0075, 0.0479, 0.0128, 0.0134, 0.0124, 0.0494], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0086, 0.0087, 0.0136, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 16:29:51,426 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:03,499 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:08,195 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230597.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:10,106 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230598.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:18,094 INFO [train.py:904] (5/8) Epoch 23, batch 7300, loss[loss=0.2177, simple_loss=0.3071, pruned_loss=0.0642, over 16753.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2869, pruned_loss=0.05662, over 3083390.84 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:30:39,636 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.620e+02 3.228e+02 4.011e+02 8.387e+02, threshold=6.456e+02, percent-clipped=2.0 2023-05-01 16:31:19,690 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230643.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:31:19,984 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 2023-05-01 16:31:33,677 INFO [train.py:904] (5/8) Epoch 23, batch 7350, loss[loss=0.2013, simple_loss=0.2824, pruned_loss=0.0601, over 16795.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2879, pruned_loss=0.0577, over 3072713.79 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:32:33,480 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230691.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:32:53,493 INFO [train.py:904] (5/8) Epoch 23, batch 7400, loss[loss=0.2157, simple_loss=0.3057, pruned_loss=0.06292, over 16325.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.289, pruned_loss=0.05806, over 3092202.41 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:33:16,064 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 2.706e+02 3.245e+02 4.052e+02 5.999e+02, threshold=6.490e+02, percent-clipped=0.0 2023-05-01 16:33:17,908 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8184, 2.0016, 2.3671, 3.0307, 2.1687, 2.2046, 2.2320, 2.0806], device='cuda:5'), covar=tensor([0.1746, 0.4143, 0.2623, 0.0911, 0.4983, 0.2838, 0.3614, 0.4104], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0451, 0.0368, 0.0326, 0.0436, 0.0519, 0.0423, 0.0528], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:34:13,437 INFO [train.py:904] (5/8) Epoch 23, batch 7450, loss[loss=0.2589, simple_loss=0.3135, pruned_loss=0.1022, over 11732.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.29, pruned_loss=0.05889, over 3104881.46 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:35:19,273 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230792.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:35,340 INFO [train.py:904] (5/8) Epoch 23, batch 7500, loss[loss=0.1975, simple_loss=0.2826, pruned_loss=0.05619, over 16544.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2902, pruned_loss=0.05863, over 3078623.01 frames. ], batch size: 75, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:35:40,990 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:41,237 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0950, 2.2008, 2.2101, 3.6375, 2.1819, 2.5876, 2.2605, 2.3457], device='cuda:5'), covar=tensor([0.1359, 0.3372, 0.2918, 0.0622, 0.3895, 0.2385, 0.3386, 0.3268], device='cuda:5'), in_proj_covar=tensor([0.0402, 0.0450, 0.0367, 0.0325, 0.0435, 0.0517, 0.0422, 0.0526], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:35:58,129 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.839e+02 3.579e+02 4.414e+02 1.002e+03, threshold=7.158e+02, percent-clipped=2.0 2023-05-01 16:36:09,661 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230824.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 16:36:52,040 INFO [train.py:904] (5/8) Epoch 23, batch 7550, loss[loss=0.1983, simple_loss=0.2873, pruned_loss=0.05463, over 16721.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2885, pruned_loss=0.05814, over 3099892.15 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:36:52,601 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:36:53,664 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:33,356 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230881.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:35,726 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 16:37:45,742 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:47,097 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:58,359 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230897.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:38:06,500 INFO [train.py:904] (5/8) Epoch 23, batch 7600, loss[loss=0.186, simple_loss=0.2693, pruned_loss=0.0514, over 16508.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2879, pruned_loss=0.05841, over 3104393.23 frames. ], batch size: 68, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:38:10,153 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7802, 4.8392, 4.6728, 4.3138, 4.3586, 4.7464, 4.5525, 4.4378], device='cuda:5'), covar=tensor([0.0583, 0.0478, 0.0306, 0.0314, 0.0888, 0.0458, 0.0459, 0.0648], device='cuda:5'), in_proj_covar=tensor([0.0288, 0.0429, 0.0335, 0.0335, 0.0341, 0.0391, 0.0230, 0.0403], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:38:18,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3556, 5.3715, 5.1370, 4.4481, 5.2497, 1.9432, 5.0275, 4.9761], device='cuda:5'), covar=tensor([0.0099, 0.0093, 0.0177, 0.0392, 0.0090, 0.2545, 0.0109, 0.0192], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0159, 0.0200, 0.0176, 0.0175, 0.0207, 0.0188, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:38:27,916 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.696e+02 3.168e+02 3.892e+02 6.127e+02, threshold=6.336e+02, percent-clipped=0.0 2023-05-01 16:39:10,347 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:19,947 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:21,941 INFO [train.py:904] (5/8) Epoch 23, batch 7650, loss[loss=0.2177, simple_loss=0.3036, pruned_loss=0.06587, over 16700.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2881, pruned_loss=0.05849, over 3114262.40 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:40:35,945 INFO [train.py:904] (5/8) Epoch 23, batch 7700, loss[loss=0.1615, simple_loss=0.2628, pruned_loss=0.03012, over 16848.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2885, pruned_loss=0.05888, over 3101452.07 frames. ], batch size: 96, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:40:57,684 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.849e+02 3.493e+02 4.384e+02 8.642e+02, threshold=6.986e+02, percent-clipped=7.0 2023-05-01 16:41:53,927 INFO [train.py:904] (5/8) Epoch 23, batch 7750, loss[loss=0.1938, simple_loss=0.286, pruned_loss=0.05087, over 16726.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2885, pruned_loss=0.05847, over 3107489.73 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:42:58,123 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9599, 3.2030, 3.3411, 1.9652, 2.9209, 2.1906, 3.4549, 3.4984], device='cuda:5'), covar=tensor([0.0256, 0.0865, 0.0633, 0.2283, 0.0934, 0.1052, 0.0643, 0.0969], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0165, 0.0168, 0.0154, 0.0146, 0.0130, 0.0143, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:43:09,649 INFO [train.py:904] (5/8) Epoch 23, batch 7800, loss[loss=0.1782, simple_loss=0.2718, pruned_loss=0.04229, over 16429.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2895, pruned_loss=0.05923, over 3083891.16 frames. ], batch size: 68, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:43:28,613 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 16:43:30,327 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.907e+02 3.422e+02 4.020e+02 9.124e+02, threshold=6.845e+02, percent-clipped=1.0 2023-05-01 16:43:41,730 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231124.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:44:17,336 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:44:24,898 INFO [train.py:904] (5/8) Epoch 23, batch 7850, loss[loss=0.1734, simple_loss=0.2621, pruned_loss=0.04231, over 16478.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2908, pruned_loss=0.05999, over 3055695.43 frames. ], batch size: 75, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:44:52,236 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:05,973 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:16,810 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231189.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:37,203 INFO [train.py:904] (5/8) Epoch 23, batch 7900, loss[loss=0.2066, simple_loss=0.304, pruned_loss=0.0546, over 16728.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2899, pruned_loss=0.05896, over 3088526.94 frames. ], batch size: 83, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:45:57,447 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.682e+02 3.251e+02 4.238e+02 6.670e+02, threshold=6.501e+02, percent-clipped=0.0 2023-05-01 16:46:16,892 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231229.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:31,363 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:44,782 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:52,768 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231251.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:54,848 INFO [train.py:904] (5/8) Epoch 23, batch 7950, loss[loss=0.181, simple_loss=0.2698, pruned_loss=0.04611, over 16316.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2902, pruned_loss=0.05967, over 3067355.69 frames. ], batch size: 35, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:47:19,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9524, 2.7798, 2.7934, 2.1318, 2.6775, 2.1137, 2.7733, 2.9770], device='cuda:5'), covar=tensor([0.0295, 0.0761, 0.0540, 0.1823, 0.0843, 0.0973, 0.0599, 0.0699], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0155, 0.0146, 0.0130, 0.0144, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:48:11,523 INFO [train.py:904] (5/8) Epoch 23, batch 8000, loss[loss=0.2638, simple_loss=0.323, pruned_loss=0.1023, over 11404.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2908, pruned_loss=0.06045, over 3054240.64 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:48:26,462 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231312.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:48:26,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5593, 2.6361, 2.3689, 3.9355, 2.6701, 3.8146, 1.5619, 2.7565], device='cuda:5'), covar=tensor([0.1525, 0.0803, 0.1336, 0.0244, 0.0232, 0.0428, 0.1739, 0.0907], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0175, 0.0196, 0.0191, 0.0207, 0.0215, 0.0203, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:48:33,455 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.670e+02 3.387e+02 4.030e+02 6.156e+02, threshold=6.774e+02, percent-clipped=0.0 2023-05-01 16:49:26,912 INFO [train.py:904] (5/8) Epoch 23, batch 8050, loss[loss=0.1942, simple_loss=0.2928, pruned_loss=0.04776, over 16528.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2907, pruned_loss=0.06, over 3057767.12 frames. ], batch size: 68, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:50:04,057 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5141, 5.8721, 5.6084, 5.6186, 5.2253, 5.2822, 5.2720, 5.9813], device='cuda:5'), covar=tensor([0.1154, 0.0773, 0.0980, 0.0934, 0.0890, 0.0653, 0.1231, 0.0871], device='cuda:5'), in_proj_covar=tensor([0.0672, 0.0814, 0.0678, 0.0627, 0.0518, 0.0530, 0.0689, 0.0645], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:50:42,768 INFO [train.py:904] (5/8) Epoch 23, batch 8100, loss[loss=0.2094, simple_loss=0.2846, pruned_loss=0.06712, over 11728.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2897, pruned_loss=0.05904, over 3067892.14 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:50:43,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4997, 1.6781, 2.1105, 2.3682, 2.4815, 2.7394, 1.6799, 2.6280], device='cuda:5'), covar=tensor([0.0228, 0.0613, 0.0344, 0.0400, 0.0332, 0.0211, 0.0669, 0.0190], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0191, 0.0177, 0.0183, 0.0197, 0.0154, 0.0197, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:51:04,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.502e+02 3.046e+02 3.676e+02 7.002e+02, threshold=6.092e+02, percent-clipped=1.0 2023-05-01 16:51:09,673 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8433, 4.8917, 4.7685, 4.4208, 4.4190, 4.8412, 4.6442, 4.5054], device='cuda:5'), covar=tensor([0.0567, 0.0425, 0.0295, 0.0316, 0.0926, 0.0427, 0.0414, 0.0664], device='cuda:5'), in_proj_covar=tensor([0.0289, 0.0434, 0.0337, 0.0337, 0.0343, 0.0394, 0.0232, 0.0407], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:51:51,899 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:51:59,319 INFO [train.py:904] (5/8) Epoch 23, batch 8150, loss[loss=0.1787, simple_loss=0.2655, pruned_loss=0.04593, over 16700.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2869, pruned_loss=0.05776, over 3076228.15 frames. ], batch size: 89, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:52:03,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5125, 3.4671, 3.4504, 2.6965, 3.3175, 2.0472, 3.1107, 2.8757], device='cuda:5'), covar=tensor([0.0168, 0.0143, 0.0199, 0.0222, 0.0110, 0.2319, 0.0144, 0.0260], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0159, 0.0201, 0.0178, 0.0176, 0.0208, 0.0188, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:53:05,174 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:53:14,931 INFO [train.py:904] (5/8) Epoch 23, batch 8200, loss[loss=0.1723, simple_loss=0.2643, pruned_loss=0.04016, over 16448.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2841, pruned_loss=0.05681, over 3078222.20 frames. ], batch size: 68, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:53:30,107 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7503, 1.8393, 1.6536, 1.5321, 1.9839, 1.6672, 1.6353, 1.9346], device='cuda:5'), covar=tensor([0.0240, 0.0306, 0.0445, 0.0373, 0.0250, 0.0292, 0.0206, 0.0204], device='cuda:5'), in_proj_covar=tensor([0.0212, 0.0234, 0.0225, 0.0227, 0.0235, 0.0233, 0.0233, 0.0231], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 16:53:38,108 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.780e+02 3.369e+02 4.151e+02 6.479e+02, threshold=6.737e+02, percent-clipped=3.0 2023-05-01 16:54:26,264 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:54:37,502 INFO [train.py:904] (5/8) Epoch 23, batch 8250, loss[loss=0.1538, simple_loss=0.246, pruned_loss=0.03084, over 11995.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2827, pruned_loss=0.0543, over 3054149.87 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:55:05,258 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 16:55:43,804 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:55:57,545 INFO [train.py:904] (5/8) Epoch 23, batch 8300, loss[loss=0.1703, simple_loss=0.2561, pruned_loss=0.04224, over 12238.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2804, pruned_loss=0.0517, over 3035356.84 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:56:05,106 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231607.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:56:22,165 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.236e+02 2.512e+02 2.973e+02 5.364e+02, threshold=5.024e+02, percent-clipped=0.0 2023-05-01 16:56:25,477 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1979, 3.2385, 2.0439, 3.5079, 2.4320, 3.5010, 2.0887, 2.6805], device='cuda:5'), covar=tensor([0.0303, 0.0337, 0.1509, 0.0241, 0.0853, 0.0511, 0.1516, 0.0738], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0175, 0.0194, 0.0163, 0.0176, 0.0216, 0.0203, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:56:39,936 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-01 16:57:03,047 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3431, 3.4708, 3.6641, 3.6442, 3.6609, 3.5092, 3.4860, 3.5548], device='cuda:5'), covar=tensor([0.0483, 0.0829, 0.0558, 0.0491, 0.0545, 0.0657, 0.0930, 0.0551], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0456, 0.0443, 0.0411, 0.0488, 0.0464, 0.0548, 0.0370], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 16:57:11,265 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231647.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:57:19,287 INFO [train.py:904] (5/8) Epoch 23, batch 8350, loss[loss=0.193, simple_loss=0.2754, pruned_loss=0.05536, over 12262.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2808, pruned_loss=0.05005, over 3053154.76 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:57:34,442 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5065, 3.6286, 2.3610, 4.0052, 2.7988, 3.9429, 2.3737, 2.9250], device='cuda:5'), covar=tensor([0.0310, 0.0340, 0.1428, 0.0256, 0.0813, 0.0576, 0.1505, 0.0733], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0174, 0.0193, 0.0163, 0.0176, 0.0215, 0.0202, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 16:58:41,953 INFO [train.py:904] (5/8) Epoch 23, batch 8400, loss[loss=0.1694, simple_loss=0.2688, pruned_loss=0.03505, over 16730.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2778, pruned_loss=0.04788, over 3048920.85 frames. ], batch size: 83, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:58:50,882 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231708.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 16:59:06,592 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.191e+02 2.741e+02 3.304e+02 6.516e+02, threshold=5.483e+02, percent-clipped=5.0 2023-05-01 17:00:04,856 INFO [train.py:904] (5/8) Epoch 23, batch 8450, loss[loss=0.1868, simple_loss=0.2888, pruned_loss=0.04243, over 16320.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.276, pruned_loss=0.04625, over 3048850.48 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:24,145 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8959, 4.2035, 4.0479, 4.0745, 3.7377, 3.8189, 3.8386, 4.1984], device='cuda:5'), covar=tensor([0.1171, 0.0989, 0.1007, 0.0804, 0.0871, 0.1680, 0.1042, 0.1042], device='cuda:5'), in_proj_covar=tensor([0.0670, 0.0811, 0.0671, 0.0621, 0.0514, 0.0528, 0.0683, 0.0639], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:01:24,910 INFO [train.py:904] (5/8) Epoch 23, batch 8500, loss[loss=0.1618, simple_loss=0.2403, pruned_loss=0.04163, over 11592.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2722, pruned_loss=0.04389, over 3059899.05 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:27,234 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6790, 2.6142, 1.8650, 2.7842, 2.1233, 2.7905, 2.1598, 2.4461], device='cuda:5'), covar=tensor([0.0305, 0.0361, 0.1299, 0.0276, 0.0616, 0.0458, 0.1202, 0.0554], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0174, 0.0192, 0.0162, 0.0175, 0.0214, 0.0201, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 17:01:43,204 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5380, 4.8630, 4.6923, 4.6671, 4.3925, 4.3557, 4.3314, 4.9302], device='cuda:5'), covar=tensor([0.1228, 0.0949, 0.1005, 0.0874, 0.0843, 0.1269, 0.1279, 0.0948], device='cuda:5'), in_proj_covar=tensor([0.0671, 0.0812, 0.0673, 0.0622, 0.0515, 0.0529, 0.0684, 0.0640], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:01:48,518 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 2.174e+02 2.671e+02 3.276e+02 7.093e+02, threshold=5.341e+02, percent-clipped=2.0 2023-05-01 17:02:48,227 INFO [train.py:904] (5/8) Epoch 23, batch 8550, loss[loss=0.1993, simple_loss=0.2932, pruned_loss=0.05273, over 16408.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2709, pruned_loss=0.04329, over 3050031.89 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:03:08,867 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 17:04:28,278 INFO [train.py:904] (5/8) Epoch 23, batch 8600, loss[loss=0.1593, simple_loss=0.2616, pruned_loss=0.02847, over 16656.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2708, pruned_loss=0.04211, over 3037413.53 frames. ], batch size: 89, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:04:36,665 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:05:00,381 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.309e+02 2.625e+02 3.302e+02 5.565e+02, threshold=5.250e+02, percent-clipped=1.0 2023-05-01 17:06:03,609 INFO [train.py:904] (5/8) Epoch 23, batch 8650, loss[loss=0.1651, simple_loss=0.2649, pruned_loss=0.03266, over 16202.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2691, pruned_loss=0.04067, over 3042759.85 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:06:10,285 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:06:20,959 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 17:07:42,601 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7771, 3.8901, 4.0043, 2.9275, 3.4999, 3.9640, 3.6996, 2.3851], device='cuda:5'), covar=tensor([0.0415, 0.0061, 0.0044, 0.0323, 0.0121, 0.0086, 0.0071, 0.0483], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0083, 0.0084, 0.0132, 0.0097, 0.0108, 0.0093, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 17:07:53,209 INFO [train.py:904] (5/8) Epoch 23, batch 8700, loss[loss=0.1677, simple_loss=0.2645, pruned_loss=0.03544, over 16409.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2668, pruned_loss=0.03961, over 3055310.12 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:07:54,770 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232003.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:08:23,062 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 2.142e+02 2.520e+02 3.226e+02 6.058e+02, threshold=5.039e+02, percent-clipped=2.0 2023-05-01 17:09:27,791 INFO [train.py:904] (5/8) Epoch 23, batch 8750, loss[loss=0.1789, simple_loss=0.2781, pruned_loss=0.03984, over 16761.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2664, pruned_loss=0.03923, over 3048651.47 frames. ], batch size: 124, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:11:15,141 INFO [train.py:904] (5/8) Epoch 23, batch 8800, loss[loss=0.161, simple_loss=0.2591, pruned_loss=0.0314, over 16252.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2639, pruned_loss=0.03804, over 3037225.44 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:11:29,825 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1624, 3.2167, 1.7301, 3.4951, 2.3911, 3.4487, 1.9806, 2.6324], device='cuda:5'), covar=tensor([0.0348, 0.0387, 0.2012, 0.0248, 0.0950, 0.0581, 0.1780, 0.0846], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0172, 0.0191, 0.0160, 0.0173, 0.0211, 0.0200, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 17:11:46,653 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.145e+02 2.530e+02 3.020e+02 7.216e+02, threshold=5.060e+02, percent-clipped=1.0 2023-05-01 17:12:28,662 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4372, 3.5200, 3.6962, 3.6751, 3.6974, 3.5410, 3.5736, 3.5738], device='cuda:5'), covar=tensor([0.0365, 0.0591, 0.0440, 0.0426, 0.0378, 0.0453, 0.0645, 0.0431], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0447, 0.0436, 0.0403, 0.0479, 0.0454, 0.0536, 0.0365], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 17:12:57,917 INFO [train.py:904] (5/8) Epoch 23, batch 8850, loss[loss=0.1889, simple_loss=0.2915, pruned_loss=0.04311, over 16093.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2669, pruned_loss=0.0378, over 3022673.02 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:14:42,076 INFO [train.py:904] (5/8) Epoch 23, batch 8900, loss[loss=0.1712, simple_loss=0.2708, pruned_loss=0.0358, over 16890.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.267, pruned_loss=0.0372, over 3025268.89 frames. ], batch size: 102, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:15:12,047 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.232e+02 2.655e+02 3.319e+02 5.405e+02, threshold=5.309e+02, percent-clipped=1.0 2023-05-01 17:15:12,633 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3622, 6.0007, 6.1994, 5.8855, 5.9482, 6.4429, 6.0441, 5.7245], device='cuda:5'), covar=tensor([0.0815, 0.1694, 0.1886, 0.1728, 0.2370, 0.0860, 0.1278, 0.2135], device='cuda:5'), in_proj_covar=tensor([0.0396, 0.0576, 0.0637, 0.0475, 0.0634, 0.0663, 0.0497, 0.0638], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 17:16:22,571 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5262, 4.4835, 4.3432, 3.7735, 4.4361, 1.6837, 4.1420, 4.1689], device='cuda:5'), covar=tensor([0.0098, 0.0088, 0.0192, 0.0314, 0.0098, 0.2673, 0.0142, 0.0217], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0157, 0.0197, 0.0174, 0.0174, 0.0206, 0.0186, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:16:33,777 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:16:45,336 INFO [train.py:904] (5/8) Epoch 23, batch 8950, loss[loss=0.1556, simple_loss=0.2642, pruned_loss=0.02356, over 16862.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2659, pruned_loss=0.03698, over 3051005.36 frames. ], batch size: 90, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:18:31,856 INFO [train.py:904] (5/8) Epoch 23, batch 9000, loss[loss=0.135, simple_loss=0.2345, pruned_loss=0.01773, over 16859.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.263, pruned_loss=0.03587, over 3053964.47 frames. ], batch size: 102, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:18:31,857 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 17:18:38,733 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3588, 5.9087, 5.9010, 5.6002, 5.6393, 6.1534, 5.7126, 5.4863], device='cuda:5'), covar=tensor([0.0640, 0.1636, 0.1960, 0.1461, 0.2094, 0.0611, 0.1227, 0.1941], device='cuda:5'), in_proj_covar=tensor([0.0395, 0.0576, 0.0636, 0.0476, 0.0633, 0.0662, 0.0497, 0.0637], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 17:18:42,674 INFO [train.py:938] (5/8) Epoch 23, validation: loss=0.1452, simple_loss=0.249, pruned_loss=0.02066, over 944034.00 frames. 2023-05-01 17:18:42,676 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-01 17:18:43,589 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232303.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:18:43,760 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9022, 2.8515, 2.5137, 4.5400, 2.8788, 4.1335, 1.6912, 3.0274], device='cuda:5'), covar=tensor([0.1449, 0.0844, 0.1284, 0.0159, 0.0157, 0.0365, 0.1782, 0.0764], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0173, 0.0194, 0.0187, 0.0201, 0.0212, 0.0202, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 17:18:54,273 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232308.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:19:18,068 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 1.963e+02 2.332e+02 2.758e+02 5.487e+02, threshold=4.665e+02, percent-clipped=1.0 2023-05-01 17:19:44,744 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4544, 3.5011, 3.6934, 3.6729, 3.6840, 3.5298, 3.5717, 3.5982], device='cuda:5'), covar=tensor([0.0383, 0.0885, 0.0511, 0.0477, 0.0487, 0.0542, 0.0691, 0.0443], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0445, 0.0437, 0.0403, 0.0478, 0.0453, 0.0534, 0.0365], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 17:20:24,509 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232351.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:20:28,125 INFO [train.py:904] (5/8) Epoch 23, batch 9050, loss[loss=0.1491, simple_loss=0.2401, pruned_loss=0.02901, over 15421.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2646, pruned_loss=0.03672, over 3046845.61 frames. ], batch size: 191, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:21:20,944 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 17:21:23,529 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:21:34,413 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-01 17:22:12,822 INFO [train.py:904] (5/8) Epoch 23, batch 9100, loss[loss=0.1659, simple_loss=0.2533, pruned_loss=0.03929, over 12299.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2642, pruned_loss=0.03717, over 3062394.24 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:22:24,031 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 17:22:46,301 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.157e+02 2.616e+02 3.171e+02 8.876e+02, threshold=5.231e+02, percent-clipped=3.0 2023-05-01 17:23:46,408 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232442.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:24:09,379 INFO [train.py:904] (5/8) Epoch 23, batch 9150, loss[loss=0.1421, simple_loss=0.2404, pruned_loss=0.02189, over 16867.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2644, pruned_loss=0.03667, over 3064948.69 frames. ], batch size: 90, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:25:00,462 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:25:09,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8693, 2.2374, 1.9190, 2.0390, 2.5512, 2.2066, 2.2578, 2.6958], device='cuda:5'), covar=tensor([0.0154, 0.0511, 0.0566, 0.0553, 0.0324, 0.0487, 0.0184, 0.0310], device='cuda:5'), in_proj_covar=tensor([0.0204, 0.0228, 0.0220, 0.0222, 0.0229, 0.0229, 0.0225, 0.0224], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:25:52,929 INFO [train.py:904] (5/8) Epoch 23, batch 9200, loss[loss=0.1671, simple_loss=0.2456, pruned_loss=0.04431, over 12210.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.261, pruned_loss=0.03614, over 3063377.11 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:25:58,372 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 17:26:21,402 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:26:24,111 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 2.032e+02 2.532e+02 2.962e+02 8.451e+02, threshold=5.064e+02, percent-clipped=2.0 2023-05-01 17:26:55,365 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7319, 2.4536, 2.3062, 3.5889, 2.0725, 3.6477, 1.5289, 2.8696], device='cuda:5'), covar=tensor([0.1477, 0.0814, 0.1247, 0.0171, 0.0099, 0.0346, 0.1767, 0.0755], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0172, 0.0194, 0.0186, 0.0200, 0.0211, 0.0201, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 17:26:57,083 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:27:18,208 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 17:27:27,739 INFO [train.py:904] (5/8) Epoch 23, batch 9250, loss[loss=0.1511, simple_loss=0.249, pruned_loss=0.02665, over 16808.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2608, pruned_loss=0.03611, over 3062782.84 frames. ], batch size: 96, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:28:23,218 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232579.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:04,377 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 17:29:17,466 INFO [train.py:904] (5/8) Epoch 23, batch 9300, loss[loss=0.1604, simple_loss=0.2462, pruned_loss=0.03729, over 16655.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2597, pruned_loss=0.03558, over 3071297.05 frames. ], batch size: 62, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:29:18,451 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232603.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:41,477 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7530, 4.8292, 4.6590, 4.2553, 4.3209, 4.7271, 4.4932, 4.4277], device='cuda:5'), covar=tensor([0.0532, 0.0493, 0.0314, 0.0318, 0.0906, 0.0463, 0.0420, 0.0699], device='cuda:5'), in_proj_covar=tensor([0.0284, 0.0420, 0.0329, 0.0328, 0.0332, 0.0384, 0.0226, 0.0396], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 17:29:58,639 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.317e+02 2.074e+02 2.482e+02 2.986e+02 5.209e+02, threshold=4.963e+02, percent-clipped=1.0 2023-05-01 17:30:06,078 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9194, 2.3002, 2.2560, 3.0890, 1.8510, 3.2783, 1.6542, 2.8124], device='cuda:5'), covar=tensor([0.1276, 0.0753, 0.1099, 0.0165, 0.0092, 0.0368, 0.1602, 0.0678], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0173, 0.0194, 0.0186, 0.0199, 0.0211, 0.0201, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 17:31:04,807 INFO [train.py:904] (5/8) Epoch 23, batch 9350, loss[loss=0.1736, simple_loss=0.2613, pruned_loss=0.0429, over 16876.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2593, pruned_loss=0.0355, over 3085275.79 frames. ], batch size: 116, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:31:53,008 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5805, 3.4590, 3.5121, 2.7792, 3.3472, 2.0573, 3.1350, 2.9517], device='cuda:5'), covar=tensor([0.0152, 0.0197, 0.0171, 0.0235, 0.0128, 0.2283, 0.0153, 0.0230], device='cuda:5'), in_proj_covar=tensor([0.0164, 0.0156, 0.0195, 0.0171, 0.0173, 0.0204, 0.0184, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:32:47,943 INFO [train.py:904] (5/8) Epoch 23, batch 9400, loss[loss=0.15, simple_loss=0.2366, pruned_loss=0.03171, over 12575.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2584, pruned_loss=0.03507, over 3078904.20 frames. ], batch size: 247, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:33:21,843 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.048e+02 2.308e+02 2.875e+02 5.038e+02, threshold=4.615e+02, percent-clipped=1.0 2023-05-01 17:33:57,952 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232737.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:34:29,652 INFO [train.py:904] (5/8) Epoch 23, batch 9450, loss[loss=0.1745, simple_loss=0.2689, pruned_loss=0.0401, over 16173.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2602, pruned_loss=0.03501, over 3093722.36 frames. ], batch size: 165, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:35:12,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8712, 2.4328, 2.5302, 1.8330, 2.5652, 2.8899, 2.5439, 2.3516], device='cuda:5'), covar=tensor([0.1038, 0.0219, 0.0226, 0.1191, 0.0119, 0.0234, 0.0445, 0.0566], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0103, 0.0093, 0.0134, 0.0078, 0.0120, 0.0122, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 17:36:09,224 INFO [train.py:904] (5/8) Epoch 23, batch 9500, loss[loss=0.1483, simple_loss=0.2343, pruned_loss=0.0311, over 12821.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2594, pruned_loss=0.03491, over 3086361.80 frames. ], batch size: 247, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:36:44,543 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.194e+02 2.525e+02 3.286e+02 6.197e+02, threshold=5.051e+02, percent-clipped=6.0 2023-05-01 17:37:08,361 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:37:54,059 INFO [train.py:904] (5/8) Epoch 23, batch 9550, loss[loss=0.1828, simple_loss=0.2839, pruned_loss=0.04079, over 16189.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2589, pruned_loss=0.03509, over 3069989.24 frames. ], batch size: 165, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:38:38,057 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:39:33,910 INFO [train.py:904] (5/8) Epoch 23, batch 9600, loss[loss=0.1563, simple_loss=0.2434, pruned_loss=0.03455, over 12203.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2605, pruned_loss=0.03617, over 3050997.17 frames. ], batch size: 247, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:39:34,742 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232903.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:40:05,777 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.131e+02 2.488e+02 2.932e+02 6.217e+02, threshold=4.975e+02, percent-clipped=3.0 2023-05-01 17:41:17,236 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:41:20,698 INFO [train.py:904] (5/8) Epoch 23, batch 9650, loss[loss=0.1726, simple_loss=0.2702, pruned_loss=0.03748, over 15358.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2627, pruned_loss=0.03617, over 3068748.14 frames. ], batch size: 191, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:41:48,115 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 17:42:39,069 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0102, 5.2745, 5.0866, 5.0736, 4.8065, 4.7459, 4.6102, 5.3635], device='cuda:5'), covar=tensor([0.1197, 0.0861, 0.0951, 0.0821, 0.0708, 0.0967, 0.1225, 0.0926], device='cuda:5'), in_proj_covar=tensor([0.0663, 0.0804, 0.0663, 0.0616, 0.0510, 0.0521, 0.0677, 0.0635], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:42:42,002 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:09,193 INFO [train.py:904] (5/8) Epoch 23, batch 9700, loss[loss=0.1687, simple_loss=0.2761, pruned_loss=0.03066, over 16848.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2619, pruned_loss=0.036, over 3075050.99 frames. ], batch size: 102, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:43:40,111 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233018.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:43,788 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.165e+02 2.462e+02 3.116e+02 5.593e+02, threshold=4.924e+02, percent-clipped=3.0 2023-05-01 17:44:17,996 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1666, 2.4504, 2.0714, 2.1676, 2.7771, 2.4761, 2.6751, 2.9342], device='cuda:5'), covar=tensor([0.0159, 0.0442, 0.0550, 0.0520, 0.0285, 0.0403, 0.0244, 0.0256], device='cuda:5'), in_proj_covar=tensor([0.0203, 0.0227, 0.0219, 0.0220, 0.0227, 0.0227, 0.0221, 0.0221], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:44:22,593 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:44:50,832 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233051.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:44:53,550 INFO [train.py:904] (5/8) Epoch 23, batch 9750, loss[loss=0.1844, simple_loss=0.2776, pruned_loss=0.04558, over 15311.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2601, pruned_loss=0.0359, over 3051181.56 frames. ], batch size: 190, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:44:54,297 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4061, 3.3258, 2.7368, 2.1654, 2.1599, 2.3429, 3.5073, 2.9833], device='cuda:5'), covar=tensor([0.3179, 0.0668, 0.1772, 0.2834, 0.2956, 0.2264, 0.0429, 0.1423], device='cuda:5'), in_proj_covar=tensor([0.0321, 0.0261, 0.0299, 0.0308, 0.0286, 0.0256, 0.0289, 0.0329], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 17:45:14,280 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6816, 2.4865, 2.2600, 4.1488, 2.4872, 3.8923, 1.2794, 2.7460], device='cuda:5'), covar=tensor([0.1512, 0.0924, 0.1398, 0.0167, 0.0190, 0.0404, 0.1998, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0172, 0.0194, 0.0185, 0.0198, 0.0211, 0.0202, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 17:45:43,973 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:45:59,983 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233085.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:46:29,594 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2250, 4.3106, 4.1520, 3.8513, 3.8815, 4.2218, 3.8850, 3.9397], device='cuda:5'), covar=tensor([0.0581, 0.0667, 0.0309, 0.0294, 0.0774, 0.0509, 0.0842, 0.0600], device='cuda:5'), in_proj_covar=tensor([0.0281, 0.0415, 0.0328, 0.0326, 0.0329, 0.0380, 0.0224, 0.0391], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 17:46:31,580 INFO [train.py:904] (5/8) Epoch 23, batch 9800, loss[loss=0.1718, simple_loss=0.275, pruned_loss=0.03428, over 15416.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2607, pruned_loss=0.03555, over 3063953.10 frames. ], batch size: 190, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:47:03,693 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.119e+02 2.583e+02 3.356e+02 7.260e+02, threshold=5.167e+02, percent-clipped=1.0 2023-05-01 17:47:26,798 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:47:35,946 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4205, 3.3050, 3.3607, 3.5512, 3.5445, 3.3222, 3.5493, 3.6345], device='cuda:5'), covar=tensor([0.1458, 0.1322, 0.1611, 0.0951, 0.0946, 0.3086, 0.1267, 0.1056], device='cuda:5'), in_proj_covar=tensor([0.0614, 0.0757, 0.0871, 0.0768, 0.0584, 0.0608, 0.0636, 0.0740], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:47:47,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6203, 3.6881, 3.4828, 3.1658, 3.3268, 3.5668, 3.3712, 3.3998], device='cuda:5'), covar=tensor([0.0522, 0.0549, 0.0283, 0.0238, 0.0461, 0.0414, 0.1182, 0.0426], device='cuda:5'), in_proj_covar=tensor([0.0280, 0.0412, 0.0327, 0.0324, 0.0327, 0.0378, 0.0222, 0.0389], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 17:48:10,593 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3651, 3.6233, 3.8316, 2.1349, 3.2603, 2.6486, 3.7418, 3.8655], device='cuda:5'), covar=tensor([0.0253, 0.0764, 0.0561, 0.2187, 0.0785, 0.0937, 0.0629, 0.0894], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0157, 0.0163, 0.0150, 0.0142, 0.0126, 0.0139, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 17:48:16,806 INFO [train.py:904] (5/8) Epoch 23, batch 9850, loss[loss=0.1547, simple_loss=0.2567, pruned_loss=0.0264, over 15412.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2621, pruned_loss=0.0353, over 3073942.08 frames. ], batch size: 191, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:49:01,285 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233174.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:49:13,855 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233180.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:50:08,901 INFO [train.py:904] (5/8) Epoch 23, batch 9900, loss[loss=0.1619, simple_loss=0.2667, pruned_loss=0.02852, over 16324.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2618, pruned_loss=0.03515, over 3042491.11 frames. ], batch size: 166, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:50:21,529 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1842, 3.1729, 2.0120, 3.5080, 2.3773, 3.4694, 2.2005, 2.6959], device='cuda:5'), covar=tensor([0.0312, 0.0427, 0.1618, 0.0261, 0.0894, 0.0627, 0.1483, 0.0743], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0168, 0.0186, 0.0156, 0.0169, 0.0205, 0.0197, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 17:50:46,350 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.034e+02 2.487e+02 3.058e+02 5.815e+02, threshold=4.974e+02, percent-clipped=3.0 2023-05-01 17:50:52,785 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:51:09,698 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3896, 3.4299, 2.1570, 3.9055, 2.5630, 3.8170, 2.3447, 2.7841], device='cuda:5'), covar=tensor([0.0334, 0.0415, 0.1669, 0.0223, 0.0892, 0.0599, 0.1488, 0.0825], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0168, 0.0186, 0.0155, 0.0169, 0.0205, 0.0197, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-01 17:51:45,265 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5252, 1.7045, 2.0807, 2.5122, 2.4452, 2.7795, 1.8428, 2.7725], device='cuda:5'), covar=tensor([0.0217, 0.0582, 0.0377, 0.0321, 0.0369, 0.0209, 0.0558, 0.0162], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0187, 0.0173, 0.0177, 0.0192, 0.0149, 0.0192, 0.0147], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:52:06,040 INFO [train.py:904] (5/8) Epoch 23, batch 9950, loss[loss=0.1665, simple_loss=0.262, pruned_loss=0.0355, over 16903.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2634, pruned_loss=0.03507, over 3064742.58 frames. ], batch size: 116, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:52:10,795 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2097, 3.3023, 3.6301, 1.9970, 3.1315, 2.3814, 3.7054, 3.6565], device='cuda:5'), covar=tensor([0.0206, 0.0892, 0.0601, 0.2172, 0.0756, 0.0974, 0.0542, 0.0835], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0156, 0.0162, 0.0149, 0.0141, 0.0125, 0.0139, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 17:52:18,788 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4815, 2.5688, 2.1698, 2.2821, 2.9535, 2.6638, 2.9429, 3.1326], device='cuda:5'), covar=tensor([0.0146, 0.0458, 0.0578, 0.0516, 0.0271, 0.0406, 0.0304, 0.0262], device='cuda:5'), in_proj_covar=tensor([0.0205, 0.0229, 0.0221, 0.0223, 0.0229, 0.0230, 0.0223, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:52:20,937 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:53:11,708 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-01 17:54:07,454 INFO [train.py:904] (5/8) Epoch 23, batch 10000, loss[loss=0.1602, simple_loss=0.2616, pruned_loss=0.0294, over 16655.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2622, pruned_loss=0.03477, over 3075998.08 frames. ], batch size: 76, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:54:17,935 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 17:54:40,345 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 2.128e+02 2.777e+02 3.344e+02 7.240e+02, threshold=5.555e+02, percent-clipped=2.0 2023-05-01 17:54:41,091 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:55:00,700 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4857, 2.0570, 1.8274, 1.8014, 2.2928, 1.9798, 1.8945, 2.3432], device='cuda:5'), covar=tensor([0.0196, 0.0424, 0.0524, 0.0476, 0.0296, 0.0404, 0.0242, 0.0290], device='cuda:5'), in_proj_covar=tensor([0.0204, 0.0229, 0.0221, 0.0223, 0.0229, 0.0230, 0.0223, 0.0223], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 17:55:36,826 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233346.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:55:47,980 INFO [train.py:904] (5/8) Epoch 23, batch 10050, loss[loss=0.1858, simple_loss=0.2829, pruned_loss=0.0444, over 16279.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2625, pruned_loss=0.03468, over 3076449.48 frames. ], batch size: 165, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:56:17,032 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 17:56:31,051 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:57:20,842 INFO [train.py:904] (5/8) Epoch 23, batch 10100, loss[loss=0.1518, simple_loss=0.2448, pruned_loss=0.02938, over 16683.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.263, pruned_loss=0.03504, over 3075744.23 frames. ], batch size: 57, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:57:28,139 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 17:57:37,556 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 17:57:53,983 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.457e+02 2.206e+02 2.577e+02 3.013e+02 4.747e+02, threshold=5.154e+02, percent-clipped=0.0 2023-05-01 17:59:06,342 INFO [train.py:904] (5/8) Epoch 24, batch 0, loss[loss=0.2257, simple_loss=0.2995, pruned_loss=0.07598, over 16338.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.2995, pruned_loss=0.07598, over 16338.00 frames. ], batch size: 165, lr: 2.84e-03, grad_scale: 8.0 2023-05-01 17:59:06,342 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 17:59:14,244 INFO [train.py:938] (5/8) Epoch 24, validation: loss=0.145, simple_loss=0.2483, pruned_loss=0.02085, over 944034.00 frames. 2023-05-01 17:59:14,244 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-01 18:00:00,002 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8542, 2.7607, 2.8995, 4.9665, 3.8618, 4.3846, 1.7051, 3.3415], device='cuda:5'), covar=tensor([0.1538, 0.0979, 0.1290, 0.0250, 0.0235, 0.0434, 0.1912, 0.0782], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0172, 0.0194, 0.0185, 0.0197, 0.0211, 0.0202, 0.0191], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:00:18,715 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8813, 4.6688, 4.7328, 4.3990, 4.3979, 4.7821, 4.6509, 4.5213], device='cuda:5'), covar=tensor([0.0638, 0.0975, 0.0443, 0.0394, 0.1061, 0.0568, 0.0424, 0.0707], device='cuda:5'), in_proj_covar=tensor([0.0280, 0.0414, 0.0327, 0.0325, 0.0329, 0.0377, 0.0223, 0.0391], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 18:00:23,703 INFO [train.py:904] (5/8) Epoch 24, batch 50, loss[loss=0.1632, simple_loss=0.2572, pruned_loss=0.03461, over 17188.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2664, pruned_loss=0.04836, over 748301.09 frames. ], batch size: 46, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:00:52,611 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.410e+02 2.907e+02 3.501e+02 5.437e+02, threshold=5.814e+02, percent-clipped=4.0 2023-05-01 18:00:57,883 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9369, 2.0770, 2.4877, 2.8450, 2.6894, 3.3791, 2.0626, 3.4188], device='cuda:5'), covar=tensor([0.0302, 0.0596, 0.0406, 0.0349, 0.0436, 0.0239, 0.0629, 0.0192], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0190, 0.0176, 0.0180, 0.0195, 0.0152, 0.0194, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:01:32,994 INFO [train.py:904] (5/8) Epoch 24, batch 100, loss[loss=0.1448, simple_loss=0.2367, pruned_loss=0.02647, over 16804.00 frames. ], tot_loss[loss=0.177, simple_loss=0.263, pruned_loss=0.04548, over 1314751.18 frames. ], batch size: 39, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:01:43,916 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.55 vs. limit=5.0 2023-05-01 18:01:53,045 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 18:02:40,667 INFO [train.py:904] (5/8) Epoch 24, batch 150, loss[loss=0.1798, simple_loss=0.2703, pruned_loss=0.0447, over 16557.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2614, pruned_loss=0.04461, over 1755383.17 frames. ], batch size: 68, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:02:56,784 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:08,348 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.737e+02 2.160e+02 2.597e+02 3.268e+02 6.927e+02, threshold=5.194e+02, percent-clipped=3.0 2023-05-01 18:03:30,591 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:39,385 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233646.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:03:48,638 INFO [train.py:904] (5/8) Epoch 24, batch 200, loss[loss=0.1765, simple_loss=0.2585, pruned_loss=0.0473, over 15405.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2611, pruned_loss=0.0445, over 2103572.91 frames. ], batch size: 190, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:04:18,034 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:24,121 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6026, 3.6591, 2.2246, 3.8629, 2.9222, 3.7829, 2.3209, 2.9902], device='cuda:5'), covar=tensor([0.0273, 0.0370, 0.1668, 0.0380, 0.0787, 0.0881, 0.1537, 0.0711], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0172, 0.0191, 0.0161, 0.0173, 0.0212, 0.0201, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:04:45,847 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:54,403 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233700.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:04:58,756 INFO [train.py:904] (5/8) Epoch 24, batch 250, loss[loss=0.156, simple_loss=0.242, pruned_loss=0.03498, over 17183.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2591, pruned_loss=0.04309, over 2383154.59 frames. ], batch size: 46, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:05:07,845 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233710.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:05:25,968 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:05:26,868 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.284e+02 2.752e+02 3.292e+02 5.130e+02, threshold=5.503e+02, percent-clipped=0.0 2023-05-01 18:06:08,093 INFO [train.py:904] (5/8) Epoch 24, batch 300, loss[loss=0.1581, simple_loss=0.2586, pruned_loss=0.02879, over 17121.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2573, pruned_loss=0.04249, over 2595901.81 frames. ], batch size: 47, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:06:33,449 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233771.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:07:16,410 INFO [train.py:904] (5/8) Epoch 24, batch 350, loss[loss=0.153, simple_loss=0.2524, pruned_loss=0.02677, over 17133.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2547, pruned_loss=0.04086, over 2759636.81 frames. ], batch size: 49, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:07:43,748 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 2.126e+02 2.409e+02 2.975e+02 6.590e+02, threshold=4.818e+02, percent-clipped=2.0 2023-05-01 18:07:49,811 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7939, 3.9704, 3.9281, 3.0845, 3.5558, 3.9749, 3.7494, 1.9460], device='cuda:5'), covar=tensor([0.0563, 0.0142, 0.0108, 0.0431, 0.0196, 0.0220, 0.0175, 0.0746], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0086, 0.0085, 0.0135, 0.0098, 0.0110, 0.0095, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 18:07:56,900 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-01 18:08:25,454 INFO [train.py:904] (5/8) Epoch 24, batch 400, loss[loss=0.1682, simple_loss=0.2649, pruned_loss=0.03572, over 17240.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2531, pruned_loss=0.04031, over 2875059.19 frames. ], batch size: 52, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:32,984 INFO [train.py:904] (5/8) Epoch 24, batch 450, loss[loss=0.1507, simple_loss=0.2409, pruned_loss=0.0302, over 17244.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2525, pruned_loss=0.03959, over 2979562.34 frames. ], batch size: 45, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:50,109 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:09:59,641 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.087e+02 2.389e+02 2.800e+02 4.652e+02, threshold=4.779e+02, percent-clipped=0.0 2023-05-01 18:10:04,135 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3359, 2.6306, 2.2054, 2.4313, 2.9548, 2.7072, 2.9657, 3.0821], device='cuda:5'), covar=tensor([0.0263, 0.0485, 0.0574, 0.0527, 0.0319, 0.0431, 0.0340, 0.0330], device='cuda:5'), in_proj_covar=tensor([0.0217, 0.0238, 0.0229, 0.0230, 0.0239, 0.0238, 0.0235, 0.0233], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:10:39,004 INFO [train.py:904] (5/8) Epoch 24, batch 500, loss[loss=0.1757, simple_loss=0.2486, pruned_loss=0.05141, over 16859.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2515, pruned_loss=0.03907, over 3054017.23 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:10:53,522 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233963.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:11:30,633 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5486, 2.3714, 2.4606, 4.3607, 2.4059, 2.7633, 2.4810, 2.5201], device='cuda:5'), covar=tensor([0.1291, 0.3692, 0.3179, 0.0526, 0.4121, 0.2673, 0.3767, 0.3717], device='cuda:5'), in_proj_covar=tensor([0.0405, 0.0454, 0.0374, 0.0327, 0.0439, 0.0518, 0.0426, 0.0530], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:11:37,612 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233995.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:11:37,888 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0240, 4.5275, 3.2115, 2.3968, 2.7176, 2.6275, 4.9410, 3.6064], device='cuda:5'), covar=tensor([0.2697, 0.0497, 0.1732, 0.2993, 0.3029, 0.2198, 0.0288, 0.1561], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0268, 0.0306, 0.0316, 0.0295, 0.0264, 0.0295, 0.0339], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 18:11:50,840 INFO [train.py:904] (5/8) Epoch 24, batch 550, loss[loss=0.1751, simple_loss=0.258, pruned_loss=0.04608, over 16400.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2501, pruned_loss=0.03851, over 3114964.09 frames. ], batch size: 68, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:12:17,100 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.103e+02 2.436e+02 3.122e+02 5.008e+02, threshold=4.871e+02, percent-clipped=2.0 2023-05-01 18:12:20,427 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:12:57,942 INFO [train.py:904] (5/8) Epoch 24, batch 600, loss[loss=0.1718, simple_loss=0.2599, pruned_loss=0.04183, over 16726.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2494, pruned_loss=0.03882, over 3167019.35 frames. ], batch size: 62, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:13:17,459 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234066.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:13:23,010 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5280, 5.9197, 5.6756, 5.7304, 5.3286, 5.3793, 5.2992, 6.0662], device='cuda:5'), covar=tensor([0.1549, 0.1035, 0.1096, 0.0864, 0.0930, 0.0704, 0.1268, 0.0957], device='cuda:5'), in_proj_covar=tensor([0.0688, 0.0835, 0.0688, 0.0640, 0.0530, 0.0538, 0.0707, 0.0657], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:13:45,321 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234086.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:14:08,031 INFO [train.py:904] (5/8) Epoch 24, batch 650, loss[loss=0.1663, simple_loss=0.2462, pruned_loss=0.04316, over 16875.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.249, pruned_loss=0.03866, over 3207410.34 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:14:36,191 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.015e+02 2.288e+02 2.913e+02 5.815e+02, threshold=4.575e+02, percent-clipped=2.0 2023-05-01 18:15:16,431 INFO [train.py:904] (5/8) Epoch 24, batch 700, loss[loss=0.1692, simple_loss=0.2648, pruned_loss=0.03679, over 16694.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2482, pruned_loss=0.0385, over 3228754.28 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:15:24,304 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6846, 6.1020, 5.7688, 5.8381, 5.3145, 5.4076, 5.4160, 6.2086], device='cuda:5'), covar=tensor([0.1439, 0.1057, 0.1251, 0.0931, 0.1005, 0.0768, 0.1410, 0.0990], device='cuda:5'), in_proj_covar=tensor([0.0685, 0.0833, 0.0686, 0.0638, 0.0529, 0.0536, 0.0705, 0.0655], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:16:23,766 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:16:24,501 INFO [train.py:904] (5/8) Epoch 24, batch 750, loss[loss=0.1438, simple_loss=0.2292, pruned_loss=0.02925, over 16836.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2484, pruned_loss=0.03872, over 3246377.48 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:16:44,283 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 18:16:45,137 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4318, 3.5023, 3.7333, 2.5403, 3.3918, 3.8088, 3.5107, 2.1606], device='cuda:5'), covar=tensor([0.0535, 0.0155, 0.0062, 0.0436, 0.0136, 0.0106, 0.0118, 0.0501], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0086, 0.0087, 0.0135, 0.0099, 0.0111, 0.0096, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 18:16:52,426 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.040e+02 2.434e+02 2.884e+02 6.768e+02, threshold=4.867e+02, percent-clipped=4.0 2023-05-01 18:17:04,964 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234231.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:17:33,990 INFO [train.py:904] (5/8) Epoch 24, batch 800, loss[loss=0.1393, simple_loss=0.229, pruned_loss=0.02478, over 17205.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2484, pruned_loss=0.03868, over 3268640.28 frames. ], batch size: 45, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:17:48,337 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234263.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:17:59,667 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 18:18:29,130 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234292.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:18:33,013 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234295.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:18:43,651 INFO [train.py:904] (5/8) Epoch 24, batch 850, loss[loss=0.1826, simple_loss=0.2672, pruned_loss=0.04899, over 16535.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2482, pruned_loss=0.03851, over 3284062.84 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:19:11,886 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.010e+02 2.479e+02 2.918e+02 4.237e+02, threshold=4.958e+02, percent-clipped=0.0 2023-05-01 18:19:40,595 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234343.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:19:49,547 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2419, 4.1944, 4.1615, 3.8925, 3.9336, 4.2090, 3.8905, 4.0212], device='cuda:5'), covar=tensor([0.0631, 0.0787, 0.0309, 0.0292, 0.0692, 0.0470, 0.0784, 0.0593], device='cuda:5'), in_proj_covar=tensor([0.0300, 0.0444, 0.0351, 0.0350, 0.0353, 0.0407, 0.0238, 0.0420], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:19:52,509 INFO [train.py:904] (5/8) Epoch 24, batch 900, loss[loss=0.1922, simple_loss=0.2703, pruned_loss=0.05707, over 16881.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2472, pruned_loss=0.03782, over 3294725.88 frames. ], batch size: 96, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:20:11,510 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234366.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:20:33,664 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:21:02,827 INFO [train.py:904] (5/8) Epoch 24, batch 950, loss[loss=0.1561, simple_loss=0.2357, pruned_loss=0.03827, over 16075.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2477, pruned_loss=0.03773, over 3314626.30 frames. ], batch size: 35, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:21:17,303 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:21:30,309 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.144e+02 2.493e+02 3.109e+02 1.071e+03, threshold=4.986e+02, percent-clipped=5.0 2023-05-01 18:21:37,388 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7081, 1.9689, 2.2976, 2.5112, 2.6345, 2.5278, 1.9168, 2.7559], device='cuda:5'), covar=tensor([0.0185, 0.0470, 0.0334, 0.0292, 0.0325, 0.0315, 0.0530, 0.0184], device='cuda:5'), in_proj_covar=tensor([0.0192, 0.0195, 0.0181, 0.0186, 0.0200, 0.0158, 0.0198, 0.0155], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:21:38,411 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8414, 3.9401, 4.1387, 4.1294, 4.1569, 3.9259, 3.9604, 3.9169], device='cuda:5'), covar=tensor([0.0445, 0.0727, 0.0488, 0.0438, 0.0555, 0.0541, 0.0805, 0.0561], device='cuda:5'), in_proj_covar=tensor([0.0427, 0.0473, 0.0462, 0.0424, 0.0503, 0.0483, 0.0565, 0.0386], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 18:22:10,614 INFO [train.py:904] (5/8) Epoch 24, batch 1000, loss[loss=0.1464, simple_loss=0.2225, pruned_loss=0.0351, over 16506.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2465, pruned_loss=0.03759, over 3314615.44 frames. ], batch size: 146, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:22:50,709 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7615, 3.9397, 2.6628, 4.5077, 3.0715, 4.4208, 2.7125, 3.2150], device='cuda:5'), covar=tensor([0.0351, 0.0423, 0.1480, 0.0318, 0.0881, 0.0525, 0.1523, 0.0806], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0179, 0.0196, 0.0169, 0.0178, 0.0220, 0.0205, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:23:22,530 INFO [train.py:904] (5/8) Epoch 24, batch 1050, loss[loss=0.1697, simple_loss=0.2452, pruned_loss=0.04706, over 16861.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2465, pruned_loss=0.03792, over 3314395.91 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:23:28,421 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 18:23:50,906 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.145e+02 2.609e+02 3.101e+02 1.491e+03, threshold=5.219e+02, percent-clipped=5.0 2023-05-01 18:24:30,880 INFO [train.py:904] (5/8) Epoch 24, batch 1100, loss[loss=0.168, simple_loss=0.2588, pruned_loss=0.03864, over 16615.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2464, pruned_loss=0.03832, over 3317507.92 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:24:37,566 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234558.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:25:18,086 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:25:38,668 INFO [train.py:904] (5/8) Epoch 24, batch 1150, loss[loss=0.1778, simple_loss=0.2449, pruned_loss=0.05541, over 16899.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2465, pruned_loss=0.03809, over 3310367.38 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:25:53,425 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0384, 4.4344, 2.9859, 2.5159, 2.8085, 2.3730, 4.6457, 3.7514], device='cuda:5'), covar=tensor([0.3111, 0.0688, 0.2185, 0.3296, 0.3338, 0.2666, 0.0586, 0.1465], device='cuda:5'), in_proj_covar=tensor([0.0332, 0.0271, 0.0309, 0.0318, 0.0298, 0.0266, 0.0298, 0.0342], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 18:26:06,113 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.089e+02 2.435e+02 2.920e+02 5.927e+02, threshold=4.869e+02, percent-clipped=1.0 2023-05-01 18:26:33,638 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 18:26:46,789 INFO [train.py:904] (5/8) Epoch 24, batch 1200, loss[loss=0.1592, simple_loss=0.2455, pruned_loss=0.03649, over 16391.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2457, pruned_loss=0.03737, over 3313260.67 frames. ], batch size: 68, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:27:09,100 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:25,009 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:53,728 INFO [train.py:904] (5/8) Epoch 24, batch 1250, loss[loss=0.1681, simple_loss=0.2486, pruned_loss=0.04378, over 16699.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2457, pruned_loss=0.03731, over 3323751.68 frames. ], batch size: 62, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:28:21,125 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.044e+02 2.334e+02 2.766e+02 4.867e+02, threshold=4.669e+02, percent-clipped=0.0 2023-05-01 18:28:25,309 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:30,423 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234729.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:31,760 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234730.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:28:35,907 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6442, 3.7066, 2.6594, 2.2167, 2.2917, 2.1703, 3.7431, 3.0796], device='cuda:5'), covar=tensor([0.2988, 0.0625, 0.2152, 0.3206, 0.2963, 0.2601, 0.0642, 0.1766], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0271, 0.0309, 0.0318, 0.0298, 0.0266, 0.0298, 0.0342], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 18:28:37,176 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 18:29:02,403 INFO [train.py:904] (5/8) Epoch 24, batch 1300, loss[loss=0.1394, simple_loss=0.2274, pruned_loss=0.02572, over 16801.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2455, pruned_loss=0.03698, over 3321342.89 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:29:41,603 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:29:47,471 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:29:59,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0894, 3.8407, 4.2678, 2.3810, 4.4553, 4.5543, 3.2662, 3.5014], device='cuda:5'), covar=tensor([0.0713, 0.0310, 0.0259, 0.1149, 0.0099, 0.0189, 0.0475, 0.0423], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0140, 0.0083, 0.0128, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:30:10,010 INFO [train.py:904] (5/8) Epoch 24, batch 1350, loss[loss=0.148, simple_loss=0.2453, pruned_loss=0.02538, over 17107.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2463, pruned_loss=0.03717, over 3321496.30 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:30:38,937 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.157e+02 2.428e+02 2.986e+02 5.268e+02, threshold=4.855e+02, percent-clipped=4.0 2023-05-01 18:31:07,889 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234843.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:31:19,628 INFO [train.py:904] (5/8) Epoch 24, batch 1400, loss[loss=0.1804, simple_loss=0.2636, pruned_loss=0.04861, over 16847.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2465, pruned_loss=0.03703, over 3324432.65 frames. ], batch size: 102, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:31:27,921 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:31:37,972 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 18:31:51,468 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7547, 4.8772, 5.0424, 4.8388, 4.8683, 5.5046, 4.9925, 4.6221], device='cuda:5'), covar=tensor([0.1440, 0.2215, 0.2567, 0.2371, 0.2704, 0.1057, 0.1784, 0.2683], device='cuda:5'), in_proj_covar=tensor([0.0419, 0.0614, 0.0677, 0.0510, 0.0675, 0.0707, 0.0529, 0.0677], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 18:32:00,246 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:07,342 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234887.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:21,311 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 18:32:29,100 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 18:32:29,520 INFO [train.py:904] (5/8) Epoch 24, batch 1450, loss[loss=0.1661, simple_loss=0.2409, pruned_loss=0.04563, over 16723.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2452, pruned_loss=0.03666, over 3326581.98 frames. ], batch size: 124, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:32:34,695 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:58,967 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.989e+02 2.305e+02 2.603e+02 6.556e+02, threshold=4.610e+02, percent-clipped=2.0 2023-05-01 18:33:14,528 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:26,799 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234943.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:38,966 INFO [train.py:904] (5/8) Epoch 24, batch 1500, loss[loss=0.1634, simple_loss=0.2387, pruned_loss=0.04403, over 16827.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.245, pruned_loss=0.03739, over 3322344.21 frames. ], batch size: 102, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:33:45,668 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:49,655 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234960.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:34:21,538 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 18:34:49,734 INFO [train.py:904] (5/8) Epoch 24, batch 1550, loss[loss=0.1975, simple_loss=0.2763, pruned_loss=0.05938, over 11647.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2461, pruned_loss=0.03816, over 3311394.47 frames. ], batch size: 247, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:34:56,183 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-01 18:35:12,954 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235019.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:15,760 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:18,863 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 2.203e+02 2.509e+02 3.045e+02 6.004e+02, threshold=5.019e+02, percent-clipped=5.0 2023-05-01 18:35:20,357 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235025.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 18:35:34,161 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 18:35:58,117 INFO [train.py:904] (5/8) Epoch 24, batch 1600, loss[loss=0.1464, simple_loss=0.2304, pruned_loss=0.03119, over 16986.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2484, pruned_loss=0.03902, over 3305881.05 frames. ], batch size: 41, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:36:37,999 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:37:06,938 INFO [train.py:904] (5/8) Epoch 24, batch 1650, loss[loss=0.1732, simple_loss=0.2544, pruned_loss=0.04604, over 16534.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2509, pruned_loss=0.04006, over 3306852.05 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:37:35,281 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.234e+02 2.602e+02 3.440e+02 9.714e+02, threshold=5.204e+02, percent-clipped=4.0 2023-05-01 18:37:54,829 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235138.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:38:14,311 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8838, 2.6557, 2.5671, 4.0948, 3.3924, 4.0586, 1.6825, 2.9914], device='cuda:5'), covar=tensor([0.1288, 0.0689, 0.1095, 0.0171, 0.0117, 0.0350, 0.1544, 0.0756], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0177, 0.0197, 0.0196, 0.0205, 0.0218, 0.0206, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:38:16,053 INFO [train.py:904] (5/8) Epoch 24, batch 1700, loss[loss=0.1637, simple_loss=0.2619, pruned_loss=0.03272, over 16801.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2515, pruned_loss=0.04029, over 3314775.83 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:39:24,671 INFO [train.py:904] (5/8) Epoch 24, batch 1750, loss[loss=0.1755, simple_loss=0.2683, pruned_loss=0.04131, over 16550.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2524, pruned_loss=0.04048, over 3314211.10 frames. ], batch size: 62, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:39:52,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.256e+02 2.911e+02 3.628e+02 6.601e+02, threshold=5.823e+02, percent-clipped=7.0 2023-05-01 18:40:12,977 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235238.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:40:33,191 INFO [train.py:904] (5/8) Epoch 24, batch 1800, loss[loss=0.1943, simple_loss=0.2888, pruned_loss=0.04991, over 16359.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2543, pruned_loss=0.0405, over 3323390.83 frames. ], batch size: 146, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:40:36,385 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:04,922 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5815, 5.4908, 5.4241, 4.9540, 5.0580, 5.4767, 5.3863, 5.0414], device='cuda:5'), covar=tensor([0.0598, 0.0684, 0.0295, 0.0346, 0.1068, 0.0486, 0.0320, 0.0728], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0457, 0.0360, 0.0359, 0.0362, 0.0417, 0.0244, 0.0434], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:41:40,806 INFO [train.py:904] (5/8) Epoch 24, batch 1850, loss[loss=0.1607, simple_loss=0.24, pruned_loss=0.04072, over 15945.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2556, pruned_loss=0.04031, over 3319572.51 frames. ], batch size: 35, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:41:56,553 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,490 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,649 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:11,310 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.161e+02 2.587e+02 3.187e+02 4.962e+02, threshold=5.175e+02, percent-clipped=0.0 2023-05-01 18:42:12,328 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:45,053 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7836, 2.7178, 2.4987, 4.0917, 3.3639, 4.1363, 1.5442, 2.9158], device='cuda:5'), covar=tensor([0.1425, 0.0695, 0.1233, 0.0213, 0.0161, 0.0385, 0.1706, 0.0880], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0177, 0.0197, 0.0195, 0.0205, 0.0218, 0.0206, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:42:49,558 INFO [train.py:904] (5/8) Epoch 24, batch 1900, loss[loss=0.1659, simple_loss=0.2612, pruned_loss=0.03527, over 17074.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2552, pruned_loss=0.03958, over 3318667.02 frames. ], batch size: 53, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:43:01,040 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235360.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:43:17,591 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235373.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:43:31,579 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:00,962 INFO [train.py:904] (5/8) Epoch 24, batch 1950, loss[loss=0.1692, simple_loss=0.2601, pruned_loss=0.03913, over 15537.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2546, pruned_loss=0.03953, over 3316874.98 frames. ], batch size: 191, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:44:26,849 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235421.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:44:31,115 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.100e+02 2.476e+02 2.993e+02 5.203e+02, threshold=4.952e+02, percent-clipped=1.0 2023-05-01 18:44:37,918 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:49,020 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235438.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:52,701 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4216, 3.3720, 3.4518, 3.5179, 3.5708, 3.2857, 3.4713, 3.6429], device='cuda:5'), covar=tensor([0.1184, 0.0907, 0.0939, 0.0641, 0.0609, 0.2518, 0.1545, 0.0686], device='cuda:5'), in_proj_covar=tensor([0.0682, 0.0846, 0.0971, 0.0853, 0.0648, 0.0671, 0.0704, 0.0818], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:45:08,804 INFO [train.py:904] (5/8) Epoch 24, batch 2000, loss[loss=0.164, simple_loss=0.2608, pruned_loss=0.03365, over 17127.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2545, pruned_loss=0.03949, over 3315864.35 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:45:53,863 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235486.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:46:05,349 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-01 18:46:16,412 INFO [train.py:904] (5/8) Epoch 24, batch 2050, loss[loss=0.1663, simple_loss=0.2587, pruned_loss=0.03695, over 16608.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2539, pruned_loss=0.03962, over 3310245.12 frames. ], batch size: 68, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:46:46,285 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 2.111e+02 2.349e+02 3.134e+02 8.501e+02, threshold=4.699e+02, percent-clipped=2.0 2023-05-01 18:46:58,325 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2230, 3.2800, 2.0937, 3.4677, 2.6038, 3.4604, 2.1776, 2.6601], device='cuda:5'), covar=tensor([0.0325, 0.0428, 0.1562, 0.0355, 0.0811, 0.0701, 0.1486, 0.0794], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0182, 0.0198, 0.0173, 0.0180, 0.0224, 0.0207, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:47:03,585 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:47:24,558 INFO [train.py:904] (5/8) Epoch 24, batch 2100, loss[loss=0.1947, simple_loss=0.2618, pruned_loss=0.06376, over 16870.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2545, pruned_loss=0.04003, over 3307107.84 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:47:38,705 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9371, 3.1855, 3.1256, 2.0348, 2.7192, 2.2571, 3.4285, 3.5473], device='cuda:5'), covar=tensor([0.0292, 0.0982, 0.0721, 0.2155, 0.1029, 0.1125, 0.0630, 0.0980], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:48:10,074 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:31,077 INFO [train.py:904] (5/8) Epoch 24, batch 2150, loss[loss=0.2035, simple_loss=0.2846, pruned_loss=0.06118, over 15613.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2549, pruned_loss=0.04024, over 3317324.68 frames. ], batch size: 191, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:48:43,103 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:46,883 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:49,855 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:49:02,117 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.182e+02 2.677e+02 3.005e+02 4.991e+02, threshold=5.354e+02, percent-clipped=2.0 2023-05-01 18:49:31,535 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8808, 2.9006, 2.4570, 2.8503, 3.1777, 2.9420, 3.4430, 3.3904], device='cuda:5'), covar=tensor([0.0167, 0.0415, 0.0543, 0.0419, 0.0284, 0.0404, 0.0282, 0.0281], device='cuda:5'), in_proj_covar=tensor([0.0227, 0.0245, 0.0234, 0.0235, 0.0245, 0.0245, 0.0245, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:49:41,312 INFO [train.py:904] (5/8) Epoch 24, batch 2200, loss[loss=0.1656, simple_loss=0.2549, pruned_loss=0.03813, over 17036.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2552, pruned_loss=0.04065, over 3316459.15 frames. ], batch size: 53, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:49:54,417 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235662.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:49:57,750 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:18,232 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235679.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:50,802 INFO [train.py:904] (5/8) Epoch 24, batch 2250, loss[loss=0.157, simple_loss=0.2581, pruned_loss=0.02796, over 17111.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2554, pruned_loss=0.04086, over 3320388.68 frames. ], batch size: 47, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:51:08,557 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235716.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:51:20,161 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.445e+02 2.086e+02 2.517e+02 3.047e+02 4.896e+02, threshold=5.035e+02, percent-clipped=0.0 2023-05-01 18:51:41,285 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:51:57,981 INFO [train.py:904] (5/8) Epoch 24, batch 2300, loss[loss=0.1599, simple_loss=0.2477, pruned_loss=0.0361, over 16765.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2556, pruned_loss=0.04053, over 3328703.45 frames. ], batch size: 83, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:52:50,324 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235791.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:53:08,149 INFO [train.py:904] (5/8) Epoch 24, batch 2350, loss[loss=0.1611, simple_loss=0.2573, pruned_loss=0.0325, over 17227.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.256, pruned_loss=0.04092, over 3329928.94 frames. ], batch size: 45, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:53:13,833 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 18:53:37,791 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.284e+02 2.753e+02 3.332e+02 6.201e+02, threshold=5.507e+02, percent-clipped=2.0 2023-05-01 18:54:06,733 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 18:54:14,711 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:54:15,314 INFO [train.py:904] (5/8) Epoch 24, batch 2400, loss[loss=0.1639, simple_loss=0.2511, pruned_loss=0.0384, over 16862.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2569, pruned_loss=0.04138, over 3329641.70 frames. ], batch size: 42, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:22,154 INFO [train.py:904] (5/8) Epoch 24, batch 2450, loss[loss=0.1725, simple_loss=0.277, pruned_loss=0.03403, over 16700.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2584, pruned_loss=0.04157, over 3325065.46 frames. ], batch size: 57, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:32,552 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 18:55:33,272 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:55:51,881 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.168e+02 2.510e+02 2.858e+02 5.891e+02, threshold=5.019e+02, percent-clipped=1.0 2023-05-01 18:56:03,683 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 18:56:28,804 INFO [train.py:904] (5/8) Epoch 24, batch 2500, loss[loss=0.1821, simple_loss=0.2597, pruned_loss=0.05223, over 16775.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2581, pruned_loss=0.04129, over 3322450.04 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:56:37,274 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235959.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:57:05,205 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8845, 5.1480, 5.3278, 5.0424, 5.1046, 5.7587, 5.2158, 4.8694], device='cuda:5'), covar=tensor([0.1349, 0.2310, 0.2773, 0.2330, 0.2950, 0.1045, 0.1833, 0.2812], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0622, 0.0686, 0.0514, 0.0683, 0.0716, 0.0536, 0.0683], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 18:57:28,599 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3880, 2.3384, 2.3014, 4.1585, 2.2844, 2.7156, 2.3880, 2.4895], device='cuda:5'), covar=tensor([0.1382, 0.3839, 0.3380, 0.0594, 0.4296, 0.2904, 0.3842, 0.3829], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0461, 0.0378, 0.0334, 0.0442, 0.0529, 0.0432, 0.0539], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:57:38,966 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8593, 4.6346, 4.9785, 5.0966, 5.2996, 4.6531, 5.2847, 5.2977], device='cuda:5'), covar=tensor([0.2024, 0.1451, 0.1756, 0.0827, 0.0601, 0.1086, 0.0753, 0.0712], device='cuda:5'), in_proj_covar=tensor([0.0682, 0.0845, 0.0973, 0.0853, 0.0652, 0.0673, 0.0708, 0.0820], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 18:57:41,345 INFO [train.py:904] (5/8) Epoch 24, batch 2550, loss[loss=0.1718, simple_loss=0.2683, pruned_loss=0.03759, over 17071.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2576, pruned_loss=0.04155, over 3324918.71 frames. ], batch size: 53, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:57:47,931 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-01 18:57:59,296 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236016.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:58:01,905 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236018.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:58:10,534 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.266e+02 2.526e+02 3.100e+02 4.908e+02, threshold=5.053e+02, percent-clipped=0.0 2023-05-01 18:58:24,274 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236035.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:58:49,133 INFO [train.py:904] (5/8) Epoch 24, batch 2600, loss[loss=0.1728, simple_loss=0.2542, pruned_loss=0.04567, over 16313.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.258, pruned_loss=0.04161, over 3329519.65 frames. ], batch size: 165, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:59:03,971 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236064.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:59:25,487 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:59:34,845 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5770, 3.6809, 2.3241, 4.0976, 2.8988, 4.0247, 2.4980, 3.0681], device='cuda:5'), covar=tensor([0.0323, 0.0431, 0.1526, 0.0314, 0.0817, 0.0744, 0.1314, 0.0691], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0182, 0.0197, 0.0173, 0.0179, 0.0223, 0.0206, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 18:59:58,189 INFO [train.py:904] (5/8) Epoch 24, batch 2650, loss[loss=0.1752, simple_loss=0.2742, pruned_loss=0.03805, over 17048.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2583, pruned_loss=0.04099, over 3336557.03 frames. ], batch size: 50, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:00:01,956 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-01 19:00:12,848 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 19:00:27,606 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.167e+02 2.518e+02 3.194e+02 5.192e+02, threshold=5.036e+02, percent-clipped=2.0 2023-05-01 19:00:58,589 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:01:05,922 INFO [train.py:904] (5/8) Epoch 24, batch 2700, loss[loss=0.1781, simple_loss=0.271, pruned_loss=0.04257, over 16678.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2585, pruned_loss=0.04024, over 3333996.48 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:01:57,245 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7172, 3.8762, 2.6902, 4.5938, 3.1199, 4.5207, 2.7601, 3.3185], device='cuda:5'), covar=tensor([0.0356, 0.0433, 0.1427, 0.0343, 0.0793, 0.0541, 0.1351, 0.0710], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0182, 0.0198, 0.0174, 0.0180, 0.0224, 0.0206, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 19:02:15,470 INFO [train.py:904] (5/8) Epoch 24, batch 2750, loss[loss=0.1747, simple_loss=0.2721, pruned_loss=0.03868, over 16688.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2588, pruned_loss=0.03973, over 3343948.72 frames. ], batch size: 57, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:02:44,675 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.298e+02 1.972e+02 2.253e+02 2.676e+02 4.118e+02, threshold=4.507e+02, percent-clipped=0.0 2023-05-01 19:03:22,962 INFO [train.py:904] (5/8) Epoch 24, batch 2800, loss[loss=0.178, simple_loss=0.2567, pruned_loss=0.04968, over 16721.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.258, pruned_loss=0.03963, over 3333034.64 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:04:32,943 INFO [train.py:904] (5/8) Epoch 24, batch 2850, loss[loss=0.1557, simple_loss=0.2402, pruned_loss=0.0356, over 16953.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2569, pruned_loss=0.03986, over 3313763.97 frames. ], batch size: 41, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:04:41,881 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8429, 2.1532, 2.3782, 3.1336, 2.1677, 2.3576, 2.3312, 2.2901], device='cuda:5'), covar=tensor([0.1432, 0.3164, 0.2611, 0.0790, 0.3930, 0.2351, 0.3054, 0.3252], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0460, 0.0378, 0.0333, 0.0440, 0.0528, 0.0431, 0.0538], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:05:04,034 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.044e+02 2.504e+02 3.029e+02 5.589e+02, threshold=5.008e+02, percent-clipped=4.0 2023-05-01 19:05:17,539 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:05:43,140 INFO [train.py:904] (5/8) Epoch 24, batch 2900, loss[loss=0.1537, simple_loss=0.2333, pruned_loss=0.03707, over 16708.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2561, pruned_loss=0.04037, over 3312487.89 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:06:13,276 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:25,531 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236383.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:47,401 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1620, 4.1580, 4.1210, 3.5102, 4.1605, 1.6907, 3.9236, 3.6164], device='cuda:5'), covar=tensor([0.0170, 0.0138, 0.0214, 0.0349, 0.0118, 0.3087, 0.0156, 0.0284], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0166, 0.0208, 0.0183, 0.0183, 0.0213, 0.0196, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:06:53,580 INFO [train.py:904] (5/8) Epoch 24, batch 2950, loss[loss=0.1477, simple_loss=0.2364, pruned_loss=0.02955, over 17197.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2556, pruned_loss=0.0405, over 3322462.99 frames. ], batch size: 45, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:07:24,081 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.404e+02 2.771e+02 3.339e+02 1.037e+03, threshold=5.543e+02, percent-clipped=6.0 2023-05-01 19:07:54,696 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:08:02,723 INFO [train.py:904] (5/8) Epoch 24, batch 3000, loss[loss=0.1845, simple_loss=0.2705, pruned_loss=0.04926, over 16706.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2562, pruned_loss=0.04136, over 3314228.33 frames. ], batch size: 57, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:08:02,723 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 19:08:12,055 INFO [train.py:938] (5/8) Epoch 24, validation: loss=0.1342, simple_loss=0.2393, pruned_loss=0.0145, over 944034.00 frames. 2023-05-01 19:08:12,056 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-01 19:08:23,794 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7976, 4.8743, 5.2702, 5.2568, 5.2955, 4.9345, 4.9016, 4.7355], device='cuda:5'), covar=tensor([0.0353, 0.0554, 0.0405, 0.0444, 0.0516, 0.0372, 0.0911, 0.0480], device='cuda:5'), in_proj_covar=tensor([0.0430, 0.0478, 0.0466, 0.0428, 0.0508, 0.0484, 0.0570, 0.0388], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 19:09:12,138 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236495.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:09:22,990 INFO [train.py:904] (5/8) Epoch 24, batch 3050, loss[loss=0.175, simple_loss=0.2613, pruned_loss=0.04436, over 12269.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2553, pruned_loss=0.0409, over 3319055.42 frames. ], batch size: 246, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:09:23,499 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0958, 2.2836, 2.6954, 3.1848, 3.0154, 3.7079, 2.5248, 3.6507], device='cuda:5'), covar=tensor([0.0303, 0.0542, 0.0403, 0.0341, 0.0352, 0.0185, 0.0487, 0.0174], device='cuda:5'), in_proj_covar=tensor([0.0198, 0.0197, 0.0186, 0.0190, 0.0205, 0.0164, 0.0201, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:09:53,433 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.094e+02 2.423e+02 2.785e+02 4.383e+02, threshold=4.846e+02, percent-clipped=1.0 2023-05-01 19:10:32,494 INFO [train.py:904] (5/8) Epoch 24, batch 3100, loss[loss=0.16, simple_loss=0.2567, pruned_loss=0.03161, over 17016.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2553, pruned_loss=0.04106, over 3311614.82 frames. ], batch size: 55, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:10:51,097 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7933, 3.8729, 4.1304, 4.1111, 4.1349, 3.8864, 3.9240, 3.9005], device='cuda:5'), covar=tensor([0.0388, 0.0597, 0.0419, 0.0416, 0.0521, 0.0465, 0.0778, 0.0552], device='cuda:5'), in_proj_covar=tensor([0.0429, 0.0477, 0.0464, 0.0427, 0.0508, 0.0484, 0.0569, 0.0389], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 19:11:25,191 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7964, 2.4584, 2.4712, 3.7289, 3.0271, 3.8486, 1.5813, 2.8548], device='cuda:5'), covar=tensor([0.1359, 0.0734, 0.1158, 0.0213, 0.0168, 0.0370, 0.1587, 0.0830], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0176, 0.0195, 0.0195, 0.0205, 0.0216, 0.0204, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 19:11:43,897 INFO [train.py:904] (5/8) Epoch 24, batch 3150, loss[loss=0.1866, simple_loss=0.2593, pruned_loss=0.05699, over 16293.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2541, pruned_loss=0.04066, over 3308030.17 frames. ], batch size: 145, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:11:58,318 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 19:12:13,747 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.103e+02 2.556e+02 2.857e+02 5.533e+02, threshold=5.112e+02, percent-clipped=2.0 2023-05-01 19:12:52,327 INFO [train.py:904] (5/8) Epoch 24, batch 3200, loss[loss=0.1592, simple_loss=0.239, pruned_loss=0.0397, over 16902.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.254, pruned_loss=0.04021, over 3313909.95 frames. ], batch size: 116, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:13:06,074 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-01 19:13:22,047 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:13:43,340 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-05-01 19:14:01,578 INFO [train.py:904] (5/8) Epoch 24, batch 3250, loss[loss=0.2026, simple_loss=0.2806, pruned_loss=0.06227, over 12548.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2541, pruned_loss=0.04042, over 3311602.33 frames. ], batch size: 246, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:14:27,660 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:14:31,668 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.578e+02 2.232e+02 2.615e+02 3.014e+02 5.902e+02, threshold=5.231e+02, percent-clipped=1.0 2023-05-01 19:14:38,794 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 19:14:50,474 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 19:15:03,677 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3380, 5.2586, 5.1605, 4.6497, 4.7655, 5.2060, 5.1179, 4.7895], device='cuda:5'), covar=tensor([0.0567, 0.0432, 0.0324, 0.0362, 0.1168, 0.0434, 0.0352, 0.0828], device='cuda:5'), in_proj_covar=tensor([0.0315, 0.0468, 0.0368, 0.0367, 0.0372, 0.0426, 0.0251, 0.0445], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 19:15:11,552 INFO [train.py:904] (5/8) Epoch 24, batch 3300, loss[loss=0.1597, simple_loss=0.2539, pruned_loss=0.03277, over 17283.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.255, pruned_loss=0.04037, over 3321892.10 frames. ], batch size: 52, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:21,168 INFO [train.py:904] (5/8) Epoch 24, batch 3350, loss[loss=0.1452, simple_loss=0.2307, pruned_loss=0.02989, over 16816.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.255, pruned_loss=0.03967, over 3328617.13 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:51,755 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 2.070e+02 2.526e+02 2.944e+02 6.359e+02, threshold=5.052e+02, percent-clipped=3.0 2023-05-01 19:16:54,207 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 19:17:04,620 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8946, 4.8905, 4.7439, 4.1064, 4.8346, 1.8727, 4.5971, 4.4921], device='cuda:5'), covar=tensor([0.0133, 0.0102, 0.0220, 0.0414, 0.0112, 0.2969, 0.0150, 0.0234], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0168, 0.0210, 0.0186, 0.0186, 0.0216, 0.0199, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:17:33,222 INFO [train.py:904] (5/8) Epoch 24, batch 3400, loss[loss=0.1545, simple_loss=0.2451, pruned_loss=0.03189, over 16692.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2556, pruned_loss=0.03994, over 3328776.43 frames. ], batch size: 62, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:17:52,601 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 19:18:31,106 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236892.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:18:45,606 INFO [train.py:904] (5/8) Epoch 24, batch 3450, loss[loss=0.1737, simple_loss=0.2548, pruned_loss=0.04629, over 15547.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2549, pruned_loss=0.03973, over 3312592.24 frames. ], batch size: 190, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:18:47,378 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8692, 2.1503, 2.4979, 2.8439, 2.6719, 3.3637, 2.3675, 3.3223], device='cuda:5'), covar=tensor([0.0315, 0.0509, 0.0376, 0.0354, 0.0368, 0.0236, 0.0501, 0.0197], device='cuda:5'), in_proj_covar=tensor([0.0198, 0.0197, 0.0186, 0.0190, 0.0205, 0.0164, 0.0202, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:19:15,861 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.094e+02 2.355e+02 2.776e+02 4.395e+02, threshold=4.710e+02, percent-clipped=0.0 2023-05-01 19:19:17,673 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7504, 3.7445, 2.9402, 2.2940, 2.4572, 2.4184, 3.9129, 3.2921], device='cuda:5'), covar=tensor([0.2743, 0.0637, 0.1733, 0.3032, 0.2745, 0.2155, 0.0513, 0.1533], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0272, 0.0309, 0.0320, 0.0302, 0.0267, 0.0298, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 19:19:25,655 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 19:19:55,282 INFO [train.py:904] (5/8) Epoch 24, batch 3500, loss[loss=0.1759, simple_loss=0.2679, pruned_loss=0.04197, over 17044.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2537, pruned_loss=0.03925, over 3318573.91 frames. ], batch size: 53, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:56,858 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236953.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 19:20:13,400 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 19:21:06,722 INFO [train.py:904] (5/8) Epoch 24, batch 3550, loss[loss=0.1458, simple_loss=0.2326, pruned_loss=0.02957, over 16751.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2527, pruned_loss=0.03867, over 3319205.92 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:21:35,734 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.420e+02 1.933e+02 2.243e+02 2.593e+02 4.523e+02, threshold=4.485e+02, percent-clipped=0.0 2023-05-01 19:21:44,166 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 19:22:15,055 INFO [train.py:904] (5/8) Epoch 24, batch 3600, loss[loss=0.1351, simple_loss=0.2203, pruned_loss=0.02497, over 15944.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2513, pruned_loss=0.03846, over 3315006.46 frames. ], batch size: 35, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:22:38,228 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.64 vs. limit=5.0 2023-05-01 19:23:26,130 INFO [train.py:904] (5/8) Epoch 24, batch 3650, loss[loss=0.168, simple_loss=0.2432, pruned_loss=0.04642, over 16450.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2504, pruned_loss=0.03886, over 3313197.12 frames. ], batch size: 146, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:23:58,726 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.153e+02 2.502e+02 3.114e+02 9.969e+02, threshold=5.004e+02, percent-clipped=5.0 2023-05-01 19:24:39,867 INFO [train.py:904] (5/8) Epoch 24, batch 3700, loss[loss=0.1773, simple_loss=0.2532, pruned_loss=0.05067, over 16747.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2497, pruned_loss=0.04121, over 3295781.09 frames. ], batch size: 134, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:25:53,332 INFO [train.py:904] (5/8) Epoch 24, batch 3750, loss[loss=0.1764, simple_loss=0.25, pruned_loss=0.05139, over 16401.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2511, pruned_loss=0.04279, over 3275567.44 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:26:25,690 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.223e+02 2.609e+02 3.048e+02 5.921e+02, threshold=5.219e+02, percent-clipped=3.0 2023-05-01 19:26:56,643 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3739, 2.8378, 2.3949, 2.5254, 3.1114, 2.8237, 3.0760, 3.2809], device='cuda:5'), covar=tensor([0.0210, 0.0398, 0.0508, 0.0442, 0.0244, 0.0345, 0.0252, 0.0239], device='cuda:5'), in_proj_covar=tensor([0.0229, 0.0244, 0.0234, 0.0235, 0.0245, 0.0245, 0.0246, 0.0242], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:26:57,504 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237248.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:27:03,169 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 19:27:03,608 INFO [train.py:904] (5/8) Epoch 24, batch 3800, loss[loss=0.1884, simple_loss=0.2714, pruned_loss=0.05269, over 16899.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2517, pruned_loss=0.04406, over 3278040.40 frames. ], batch size: 109, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:27:06,183 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3380, 5.6618, 5.3948, 5.4577, 5.0665, 4.9969, 5.0577, 5.7960], device='cuda:5'), covar=tensor([0.1238, 0.0841, 0.1112, 0.0884, 0.0896, 0.0752, 0.1133, 0.0741], device='cuda:5'), in_proj_covar=tensor([0.0716, 0.0868, 0.0717, 0.0671, 0.0555, 0.0557, 0.0733, 0.0679], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:28:07,353 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 19:28:20,068 INFO [train.py:904] (5/8) Epoch 24, batch 3850, loss[loss=0.1627, simple_loss=0.2384, pruned_loss=0.04348, over 16875.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2514, pruned_loss=0.0445, over 3283256.73 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:28:52,885 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.317e+02 2.601e+02 3.052e+02 4.766e+02, threshold=5.202e+02, percent-clipped=0.0 2023-05-01 19:29:30,856 INFO [train.py:904] (5/8) Epoch 24, batch 3900, loss[loss=0.1887, simple_loss=0.2618, pruned_loss=0.05776, over 16367.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2516, pruned_loss=0.04515, over 3268572.79 frames. ], batch size: 35, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:29:51,872 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-01 19:29:59,738 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0057, 4.0357, 4.3106, 4.2798, 4.3158, 4.0674, 4.1135, 4.0520], device='cuda:5'), covar=tensor([0.0415, 0.0696, 0.0415, 0.0425, 0.0587, 0.0463, 0.0775, 0.0592], device='cuda:5'), in_proj_covar=tensor([0.0432, 0.0483, 0.0467, 0.0430, 0.0512, 0.0487, 0.0573, 0.0391], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 19:30:07,631 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1301, 2.7458, 2.1874, 2.4755, 3.0491, 2.7997, 3.0479, 3.1685], device='cuda:5'), covar=tensor([0.0216, 0.0385, 0.0575, 0.0462, 0.0258, 0.0353, 0.0241, 0.0276], device='cuda:5'), in_proj_covar=tensor([0.0228, 0.0244, 0.0233, 0.0235, 0.0245, 0.0244, 0.0246, 0.0242], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:30:42,814 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2363, 3.3347, 3.4435, 2.1961, 2.9844, 2.4410, 3.6737, 3.7216], device='cuda:5'), covar=tensor([0.0237, 0.0892, 0.0635, 0.1958, 0.0884, 0.0992, 0.0509, 0.0830], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0156, 0.0148, 0.0132, 0.0145, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-01 19:30:43,389 INFO [train.py:904] (5/8) Epoch 24, batch 3950, loss[loss=0.172, simple_loss=0.2452, pruned_loss=0.04944, over 16798.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2508, pruned_loss=0.0456, over 3266868.44 frames. ], batch size: 83, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:31:17,993 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.223e+02 2.645e+02 3.283e+02 6.262e+02, threshold=5.289e+02, percent-clipped=1.0 2023-05-01 19:31:57,063 INFO [train.py:904] (5/8) Epoch 24, batch 4000, loss[loss=0.1708, simple_loss=0.2517, pruned_loss=0.04499, over 16707.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2508, pruned_loss=0.04588, over 3274684.40 frames. ], batch size: 62, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:32:53,205 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237491.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:33:10,924 INFO [train.py:904] (5/8) Epoch 24, batch 4050, loss[loss=0.15, simple_loss=0.2451, pruned_loss=0.02747, over 16961.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2513, pruned_loss=0.04498, over 3273978.92 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:33:43,736 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 1.828e+02 2.122e+02 2.508e+02 6.067e+02, threshold=4.243e+02, percent-clipped=1.0 2023-05-01 19:34:17,672 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=237548.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:34:23,851 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237552.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:34:24,083 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-05-01 19:34:24,477 INFO [train.py:904] (5/8) Epoch 24, batch 4100, loss[loss=0.2023, simple_loss=0.2898, pruned_loss=0.05742, over 15400.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2531, pruned_loss=0.04438, over 3274896.51 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:34:34,721 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7625, 3.8646, 2.4606, 4.5777, 3.0668, 4.5220, 2.5968, 3.1895], device='cuda:5'), covar=tensor([0.0305, 0.0400, 0.1661, 0.0136, 0.0765, 0.0406, 0.1472, 0.0753], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0181, 0.0197, 0.0173, 0.0179, 0.0223, 0.0204, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 19:34:38,860 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237563.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:34:49,957 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7435, 1.8227, 1.5603, 1.5036, 1.9082, 1.6140, 1.6946, 1.9545], device='cuda:5'), covar=tensor([0.0183, 0.0325, 0.0435, 0.0353, 0.0214, 0.0267, 0.0185, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0227, 0.0243, 0.0233, 0.0235, 0.0244, 0.0244, 0.0246, 0.0242], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:35:30,113 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=237596.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:35:40,177 INFO [train.py:904] (5/8) Epoch 24, batch 4150, loss[loss=0.1808, simple_loss=0.2769, pruned_loss=0.04237, over 16746.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2602, pruned_loss=0.04672, over 3242503.03 frames. ], batch size: 39, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:36:14,079 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237624.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:36:16,368 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.139e+02 2.639e+02 3.206e+02 5.465e+02, threshold=5.279e+02, percent-clipped=4.0 2023-05-01 19:36:43,119 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 19:36:56,405 INFO [train.py:904] (5/8) Epoch 24, batch 4200, loss[loss=0.2091, simple_loss=0.2994, pruned_loss=0.05939, over 16655.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2673, pruned_loss=0.04822, over 3219979.41 frames. ], batch size: 57, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:38:10,765 INFO [train.py:904] (5/8) Epoch 24, batch 4250, loss[loss=0.176, simple_loss=0.2804, pruned_loss=0.03581, over 16783.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2705, pruned_loss=0.04804, over 3207344.55 frames. ], batch size: 83, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:38:16,756 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-05-01 19:38:45,325 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.496e+02 2.159e+02 2.541e+02 2.876e+02 4.427e+02, threshold=5.081e+02, percent-clipped=0.0 2023-05-01 19:38:48,876 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4177, 3.2933, 2.7015, 2.2221, 2.2437, 2.3658, 3.3782, 2.9630], device='cuda:5'), covar=tensor([0.3040, 0.0684, 0.1870, 0.2821, 0.2604, 0.2160, 0.0578, 0.1412], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0272, 0.0309, 0.0320, 0.0304, 0.0267, 0.0299, 0.0342], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 19:39:16,364 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 19:39:26,167 INFO [train.py:904] (5/8) Epoch 24, batch 4300, loss[loss=0.1917, simple_loss=0.2895, pruned_loss=0.04697, over 16779.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2718, pruned_loss=0.04721, over 3206927.60 frames. ], batch size: 124, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:40:01,162 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 19:40:23,402 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237790.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:40:41,745 INFO [train.py:904] (5/8) Epoch 24, batch 4350, loss[loss=0.2173, simple_loss=0.2886, pruned_loss=0.07302, over 11505.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2744, pruned_loss=0.048, over 3206956.54 frames. ], batch size: 247, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:41:14,127 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 19:41:14,404 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.331e+02 2.533e+02 3.016e+02 1.010e+03, threshold=5.065e+02, percent-clipped=1.0 2023-05-01 19:41:45,861 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:52,239 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237851.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:54,770 INFO [train.py:904] (5/8) Epoch 24, batch 4400, loss[loss=0.197, simple_loss=0.2832, pruned_loss=0.05544, over 17004.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2766, pruned_loss=0.04899, over 3207890.34 frames. ], batch size: 55, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:43:05,639 INFO [train.py:904] (5/8) Epoch 24, batch 4450, loss[loss=0.2036, simple_loss=0.2981, pruned_loss=0.05458, over 16317.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2804, pruned_loss=0.05066, over 3196020.88 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:43:18,872 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237912.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:43:28,654 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237919.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:43:38,900 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.917e+02 2.137e+02 2.680e+02 5.470e+02, threshold=4.273e+02, percent-clipped=1.0 2023-05-01 19:44:16,784 INFO [train.py:904] (5/8) Epoch 24, batch 4500, loss[loss=0.1985, simple_loss=0.2797, pruned_loss=0.05869, over 16899.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2808, pruned_loss=0.05155, over 3195202.87 frames. ], batch size: 116, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:44:37,779 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9970, 2.1743, 2.1641, 3.4744, 2.0978, 2.4680, 2.2805, 2.2855], device='cuda:5'), covar=tensor([0.1457, 0.3291, 0.3082, 0.0707, 0.4246, 0.2397, 0.3255, 0.3520], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0461, 0.0376, 0.0332, 0.0442, 0.0530, 0.0431, 0.0540], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:44:47,395 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237973.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:45:02,411 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2064, 2.0792, 1.7758, 1.8297, 2.2864, 1.9910, 2.0623, 2.4142], device='cuda:5'), covar=tensor([0.0191, 0.0359, 0.0478, 0.0414, 0.0223, 0.0332, 0.0184, 0.0237], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0239, 0.0229, 0.0231, 0.0240, 0.0240, 0.0241, 0.0238], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:45:03,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9064, 4.8927, 4.6547, 3.4764, 4.8413, 1.7063, 4.4703, 4.1477], device='cuda:5'), covar=tensor([0.0105, 0.0094, 0.0210, 0.0610, 0.0096, 0.3715, 0.0155, 0.0401], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0165, 0.0207, 0.0183, 0.0183, 0.0212, 0.0195, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:45:32,227 INFO [train.py:904] (5/8) Epoch 24, batch 4550, loss[loss=0.1973, simple_loss=0.2816, pruned_loss=0.05651, over 17119.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.282, pruned_loss=0.05275, over 3201348.14 frames. ], batch size: 47, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:46:04,577 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 1.751e+02 2.009e+02 2.366e+02 4.725e+02, threshold=4.018e+02, percent-clipped=1.0 2023-05-01 19:46:35,249 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2706, 3.4406, 3.6508, 2.0271, 3.1015, 2.2831, 3.6117, 3.6937], device='cuda:5'), covar=tensor([0.0227, 0.0810, 0.0540, 0.2224, 0.0854, 0.1045, 0.0549, 0.0858], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0167, 0.0168, 0.0154, 0.0146, 0.0131, 0.0143, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 19:46:44,399 INFO [train.py:904] (5/8) Epoch 24, batch 4600, loss[loss=0.1922, simple_loss=0.2883, pruned_loss=0.04807, over 17029.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2825, pruned_loss=0.05265, over 3211530.05 frames. ], batch size: 50, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:47:00,815 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:15,275 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238074.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:28,243 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6889, 3.6886, 2.2867, 4.3833, 2.8694, 4.2593, 2.5877, 2.9545], device='cuda:5'), covar=tensor([0.0284, 0.0418, 0.1795, 0.0178, 0.0860, 0.0485, 0.1475, 0.0855], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0179, 0.0195, 0.0169, 0.0177, 0.0220, 0.0203, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 19:47:35,644 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2729, 5.9042, 6.1451, 5.6997, 5.8390, 6.3842, 5.8856, 5.5872], device='cuda:5'), covar=tensor([0.0851, 0.1733, 0.1995, 0.2081, 0.2515, 0.0896, 0.1404, 0.2322], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0608, 0.0665, 0.0503, 0.0665, 0.0697, 0.0523, 0.0663], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 19:47:56,661 INFO [train.py:904] (5/8) Epoch 24, batch 4650, loss[loss=0.176, simple_loss=0.2636, pruned_loss=0.04416, over 16695.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2813, pruned_loss=0.05267, over 3226916.55 frames. ], batch size: 89, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:48:25,506 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-05-01 19:48:29,647 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:30,389 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 1.898e+02 2.163e+02 2.416e+02 3.900e+02, threshold=4.325e+02, percent-clipped=0.0 2023-05-01 19:48:30,827 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238126.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:43,756 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:59,284 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:49:00,470 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:49:09,299 INFO [train.py:904] (5/8) Epoch 24, batch 4700, loss[loss=0.1678, simple_loss=0.258, pruned_loss=0.03878, over 16702.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2786, pruned_loss=0.05132, over 3229827.12 frames. ], batch size: 83, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:49:58,942 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:06,591 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 19:50:09,169 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238195.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:20,676 INFO [train.py:904] (5/8) Epoch 24, batch 4750, loss[loss=0.1615, simple_loss=0.2453, pruned_loss=0.03885, over 16771.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2752, pruned_loss=0.0496, over 3215618.70 frames. ], batch size: 39, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:50:41,776 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-05-01 19:50:43,598 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238219.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 19:50:53,745 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 1.823e+02 2.174e+02 2.542e+02 5.265e+02, threshold=4.348e+02, percent-clipped=2.0 2023-05-01 19:51:31,489 INFO [train.py:904] (5/8) Epoch 24, batch 4800, loss[loss=0.1703, simple_loss=0.2607, pruned_loss=0.03992, over 16880.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2716, pruned_loss=0.04773, over 3206932.23 frames. ], batch size: 116, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:51:49,256 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4870, 3.5361, 2.1676, 4.0480, 2.6422, 3.9657, 2.3784, 2.7832], device='cuda:5'), covar=tensor([0.0313, 0.0389, 0.1741, 0.0172, 0.0971, 0.0571, 0.1583, 0.0883], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0169, 0.0177, 0.0220, 0.0203, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 19:51:52,869 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238267.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:51:54,668 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238268.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:52:25,775 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1331, 2.2737, 2.2736, 3.8531, 2.2218, 2.5999, 2.3790, 2.4085], device='cuda:5'), covar=tensor([0.1460, 0.3637, 0.2994, 0.0568, 0.4001, 0.2522, 0.3516, 0.3299], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0460, 0.0375, 0.0331, 0.0440, 0.0528, 0.0429, 0.0537], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:52:46,486 INFO [train.py:904] (5/8) Epoch 24, batch 4850, loss[loss=0.1666, simple_loss=0.2623, pruned_loss=0.03547, over 16554.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2734, pruned_loss=0.04763, over 3187720.27 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:52:53,318 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7274, 3.8852, 4.0652, 4.0273, 4.0478, 3.8669, 3.7340, 3.8484], device='cuda:5'), covar=tensor([0.0495, 0.0655, 0.0510, 0.0560, 0.0611, 0.0496, 0.1209, 0.0528], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0460, 0.0449, 0.0413, 0.0495, 0.0468, 0.0552, 0.0376], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 19:53:22,734 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.326e+02 1.927e+02 2.251e+02 2.681e+02 4.052e+02, threshold=4.502e+02, percent-clipped=0.0 2023-05-01 19:54:04,052 INFO [train.py:904] (5/8) Epoch 24, batch 4900, loss[loss=0.166, simple_loss=0.2644, pruned_loss=0.03375, over 15379.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2725, pruned_loss=0.0465, over 3175815.69 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:55:17,232 INFO [train.py:904] (5/8) Epoch 24, batch 4950, loss[loss=0.1895, simple_loss=0.2832, pruned_loss=0.04792, over 16439.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2719, pruned_loss=0.04598, over 3187576.57 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:55:41,599 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:55:41,760 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238420.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:55:49,724 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.002e+02 2.407e+02 2.827e+02 4.284e+02, threshold=4.815e+02, percent-clipped=0.0 2023-05-01 19:55:57,442 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:55:59,992 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2738, 5.2533, 5.1221, 4.1609, 5.2265, 1.6694, 4.9033, 4.8303], device='cuda:5'), covar=tensor([0.0108, 0.0092, 0.0193, 0.0579, 0.0113, 0.3081, 0.0140, 0.0279], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0163, 0.0204, 0.0181, 0.0180, 0.0209, 0.0192, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:56:20,278 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238446.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:56:29,586 INFO [train.py:904] (5/8) Epoch 24, batch 5000, loss[loss=0.1886, simple_loss=0.2821, pruned_loss=0.04758, over 15351.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2738, pruned_loss=0.04582, over 3201671.31 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:56:35,171 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8246, 3.7341, 3.8862, 3.9864, 4.0836, 3.7182, 4.0467, 4.1291], device='cuda:5'), covar=tensor([0.1432, 0.1130, 0.1218, 0.0655, 0.0510, 0.1796, 0.0779, 0.0632], device='cuda:5'), in_proj_covar=tensor([0.0644, 0.0797, 0.0918, 0.0806, 0.0615, 0.0636, 0.0665, 0.0775], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 19:57:11,145 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238481.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 19:57:12,062 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238482.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:31,131 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:42,730 INFO [train.py:904] (5/8) Epoch 24, batch 5050, loss[loss=0.1742, simple_loss=0.2728, pruned_loss=0.03775, over 16839.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2747, pruned_loss=0.04604, over 3180905.80 frames. ], batch size: 83, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:58:15,455 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 2.006e+02 2.316e+02 2.692e+02 7.292e+02, threshold=4.632e+02, percent-clipped=1.0 2023-05-01 19:58:53,438 INFO [train.py:904] (5/8) Epoch 24, batch 5100, loss[loss=0.1654, simple_loss=0.2563, pruned_loss=0.03728, over 16516.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2738, pruned_loss=0.04582, over 3169927.51 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:59:16,759 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:00:07,254 INFO [train.py:904] (5/8) Epoch 24, batch 5150, loss[loss=0.1527, simple_loss=0.2542, pruned_loss=0.02558, over 16773.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2737, pruned_loss=0.04533, over 3149508.77 frames. ], batch size: 89, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:00:26,119 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:00:36,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6641, 2.5736, 2.5190, 3.9641, 2.7348, 3.8780, 1.5561, 2.9426], device='cuda:5'), covar=tensor([0.1405, 0.0802, 0.1183, 0.0131, 0.0156, 0.0346, 0.1714, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0193, 0.0204, 0.0215, 0.0204, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:00:38,788 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.999e+02 2.272e+02 2.615e+02 4.924e+02, threshold=4.544e+02, percent-clipped=1.0 2023-05-01 20:01:17,686 INFO [train.py:904] (5/8) Epoch 24, batch 5200, loss[loss=0.1681, simple_loss=0.2596, pruned_loss=0.03828, over 16561.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2712, pruned_loss=0.04429, over 3162449.77 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:01:19,711 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 20:01:57,529 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:02:18,787 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 20:02:19,350 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3381, 3.2238, 3.6794, 1.7548, 3.7738, 3.7729, 2.9280, 2.7584], device='cuda:5'), covar=tensor([0.0798, 0.0267, 0.0137, 0.1265, 0.0072, 0.0141, 0.0391, 0.0498], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0109, 0.0100, 0.0138, 0.0082, 0.0126, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:02:28,276 INFO [train.py:904] (5/8) Epoch 24, batch 5250, loss[loss=0.1756, simple_loss=0.2674, pruned_loss=0.04189, over 16549.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2683, pruned_loss=0.04352, over 3178223.16 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:02:52,127 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238720.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:01,285 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 1.983e+02 2.265e+02 2.655e+02 9.176e+02, threshold=4.530e+02, percent-clipped=2.0 2023-05-01 20:03:07,622 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238730.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:25,217 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:39,628 INFO [train.py:904] (5/8) Epoch 24, batch 5300, loss[loss=0.1542, simple_loss=0.2429, pruned_loss=0.03271, over 16818.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2643, pruned_loss=0.04209, over 3190262.85 frames. ], batch size: 102, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:04:01,108 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:12,369 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238776.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:04:15,113 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238778.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:20,308 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:43,202 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0259, 2.6819, 2.8362, 2.1009, 2.6393, 2.1741, 2.7222, 2.9039], device='cuda:5'), covar=tensor([0.0331, 0.0827, 0.0538, 0.1843, 0.0893, 0.0885, 0.0715, 0.0771], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0166, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:04:50,566 INFO [train.py:904] (5/8) Epoch 24, batch 5350, loss[loss=0.1864, simple_loss=0.2728, pruned_loss=0.05007, over 11832.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.263, pruned_loss=0.04185, over 3166468.69 frames. ], batch size: 246, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:05:21,928 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 1.926e+02 2.251e+02 2.610e+02 1.024e+03, threshold=4.501e+02, percent-clipped=4.0 2023-05-01 20:05:27,689 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238830.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:06:01,799 INFO [train.py:904] (5/8) Epoch 24, batch 5400, loss[loss=0.1683, simple_loss=0.2704, pruned_loss=0.03309, over 16820.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.266, pruned_loss=0.04263, over 3158354.99 frames. ], batch size: 102, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:04,636 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8298, 4.0849, 3.0799, 2.5161, 2.7882, 2.7113, 4.5011, 3.6784], device='cuda:5'), covar=tensor([0.2695, 0.0574, 0.1722, 0.2452, 0.2570, 0.1846, 0.0378, 0.1147], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0271, 0.0307, 0.0318, 0.0300, 0.0265, 0.0298, 0.0340], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 20:07:11,360 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7877, 4.6548, 4.8732, 5.0152, 5.1952, 4.6600, 5.2297, 5.1880], device='cuda:5'), covar=tensor([0.1901, 0.1294, 0.1674, 0.0753, 0.0587, 0.0972, 0.0574, 0.0741], device='cuda:5'), in_proj_covar=tensor([0.0649, 0.0799, 0.0921, 0.0810, 0.0617, 0.0639, 0.0667, 0.0779], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:07:18,015 INFO [train.py:904] (5/8) Epoch 24, batch 5450, loss[loss=0.2241, simple_loss=0.2975, pruned_loss=0.07534, over 11768.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2688, pruned_loss=0.0442, over 3154811.34 frames. ], batch size: 247, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:54,290 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.137e+02 2.656e+02 3.537e+02 7.468e+02, threshold=5.312e+02, percent-clipped=13.0 2023-05-01 20:08:11,064 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238936.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:08:18,335 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5314, 3.5975, 3.3712, 3.0103, 3.2100, 3.4896, 3.3194, 3.3029], device='cuda:5'), covar=tensor([0.0612, 0.0753, 0.0290, 0.0279, 0.0497, 0.0492, 0.1536, 0.0515], device='cuda:5'), in_proj_covar=tensor([0.0298, 0.0448, 0.0351, 0.0351, 0.0354, 0.0408, 0.0238, 0.0423], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:08:28,479 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1528, 3.2281, 1.9816, 3.4655, 2.4529, 3.5025, 2.1896, 2.6620], device='cuda:5'), covar=tensor([0.0306, 0.0369, 0.1530, 0.0191, 0.0774, 0.0546, 0.1354, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0177, 0.0193, 0.0166, 0.0176, 0.0216, 0.0201, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:08:35,280 INFO [train.py:904] (5/8) Epoch 24, batch 5500, loss[loss=0.2006, simple_loss=0.2885, pruned_loss=0.05632, over 16900.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2759, pruned_loss=0.04808, over 3145637.54 frames. ], batch size: 109, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:09:44,167 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238997.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:09:53,120 INFO [train.py:904] (5/8) Epoch 24, batch 5550, loss[loss=0.2192, simple_loss=0.3089, pruned_loss=0.06478, over 15242.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.283, pruned_loss=0.05274, over 3134208.83 frames. ], batch size: 190, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:10:06,736 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8935, 2.1926, 2.4797, 3.1024, 2.2131, 2.3811, 2.3587, 2.3195], device='cuda:5'), covar=tensor([0.1363, 0.3080, 0.2194, 0.0711, 0.3801, 0.2310, 0.3081, 0.2962], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0456, 0.0373, 0.0329, 0.0436, 0.0523, 0.0427, 0.0532], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:10:27,307 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5142, 3.0569, 3.1238, 2.0111, 2.8275, 2.1422, 3.1837, 3.3022], device='cuda:5'), covar=tensor([0.0315, 0.0749, 0.0603, 0.2018, 0.0853, 0.1047, 0.0659, 0.0932], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0165, 0.0167, 0.0154, 0.0146, 0.0130, 0.0143, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:10:30,507 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 3.032e+02 3.628e+02 4.510e+02 7.943e+02, threshold=7.255e+02, percent-clipped=11.0 2023-05-01 20:10:48,344 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:11:12,519 INFO [train.py:904] (5/8) Epoch 24, batch 5600, loss[loss=0.1936, simple_loss=0.2827, pruned_loss=0.0522, over 16850.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2879, pruned_loss=0.05721, over 3100765.30 frames. ], batch size: 90, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:11:52,280 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239076.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 20:12:26,576 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 20:12:36,921 INFO [train.py:904] (5/8) Epoch 24, batch 5650, loss[loss=0.2005, simple_loss=0.2874, pruned_loss=0.05683, over 16572.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2917, pruned_loss=0.06016, over 3101996.87 frames. ], batch size: 68, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:12:54,674 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-01 20:13:11,119 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239124.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:13:14,764 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.394e+02 3.339e+02 4.323e+02 5.614e+02 1.229e+03, threshold=8.646e+02, percent-clipped=9.0 2023-05-01 20:13:47,401 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9327, 5.4954, 5.6489, 5.3614, 5.4782, 5.9734, 5.4679, 5.2459], device='cuda:5'), covar=tensor([0.0927, 0.1662, 0.2025, 0.1832, 0.2085, 0.0825, 0.1485, 0.2303], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0602, 0.0661, 0.0496, 0.0660, 0.0692, 0.0519, 0.0659], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 20:13:55,701 INFO [train.py:904] (5/8) Epoch 24, batch 5700, loss[loss=0.1849, simple_loss=0.2876, pruned_loss=0.04115, over 16687.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2936, pruned_loss=0.06163, over 3083253.12 frames. ], batch size: 89, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:15:14,253 INFO [train.py:904] (5/8) Epoch 24, batch 5750, loss[loss=0.2079, simple_loss=0.2964, pruned_loss=0.05967, over 17000.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.296, pruned_loss=0.06313, over 3072949.37 frames. ], batch size: 53, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:15:53,492 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.857e+02 3.000e+02 3.580e+02 4.341e+02 9.766e+02, threshold=7.159e+02, percent-clipped=1.0 2023-05-01 20:16:37,345 INFO [train.py:904] (5/8) Epoch 24, batch 5800, loss[loss=0.1867, simple_loss=0.2837, pruned_loss=0.04483, over 15317.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2958, pruned_loss=0.06198, over 3057875.41 frames. ], batch size: 190, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:17:40,262 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239292.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 20:17:56,804 INFO [train.py:904] (5/8) Epoch 24, batch 5850, loss[loss=0.2085, simple_loss=0.2928, pruned_loss=0.06207, over 16650.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.293, pruned_loss=0.05998, over 3051331.22 frames. ], batch size: 76, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:18:33,903 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.891e+02 3.438e+02 4.490e+02 7.448e+02, threshold=6.875e+02, percent-clipped=1.0 2023-05-01 20:18:52,958 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239337.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:19:19,097 INFO [train.py:904] (5/8) Epoch 24, batch 5900, loss[loss=0.1788, simple_loss=0.2795, pruned_loss=0.03906, over 16717.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2925, pruned_loss=0.05954, over 3064655.35 frames. ], batch size: 89, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:19:50,748 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-01 20:20:10,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0400, 4.0026, 3.9781, 3.1370, 3.9851, 1.7492, 3.7792, 3.4638], device='cuda:5'), covar=tensor([0.0187, 0.0142, 0.0218, 0.0363, 0.0126, 0.3137, 0.0186, 0.0334], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0164, 0.0205, 0.0182, 0.0180, 0.0210, 0.0192, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:20:14,709 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239385.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:20:29,370 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5923, 2.7734, 2.2945, 2.5657, 3.1435, 2.7844, 3.1820, 3.3291], device='cuda:5'), covar=tensor([0.0137, 0.0406, 0.0568, 0.0446, 0.0247, 0.0378, 0.0279, 0.0251], device='cuda:5'), in_proj_covar=tensor([0.0219, 0.0237, 0.0230, 0.0230, 0.0239, 0.0237, 0.0239, 0.0236], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:20:42,789 INFO [train.py:904] (5/8) Epoch 24, batch 5950, loss[loss=0.1833, simple_loss=0.2752, pruned_loss=0.04574, over 16544.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2931, pruned_loss=0.0585, over 3069904.94 frames. ], batch size: 68, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:21:03,328 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6154, 2.4891, 1.8704, 2.6542, 2.0999, 2.7543, 2.1896, 2.3671], device='cuda:5'), covar=tensor([0.0323, 0.0391, 0.1276, 0.0275, 0.0688, 0.0469, 0.1126, 0.0599], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0168, 0.0178, 0.0218, 0.0204, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:21:12,974 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4420, 3.9094, 4.0108, 2.5850, 3.5803, 3.9947, 3.5189, 2.3558], device='cuda:5'), covar=tensor([0.0568, 0.0063, 0.0056, 0.0447, 0.0126, 0.0126, 0.0117, 0.0469], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0087, 0.0086, 0.0134, 0.0099, 0.0111, 0.0097, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 20:21:21,253 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.477e+02 3.014e+02 3.724e+02 7.833e+02, threshold=6.029e+02, percent-clipped=2.0 2023-05-01 20:22:03,341 INFO [train.py:904] (5/8) Epoch 24, batch 6000, loss[loss=0.1855, simple_loss=0.2724, pruned_loss=0.04927, over 15472.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2922, pruned_loss=0.05818, over 3080810.53 frames. ], batch size: 191, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:22:03,341 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 20:22:14,270 INFO [train.py:938] (5/8) Epoch 24, validation: loss=0.1493, simple_loss=0.2618, pruned_loss=0.01837, over 944034.00 frames. 2023-05-01 20:22:14,271 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-01 20:23:32,279 INFO [train.py:904] (5/8) Epoch 24, batch 6050, loss[loss=0.2066, simple_loss=0.2902, pruned_loss=0.06151, over 15373.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2901, pruned_loss=0.05729, over 3104183.66 frames. ], batch size: 191, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:24:09,878 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.713e+02 3.363e+02 4.065e+02 7.660e+02, threshold=6.725e+02, percent-clipped=3.0 2023-05-01 20:24:51,435 INFO [train.py:904] (5/8) Epoch 24, batch 6100, loss[loss=0.1699, simple_loss=0.2613, pruned_loss=0.03926, over 17245.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2892, pruned_loss=0.05616, over 3106846.22 frames. ], batch size: 45, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:25:56,278 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239592.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:26:14,113 INFO [train.py:904] (5/8) Epoch 24, batch 6150, loss[loss=0.2241, simple_loss=0.2971, pruned_loss=0.07552, over 11458.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2877, pruned_loss=0.05592, over 3104043.08 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:26:24,846 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2822, 1.7035, 2.0386, 2.2837, 2.3822, 2.5945, 1.8014, 2.4766], device='cuda:5'), covar=tensor([0.0245, 0.0519, 0.0308, 0.0382, 0.0331, 0.0219, 0.0541, 0.0169], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0194, 0.0182, 0.0185, 0.0201, 0.0160, 0.0199, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:26:41,470 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8140, 3.7889, 3.9262, 3.6783, 3.8932, 4.2455, 3.9253, 3.5938], device='cuda:5'), covar=tensor([0.2020, 0.2374, 0.2384, 0.2514, 0.2466, 0.1763, 0.1629, 0.2591], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0607, 0.0666, 0.0499, 0.0664, 0.0695, 0.0521, 0.0664], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 20:26:45,914 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=239622.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:26:53,241 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.890e+02 2.573e+02 3.097e+02 3.625e+02 7.083e+02, threshold=6.193e+02, percent-clipped=1.0 2023-05-01 20:27:14,514 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239640.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:27:34,967 INFO [train.py:904] (5/8) Epoch 24, batch 6200, loss[loss=0.2013, simple_loss=0.2822, pruned_loss=0.06019, over 16931.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2863, pruned_loss=0.05565, over 3103842.67 frames. ], batch size: 116, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:28:13,187 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 20:28:22,535 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=239683.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:28:52,043 INFO [train.py:904] (5/8) Epoch 24, batch 6250, loss[loss=0.1926, simple_loss=0.2835, pruned_loss=0.05079, over 16622.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2857, pruned_loss=0.05542, over 3101844.30 frames. ], batch size: 57, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:29:29,988 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 2.614e+02 3.057e+02 3.659e+02 7.500e+02, threshold=6.114e+02, percent-clipped=2.0 2023-05-01 20:29:43,270 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4942, 3.3497, 3.7461, 1.8494, 3.9115, 3.9306, 3.0226, 2.8522], device='cuda:5'), covar=tensor([0.0850, 0.0283, 0.0225, 0.1271, 0.0083, 0.0187, 0.0441, 0.0510], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0109, 0.0100, 0.0139, 0.0083, 0.0129, 0.0129, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:30:05,932 INFO [train.py:904] (5/8) Epoch 24, batch 6300, loss[loss=0.1979, simple_loss=0.2748, pruned_loss=0.06054, over 17067.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2858, pruned_loss=0.05493, over 3139557.52 frames. ], batch size: 53, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:30:20,210 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9431, 2.0715, 2.5317, 2.9111, 2.7349, 3.3137, 2.1318, 3.3243], device='cuda:5'), covar=tensor([0.0212, 0.0527, 0.0363, 0.0347, 0.0348, 0.0183, 0.0608, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0194, 0.0182, 0.0185, 0.0202, 0.0160, 0.0200, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:30:47,823 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9311, 3.2878, 3.4195, 2.1406, 2.9515, 2.2851, 3.4510, 3.5715], device='cuda:5'), covar=tensor([0.0261, 0.0810, 0.0650, 0.2160, 0.0931, 0.1086, 0.0634, 0.0954], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0166, 0.0169, 0.0155, 0.0147, 0.0132, 0.0144, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:31:13,318 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 20:31:24,193 INFO [train.py:904] (5/8) Epoch 24, batch 6350, loss[loss=0.1822, simple_loss=0.2663, pruned_loss=0.04902, over 16616.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2865, pruned_loss=0.05614, over 3122037.77 frames. ], batch size: 57, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:31:25,580 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 20:32:03,928 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.790e+02 3.396e+02 4.270e+02 1.026e+03, threshold=6.791e+02, percent-clipped=1.0 2023-05-01 20:32:15,597 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1656, 5.1150, 4.9907, 4.3012, 5.0625, 1.8208, 4.8054, 4.6478], device='cuda:5'), covar=tensor([0.0079, 0.0074, 0.0165, 0.0392, 0.0082, 0.2911, 0.0111, 0.0238], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0165, 0.0206, 0.0182, 0.0181, 0.0212, 0.0193, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:32:17,775 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2712, 2.2846, 2.1891, 3.8878, 2.2263, 2.6548, 2.3330, 2.4271], device='cuda:5'), covar=tensor([0.1260, 0.3294, 0.3049, 0.0537, 0.4056, 0.2250, 0.3433, 0.3167], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0456, 0.0373, 0.0329, 0.0437, 0.0522, 0.0427, 0.0533], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:32:42,138 INFO [train.py:904] (5/8) Epoch 24, batch 6400, loss[loss=0.2601, simple_loss=0.3281, pruned_loss=0.096, over 11547.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2875, pruned_loss=0.05788, over 3083651.24 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:32:44,415 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 20:33:58,193 INFO [train.py:904] (5/8) Epoch 24, batch 6450, loss[loss=0.1934, simple_loss=0.2633, pruned_loss=0.06179, over 11810.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2876, pruned_loss=0.05762, over 3074679.22 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:34:37,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.672e+02 3.087e+02 3.802e+02 7.341e+02, threshold=6.174e+02, percent-clipped=2.0 2023-05-01 20:35:16,111 INFO [train.py:904] (5/8) Epoch 24, batch 6500, loss[loss=0.1811, simple_loss=0.2704, pruned_loss=0.04588, over 16403.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2847, pruned_loss=0.05637, over 3080888.12 frames. ], batch size: 35, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:35:55,173 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:36:00,732 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6112, 4.3469, 4.3186, 2.8803, 3.9738, 4.3921, 3.9143, 2.3436], device='cuda:5'), covar=tensor([0.0570, 0.0049, 0.0047, 0.0402, 0.0083, 0.0096, 0.0085, 0.0483], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0087, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 20:36:39,182 INFO [train.py:904] (5/8) Epoch 24, batch 6550, loss[loss=0.1965, simple_loss=0.2956, pruned_loss=0.04866, over 16298.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2878, pruned_loss=0.05685, over 3090765.65 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:36:48,716 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 20:37:16,977 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.521e+02 3.197e+02 3.720e+02 9.360e+02, threshold=6.395e+02, percent-clipped=2.0 2023-05-01 20:37:54,322 INFO [train.py:904] (5/8) Epoch 24, batch 6600, loss[loss=0.2065, simple_loss=0.293, pruned_loss=0.06, over 15276.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2913, pruned_loss=0.05842, over 3082798.65 frames. ], batch size: 190, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:39:11,717 INFO [train.py:904] (5/8) Epoch 24, batch 6650, loss[loss=0.1839, simple_loss=0.2714, pruned_loss=0.04816, over 16938.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2921, pruned_loss=0.05998, over 3066515.70 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:39:50,366 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.925e+02 3.532e+02 4.365e+02 9.170e+02, threshold=7.065e+02, percent-clipped=1.0 2023-05-01 20:40:28,845 INFO [train.py:904] (5/8) Epoch 24, batch 6700, loss[loss=0.2138, simple_loss=0.2967, pruned_loss=0.06546, over 16478.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2898, pruned_loss=0.05909, over 3082208.21 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:41:45,724 INFO [train.py:904] (5/8) Epoch 24, batch 6750, loss[loss=0.2221, simple_loss=0.2943, pruned_loss=0.07492, over 12073.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2889, pruned_loss=0.05921, over 3083170.36 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:42:23,538 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.933e+02 2.874e+02 3.326e+02 4.186e+02 6.587e+02, threshold=6.652e+02, percent-clipped=0.0 2023-05-01 20:43:01,410 INFO [train.py:904] (5/8) Epoch 24, batch 6800, loss[loss=0.2093, simple_loss=0.2914, pruned_loss=0.06357, over 16663.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2886, pruned_loss=0.05863, over 3095588.59 frames. ], batch size: 134, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:43:33,053 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-05-01 20:43:42,209 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:44:06,652 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0901, 5.0297, 4.9144, 3.9501, 4.9584, 1.6472, 4.6280, 4.4894], device='cuda:5'), covar=tensor([0.0155, 0.0141, 0.0196, 0.0565, 0.0148, 0.3208, 0.0287, 0.0281], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0165, 0.0207, 0.0183, 0.0181, 0.0212, 0.0194, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:44:07,883 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6420, 2.5280, 2.5410, 4.1610, 2.7954, 3.9820, 1.5467, 2.8918], device='cuda:5'), covar=tensor([0.1456, 0.0840, 0.1203, 0.0166, 0.0258, 0.0416, 0.1692, 0.0817], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0193, 0.0205, 0.0216, 0.0203, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:44:21,121 INFO [train.py:904] (5/8) Epoch 24, batch 6850, loss[loss=0.1896, simple_loss=0.2948, pruned_loss=0.04219, over 17250.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2893, pruned_loss=0.05817, over 3096867.69 frames. ], batch size: 52, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:44:56,067 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=240326.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:45:00,026 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.623e+02 3.086e+02 3.760e+02 7.465e+02, threshold=6.171e+02, percent-clipped=2.0 2023-05-01 20:45:17,529 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:45:35,455 INFO [train.py:904] (5/8) Epoch 24, batch 6900, loss[loss=0.2176, simple_loss=0.3024, pruned_loss=0.06643, over 15397.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2913, pruned_loss=0.05773, over 3104664.36 frames. ], batch size: 190, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:46:26,555 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 20:46:33,646 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-05-01 20:46:50,846 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240401.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:46:52,770 INFO [train.py:904] (5/8) Epoch 24, batch 6950, loss[loss=0.1774, simple_loss=0.2701, pruned_loss=0.04231, over 17048.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2932, pruned_loss=0.05979, over 3086253.51 frames. ], batch size: 50, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:47:33,344 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 2.974e+02 3.564e+02 4.384e+02 8.202e+02, threshold=7.128e+02, percent-clipped=2.0 2023-05-01 20:48:07,692 INFO [train.py:904] (5/8) Epoch 24, batch 7000, loss[loss=0.2114, simple_loss=0.3067, pruned_loss=0.05802, over 16937.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2939, pruned_loss=0.05977, over 3080668.06 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:48:32,181 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9607, 4.0189, 4.2947, 4.2586, 4.2828, 4.0370, 4.0256, 3.9809], device='cuda:5'), covar=tensor([0.0379, 0.0668, 0.0495, 0.0485, 0.0499, 0.0512, 0.0973, 0.0601], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0464, 0.0451, 0.0416, 0.0497, 0.0472, 0.0555, 0.0377], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 20:48:49,500 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3236, 3.0987, 3.4397, 1.8659, 3.5830, 3.6128, 2.8195, 2.6935], device='cuda:5'), covar=tensor([0.0861, 0.0304, 0.0216, 0.1239, 0.0092, 0.0197, 0.0489, 0.0515], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0138, 0.0082, 0.0128, 0.0129, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 20:49:14,349 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0416, 2.2235, 2.2267, 3.7591, 2.1192, 2.5186, 2.2474, 2.3461], device='cuda:5'), covar=tensor([0.1486, 0.3647, 0.3245, 0.0627, 0.4351, 0.2703, 0.3676, 0.3566], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0457, 0.0375, 0.0330, 0.0440, 0.0523, 0.0429, 0.0535], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:49:21,393 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9183, 4.8920, 4.7210, 4.0110, 4.8237, 1.8654, 4.5844, 4.4672], device='cuda:5'), covar=tensor([0.0112, 0.0101, 0.0213, 0.0426, 0.0108, 0.2864, 0.0147, 0.0259], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0165, 0.0207, 0.0183, 0.0181, 0.0213, 0.0194, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:49:23,903 INFO [train.py:904] (5/8) Epoch 24, batch 7050, loss[loss=0.2115, simple_loss=0.3019, pruned_loss=0.06049, over 16242.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2939, pruned_loss=0.05897, over 3096917.99 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:49:24,965 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0722, 5.1874, 5.5457, 5.4893, 5.5193, 5.1624, 5.0891, 4.8442], device='cuda:5'), covar=tensor([0.0320, 0.0503, 0.0321, 0.0375, 0.0511, 0.0395, 0.0933, 0.0548], device='cuda:5'), in_proj_covar=tensor([0.0416, 0.0465, 0.0451, 0.0416, 0.0498, 0.0473, 0.0555, 0.0377], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 20:50:06,659 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.536e+02 3.115e+02 4.008e+02 6.614e+02, threshold=6.229e+02, percent-clipped=0.0 2023-05-01 20:50:42,264 INFO [train.py:904] (5/8) Epoch 24, batch 7100, loss[loss=0.1867, simple_loss=0.2761, pruned_loss=0.04864, over 16625.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2915, pruned_loss=0.05803, over 3100476.30 frames. ], batch size: 62, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:51:59,039 INFO [train.py:904] (5/8) Epoch 24, batch 7150, loss[loss=0.2235, simple_loss=0.302, pruned_loss=0.07254, over 16407.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2896, pruned_loss=0.05773, over 3105642.27 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:52:39,168 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.941e+02 3.518e+02 4.081e+02 6.999e+02, threshold=7.036e+02, percent-clipped=1.0 2023-05-01 20:52:49,458 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 20:53:12,654 INFO [train.py:904] (5/8) Epoch 24, batch 7200, loss[loss=0.1735, simple_loss=0.2712, pruned_loss=0.03786, over 16514.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2875, pruned_loss=0.05652, over 3093505.31 frames. ], batch size: 68, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:54:20,553 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=240696.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:54:32,017 INFO [train.py:904] (5/8) Epoch 24, batch 7250, loss[loss=0.1898, simple_loss=0.2767, pruned_loss=0.05143, over 16402.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.285, pruned_loss=0.0554, over 3089152.09 frames. ], batch size: 68, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:54:44,328 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 20:55:12,207 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.642e+02 3.264e+02 4.366e+02 7.559e+02, threshold=6.528e+02, percent-clipped=1.0 2023-05-01 20:55:45,068 INFO [train.py:904] (5/8) Epoch 24, batch 7300, loss[loss=0.2132, simple_loss=0.3062, pruned_loss=0.06015, over 16433.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2851, pruned_loss=0.05523, over 3111566.83 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:56:03,626 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6863, 3.8738, 2.9119, 2.3671, 2.6506, 2.5755, 4.3434, 3.4356], device='cuda:5'), covar=tensor([0.3117, 0.0678, 0.1930, 0.2727, 0.2916, 0.2114, 0.0418, 0.1423], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0272, 0.0308, 0.0321, 0.0301, 0.0266, 0.0300, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 20:56:58,478 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5000, 3.4680, 3.4424, 2.6864, 3.3815, 2.1005, 3.1540, 2.7932], device='cuda:5'), covar=tensor([0.0149, 0.0121, 0.0202, 0.0240, 0.0100, 0.2383, 0.0132, 0.0286], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0163, 0.0204, 0.0180, 0.0178, 0.0209, 0.0191, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:57:02,850 INFO [train.py:904] (5/8) Epoch 24, batch 7350, loss[loss=0.2038, simple_loss=0.2852, pruned_loss=0.06121, over 16386.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2861, pruned_loss=0.05573, over 3127629.32 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:57:44,659 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.735e+02 3.201e+02 4.098e+02 9.519e+02, threshold=6.402e+02, percent-clipped=6.0 2023-05-01 20:58:18,672 INFO [train.py:904] (5/8) Epoch 24, batch 7400, loss[loss=0.2266, simple_loss=0.3014, pruned_loss=0.07593, over 11113.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2875, pruned_loss=0.05658, over 3127736.44 frames. ], batch size: 247, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:58:29,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4912, 4.5822, 4.3871, 4.0514, 4.0661, 4.4868, 4.2005, 4.2128], device='cuda:5'), covar=tensor([0.0659, 0.0724, 0.0317, 0.0334, 0.0876, 0.0534, 0.0648, 0.0707], device='cuda:5'), in_proj_covar=tensor([0.0288, 0.0433, 0.0338, 0.0338, 0.0342, 0.0391, 0.0233, 0.0407], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:58:55,145 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5836, 2.2649, 1.8289, 2.0013, 2.5428, 2.2280, 2.3339, 2.7007], device='cuda:5'), covar=tensor([0.0224, 0.0451, 0.0594, 0.0509, 0.0286, 0.0408, 0.0224, 0.0275], device='cuda:5'), in_proj_covar=tensor([0.0211, 0.0232, 0.0225, 0.0226, 0.0234, 0.0232, 0.0232, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 20:59:34,872 INFO [train.py:904] (5/8) Epoch 24, batch 7450, loss[loss=0.2122, simple_loss=0.3051, pruned_loss=0.05971, over 15243.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2885, pruned_loss=0.0572, over 3137326.93 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:00:19,563 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.843e+02 3.471e+02 4.367e+02 9.441e+02, threshold=6.943e+02, percent-clipped=5.0 2023-05-01 21:00:22,565 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240932.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:00:55,587 INFO [train.py:904] (5/8) Epoch 24, batch 7500, loss[loss=0.2027, simple_loss=0.2914, pruned_loss=0.05703, over 16632.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2889, pruned_loss=0.0566, over 3119181.78 frames. ], batch size: 57, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:01:35,383 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8349, 1.9676, 2.2491, 3.2684, 1.9406, 2.1889, 2.1071, 2.0996], device='cuda:5'), covar=tensor([0.1795, 0.4367, 0.3061, 0.0843, 0.5164, 0.3179, 0.4165, 0.3863], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0458, 0.0373, 0.0329, 0.0439, 0.0523, 0.0429, 0.0533], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:01:46,142 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240986.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:01:57,385 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240993.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:02:00,997 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240996.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:02:11,657 INFO [train.py:904] (5/8) Epoch 24, batch 7550, loss[loss=0.2044, simple_loss=0.2905, pruned_loss=0.05918, over 16171.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.288, pruned_loss=0.05729, over 3102347.86 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:02:53,687 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.697e+02 3.200e+02 3.848e+02 7.597e+02, threshold=6.400e+02, percent-clipped=1.0 2023-05-01 21:03:15,842 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:20,296 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241047.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:29,259 INFO [train.py:904] (5/8) Epoch 24, batch 7600, loss[loss=0.1869, simple_loss=0.2801, pruned_loss=0.04687, over 16791.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2871, pruned_loss=0.05775, over 3086566.56 frames. ], batch size: 102, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:03:48,167 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5775, 2.6541, 2.4784, 4.1529, 3.0198, 3.9808, 1.4594, 2.9430], device='cuda:5'), covar=tensor([0.1469, 0.0815, 0.1312, 0.0202, 0.0287, 0.0394, 0.1785, 0.0813], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0177, 0.0197, 0.0194, 0.0206, 0.0216, 0.0205, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 21:04:22,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4932, 3.5000, 2.6041, 2.1755, 2.2818, 2.2719, 3.7100, 3.1374], device='cuda:5'), covar=tensor([0.3106, 0.0733, 0.2032, 0.2938, 0.2907, 0.2343, 0.0539, 0.1419], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0270, 0.0306, 0.0318, 0.0299, 0.0265, 0.0298, 0.0340], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 21:04:47,295 INFO [train.py:904] (5/8) Epoch 24, batch 7650, loss[loss=0.1999, simple_loss=0.2864, pruned_loss=0.05672, over 16662.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2872, pruned_loss=0.05762, over 3095443.12 frames. ], batch size: 134, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:05:30,618 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 2.751e+02 3.488e+02 4.821e+02 1.001e+03, threshold=6.976e+02, percent-clipped=5.0 2023-05-01 21:06:04,723 INFO [train.py:904] (5/8) Epoch 24, batch 7700, loss[loss=0.1855, simple_loss=0.2782, pruned_loss=0.04643, over 16720.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2873, pruned_loss=0.05797, over 3090760.43 frames. ], batch size: 134, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:06:09,638 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4730, 3.4417, 3.4350, 2.6730, 3.3050, 2.1281, 3.1175, 2.7253], device='cuda:5'), covar=tensor([0.0170, 0.0150, 0.0199, 0.0241, 0.0113, 0.2266, 0.0145, 0.0246], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0163, 0.0204, 0.0180, 0.0179, 0.0209, 0.0191, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:07:23,967 INFO [train.py:904] (5/8) Epoch 24, batch 7750, loss[loss=0.2177, simple_loss=0.3103, pruned_loss=0.06252, over 16718.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2881, pruned_loss=0.05807, over 3095920.64 frames. ], batch size: 124, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:06,386 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.844e+02 3.408e+02 3.883e+02 6.664e+02, threshold=6.815e+02, percent-clipped=0.0 2023-05-01 21:08:38,523 INFO [train.py:904] (5/8) Epoch 24, batch 7800, loss[loss=0.1951, simple_loss=0.2798, pruned_loss=0.05519, over 16874.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2886, pruned_loss=0.05866, over 3081516.78 frames. ], batch size: 116, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:53,106 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1676, 2.4173, 2.0270, 2.2033, 2.7807, 2.4453, 2.7534, 2.9840], device='cuda:5'), covar=tensor([0.0193, 0.0481, 0.0599, 0.0492, 0.0314, 0.0402, 0.0279, 0.0303], device='cuda:5'), in_proj_covar=tensor([0.0213, 0.0234, 0.0226, 0.0227, 0.0236, 0.0233, 0.0234, 0.0231], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:09:31,586 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-05-01 21:09:34,096 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241288.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 21:09:56,327 INFO [train.py:904] (5/8) Epoch 24, batch 7850, loss[loss=0.2381, simple_loss=0.3092, pruned_loss=0.08344, over 11894.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2894, pruned_loss=0.05854, over 3073714.92 frames. ], batch size: 247, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:10:38,482 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.020e+02 2.608e+02 3.095e+02 3.566e+02 5.691e+02, threshold=6.191e+02, percent-clipped=0.0 2023-05-01 21:10:55,480 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241342.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:10:59,490 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4428, 4.4122, 4.2474, 3.2727, 4.3639, 1.5797, 4.0716, 3.9365], device='cuda:5'), covar=tensor([0.0175, 0.0135, 0.0269, 0.0612, 0.0143, 0.3460, 0.0198, 0.0364], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0163, 0.0205, 0.0181, 0.0179, 0.0210, 0.0191, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:11:11,149 INFO [train.py:904] (5/8) Epoch 24, batch 7900, loss[loss=0.2048, simple_loss=0.2898, pruned_loss=0.05986, over 15550.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2885, pruned_loss=0.05851, over 3058212.48 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:12:24,153 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4836, 2.2637, 1.8503, 2.0691, 2.5040, 2.2601, 2.3208, 2.6848], device='cuda:5'), covar=tensor([0.0208, 0.0376, 0.0542, 0.0433, 0.0264, 0.0330, 0.0251, 0.0237], device='cuda:5'), in_proj_covar=tensor([0.0210, 0.0231, 0.0223, 0.0224, 0.0233, 0.0230, 0.0231, 0.0228], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:12:29,011 INFO [train.py:904] (5/8) Epoch 24, batch 7950, loss[loss=0.2666, simple_loss=0.3234, pruned_loss=0.1049, over 11799.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2887, pruned_loss=0.0584, over 3076149.48 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:12:35,111 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241406.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:13:12,618 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 2.663e+02 3.306e+02 3.929e+02 7.099e+02, threshold=6.611e+02, percent-clipped=2.0 2023-05-01 21:13:16,017 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 21:13:46,686 INFO [train.py:904] (5/8) Epoch 24, batch 8000, loss[loss=0.1971, simple_loss=0.2936, pruned_loss=0.0503, over 16863.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2886, pruned_loss=0.05866, over 3072309.77 frames. ], batch size: 102, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:14:09,563 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241467.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:14:41,732 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-01 21:14:45,016 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3594, 4.0237, 3.9731, 2.6184, 3.5905, 4.0429, 3.5876, 2.4024], device='cuda:5'), covar=tensor([0.0582, 0.0059, 0.0059, 0.0438, 0.0115, 0.0104, 0.0110, 0.0439], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0086, 0.0087, 0.0135, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 21:15:04,299 INFO [train.py:904] (5/8) Epoch 24, batch 8050, loss[loss=0.2348, simple_loss=0.2974, pruned_loss=0.08608, over 11880.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2886, pruned_loss=0.05854, over 3063591.18 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:15:47,626 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.671e+02 3.058e+02 3.888e+02 6.821e+02, threshold=6.115e+02, percent-clipped=1.0 2023-05-01 21:16:22,189 INFO [train.py:904] (5/8) Epoch 24, batch 8100, loss[loss=0.1984, simple_loss=0.2866, pruned_loss=0.05509, over 16584.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2879, pruned_loss=0.05793, over 3069861.30 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:17:14,368 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241588.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:17:20,010 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 21:17:38,325 INFO [train.py:904] (5/8) Epoch 24, batch 8150, loss[loss=0.1625, simple_loss=0.2498, pruned_loss=0.0376, over 16453.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2856, pruned_loss=0.05675, over 3101512.44 frames. ], batch size: 75, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:18:06,891 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:22,059 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.026e+02 2.691e+02 3.345e+02 4.318e+02 1.089e+03, threshold=6.691e+02, percent-clipped=5.0 2023-05-01 21:18:30,850 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241636.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:18:37,210 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2358, 2.3425, 2.4337, 3.9947, 2.3219, 2.6409, 2.4240, 2.5224], device='cuda:5'), covar=tensor([0.1392, 0.3502, 0.2938, 0.0584, 0.4104, 0.2675, 0.3492, 0.3192], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0459, 0.0374, 0.0331, 0.0441, 0.0524, 0.0429, 0.0535], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:18:39,519 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241642.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:55,847 INFO [train.py:904] (5/8) Epoch 24, batch 8200, loss[loss=0.1796, simple_loss=0.2724, pruned_loss=0.04345, over 16195.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2834, pruned_loss=0.05654, over 3068815.61 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:19:38,374 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4030, 3.3521, 3.4839, 3.5485, 3.6061, 3.3177, 3.5304, 3.6499], device='cuda:5'), covar=tensor([0.1425, 0.1077, 0.1128, 0.0672, 0.0725, 0.2473, 0.1173, 0.0889], device='cuda:5'), in_proj_covar=tensor([0.0638, 0.0789, 0.0904, 0.0793, 0.0609, 0.0627, 0.0661, 0.0767], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:19:42,968 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:19:53,197 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 21:19:56,097 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241690.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:20:16,389 INFO [train.py:904] (5/8) Epoch 24, batch 8250, loss[loss=0.1828, simple_loss=0.2808, pruned_loss=0.04246, over 16252.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2825, pruned_loss=0.05413, over 3054053.17 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:21:03,234 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.441e+02 2.851e+02 3.404e+02 7.117e+02, threshold=5.702e+02, percent-clipped=1.0 2023-05-01 21:21:05,752 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-05-01 21:21:20,194 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9588, 2.6972, 2.6174, 2.0028, 2.5299, 2.7587, 2.6355, 2.0486], device='cuda:5'), covar=tensor([0.0406, 0.0101, 0.0096, 0.0341, 0.0141, 0.0117, 0.0115, 0.0417], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0085, 0.0086, 0.0133, 0.0098, 0.0110, 0.0095, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 21:21:39,008 INFO [train.py:904] (5/8) Epoch 24, batch 8300, loss[loss=0.1799, simple_loss=0.2791, pruned_loss=0.04036, over 15295.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2799, pruned_loss=0.05092, over 3064563.91 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:21:41,901 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5331, 4.4985, 4.3542, 3.5970, 4.4294, 1.6769, 4.1748, 3.9958], device='cuda:5'), covar=tensor([0.0102, 0.0115, 0.0212, 0.0348, 0.0098, 0.2981, 0.0152, 0.0279], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0162, 0.0203, 0.0179, 0.0178, 0.0208, 0.0191, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:21:54,684 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241762.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:23:02,699 INFO [train.py:904] (5/8) Epoch 24, batch 8350, loss[loss=0.1642, simple_loss=0.2649, pruned_loss=0.03171, over 16896.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2798, pruned_loss=0.0497, over 3058342.88 frames. ], batch size: 83, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:23:03,888 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241803.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:23:37,829 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4961, 3.1489, 3.2734, 1.8635, 3.4683, 3.5057, 2.9517, 2.8384], device='cuda:5'), covar=tensor([0.0675, 0.0260, 0.0245, 0.1227, 0.0090, 0.0208, 0.0419, 0.0418], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0107, 0.0097, 0.0135, 0.0080, 0.0125, 0.0126, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 21:23:48,199 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.476e+02 2.114e+02 2.559e+02 3.113e+02 5.727e+02, threshold=5.118e+02, percent-clipped=1.0 2023-05-01 21:23:48,818 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241831.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:24:23,530 INFO [train.py:904] (5/8) Epoch 24, batch 8400, loss[loss=0.1701, simple_loss=0.2652, pruned_loss=0.03751, over 16915.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2769, pruned_loss=0.04759, over 3040539.01 frames. ], batch size: 90, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:24:42,815 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241864.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:25:14,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9504, 2.1899, 2.3972, 3.2219, 2.2270, 2.3592, 2.3613, 2.2974], device='cuda:5'), covar=tensor([0.1283, 0.3660, 0.2777, 0.0713, 0.4588, 0.2618, 0.3545, 0.3462], device='cuda:5'), in_proj_covar=tensor([0.0401, 0.0450, 0.0368, 0.0324, 0.0433, 0.0513, 0.0421, 0.0525], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:25:28,873 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241892.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 21:25:45,959 INFO [train.py:904] (5/8) Epoch 24, batch 8450, loss[loss=0.1726, simple_loss=0.2642, pruned_loss=0.0405, over 16855.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2754, pruned_loss=0.04618, over 3043089.66 frames. ], batch size: 116, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:26:31,269 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.077e+02 2.419e+02 2.958e+02 5.415e+02, threshold=4.839e+02, percent-clipped=2.0 2023-05-01 21:27:07,283 INFO [train.py:904] (5/8) Epoch 24, batch 8500, loss[loss=0.1727, simple_loss=0.2637, pruned_loss=0.04083, over 16817.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2717, pruned_loss=0.04393, over 3038302.91 frames. ], batch size: 116, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:27:47,695 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:28:00,038 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6785, 4.9531, 4.7604, 4.7724, 4.5180, 4.5360, 4.4464, 5.0424], device='cuda:5'), covar=tensor([0.1295, 0.1033, 0.1114, 0.0877, 0.0895, 0.1103, 0.1120, 0.0927], device='cuda:5'), in_proj_covar=tensor([0.0684, 0.0821, 0.0682, 0.0638, 0.0522, 0.0530, 0.0692, 0.0644], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:28:29,785 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-01 21:28:33,832 INFO [train.py:904] (5/8) Epoch 24, batch 8550, loss[loss=0.1707, simple_loss=0.2553, pruned_loss=0.04303, over 12108.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2693, pruned_loss=0.04302, over 3003844.38 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:29:05,811 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 21:29:26,921 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.227e+02 2.590e+02 3.113e+02 4.507e+02, threshold=5.180e+02, percent-clipped=0.0 2023-05-01 21:30:13,222 INFO [train.py:904] (5/8) Epoch 24, batch 8600, loss[loss=0.1715, simple_loss=0.2709, pruned_loss=0.03607, over 15446.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2701, pruned_loss=0.0418, over 3027643.42 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:30:31,543 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:31:51,613 INFO [train.py:904] (5/8) Epoch 24, batch 8650, loss[loss=0.156, simple_loss=0.2474, pruned_loss=0.03232, over 12058.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2683, pruned_loss=0.04026, over 3025912.81 frames. ], batch size: 250, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:32:10,579 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242110.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:32:56,844 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.159e+02 2.524e+02 3.190e+02 5.272e+02, threshold=5.049e+02, percent-clipped=1.0 2023-05-01 21:33:30,866 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 21:33:37,189 INFO [train.py:904] (5/8) Epoch 24, batch 8700, loss[loss=0.1725, simple_loss=0.2608, pruned_loss=0.04208, over 12426.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2655, pruned_loss=0.0391, over 3039719.04 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:33:38,180 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242153.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:33:38,341 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0712, 2.1148, 2.1959, 3.5869, 2.1075, 2.4212, 2.2460, 2.2487], device='cuda:5'), covar=tensor([0.1339, 0.3761, 0.3171, 0.0593, 0.4294, 0.2614, 0.3820, 0.3500], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0448, 0.0367, 0.0322, 0.0431, 0.0510, 0.0419, 0.0522], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:33:50,500 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:34:42,449 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242187.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:35:13,456 INFO [train.py:904] (5/8) Epoch 24, batch 8750, loss[loss=0.1564, simple_loss=0.2464, pruned_loss=0.03324, over 12213.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2647, pruned_loss=0.03831, over 3033377.40 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:35:42,289 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:10,188 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:21,486 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.077e+02 2.590e+02 3.155e+02 6.830e+02, threshold=5.181e+02, percent-clipped=2.0 2023-05-01 21:37:05,081 INFO [train.py:904] (5/8) Epoch 24, batch 8800, loss[loss=0.181, simple_loss=0.2735, pruned_loss=0.04418, over 12322.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2638, pruned_loss=0.03746, over 3048176.85 frames. ], batch size: 247, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:37:56,101 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242277.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:18,011 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242287.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:23,578 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-01 21:38:50,070 INFO [train.py:904] (5/8) Epoch 24, batch 8850, loss[loss=0.1713, simple_loss=0.2776, pruned_loss=0.03253, over 16385.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2666, pruned_loss=0.03717, over 3037729.96 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:39:03,990 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242309.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:38,583 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:51,889 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7210, 4.5495, 4.7934, 4.9297, 5.1045, 4.5833, 5.0926, 5.1025], device='cuda:5'), covar=tensor([0.2042, 0.1270, 0.1628, 0.0761, 0.0516, 0.0838, 0.0552, 0.0617], device='cuda:5'), in_proj_covar=tensor([0.0624, 0.0772, 0.0884, 0.0780, 0.0596, 0.0616, 0.0647, 0.0750], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:39:52,558 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.137e+02 2.486e+02 3.124e+02 5.230e+02, threshold=4.972e+02, percent-clipped=1.0 2023-05-01 21:40:34,039 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8324, 3.8824, 4.0420, 3.7733, 3.9605, 4.3449, 4.0017, 3.7444], device='cuda:5'), covar=tensor([0.2220, 0.2118, 0.2170, 0.2630, 0.2520, 0.1539, 0.1532, 0.2594], device='cuda:5'), in_proj_covar=tensor([0.0404, 0.0597, 0.0656, 0.0490, 0.0644, 0.0681, 0.0511, 0.0652], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 21:40:38,021 INFO [train.py:904] (5/8) Epoch 24, batch 8900, loss[loss=0.1539, simple_loss=0.2516, pruned_loss=0.02814, over 12890.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2676, pruned_loss=0.03672, over 3052162.17 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:41:12,523 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:42:06,391 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4856, 4.2727, 4.5471, 4.6832, 4.8468, 4.3765, 4.8287, 4.8615], device='cuda:5'), covar=tensor([0.1929, 0.1464, 0.1760, 0.0849, 0.0532, 0.0949, 0.0568, 0.0734], device='cuda:5'), in_proj_covar=tensor([0.0623, 0.0770, 0.0883, 0.0778, 0.0595, 0.0616, 0.0645, 0.0748], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:42:44,442 INFO [train.py:904] (5/8) Epoch 24, batch 8950, loss[loss=0.159, simple_loss=0.2501, pruned_loss=0.03394, over 15279.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2676, pruned_loss=0.03736, over 3073307.49 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:43:49,729 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.299e+02 2.560e+02 2.923e+02 8.049e+02, threshold=5.120e+02, percent-clipped=2.0 2023-05-01 21:44:35,831 INFO [train.py:904] (5/8) Epoch 24, batch 9000, loss[loss=0.1517, simple_loss=0.2435, pruned_loss=0.02995, over 16368.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2642, pruned_loss=0.03603, over 3072843.18 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:44:35,831 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 21:44:45,533 INFO [train.py:938] (5/8) Epoch 24, validation: loss=0.1445, simple_loss=0.2484, pruned_loss=0.02026, over 944034.00 frames. 2023-05-01 21:44:45,534 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-01 21:44:59,767 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:37,073 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4906, 3.7025, 2.6885, 2.0974, 2.2140, 2.2189, 3.9436, 3.1562], device='cuda:5'), covar=tensor([0.3173, 0.0533, 0.1915, 0.3015, 0.2970, 0.2327, 0.0386, 0.1364], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0265, 0.0302, 0.0313, 0.0292, 0.0262, 0.0293, 0.0334], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 21:45:49,167 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:58,892 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242487.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:46:30,351 INFO [train.py:904] (5/8) Epoch 24, batch 9050, loss[loss=0.1956, simple_loss=0.2792, pruned_loss=0.05602, over 13059.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2653, pruned_loss=0.03643, over 3086095.60 frames. ], batch size: 246, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:46:33,849 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6873, 2.5017, 2.4203, 4.4252, 2.3464, 2.8592, 2.5109, 2.6673], device='cuda:5'), covar=tensor([0.1102, 0.3429, 0.3101, 0.0433, 0.4169, 0.2385, 0.3526, 0.3027], device='cuda:5'), in_proj_covar=tensor([0.0399, 0.0449, 0.0369, 0.0323, 0.0433, 0.0511, 0.0421, 0.0522], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 21:46:40,251 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:46:45,346 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:46:59,926 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2353, 3.3422, 3.3769, 2.3110, 3.1785, 3.4536, 3.2669, 2.0031], device='cuda:5'), covar=tensor([0.0512, 0.0065, 0.0057, 0.0393, 0.0098, 0.0081, 0.0082, 0.0506], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0084, 0.0084, 0.0131, 0.0097, 0.0107, 0.0093, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 21:47:17,052 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2388, 3.2301, 1.9412, 3.5281, 2.4567, 3.4968, 2.1580, 2.7236], device='cuda:5'), covar=tensor([0.0259, 0.0348, 0.1675, 0.0261, 0.0817, 0.0609, 0.1536, 0.0691], device='cuda:5'), in_proj_covar=tensor([0.0166, 0.0172, 0.0189, 0.0160, 0.0174, 0.0210, 0.0199, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 21:47:30,079 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.227e+02 2.575e+02 3.137e+02 5.273e+02, threshold=5.150e+02, percent-clipped=1.0 2023-05-01 21:47:35,727 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:48:00,767 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242544.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:48:17,910 INFO [train.py:904] (5/8) Epoch 24, batch 9100, loss[loss=0.1654, simple_loss=0.2672, pruned_loss=0.03183, over 16211.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2648, pruned_loss=0.03686, over 3084171.80 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:48:23,288 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242555.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:49:27,659 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242582.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:16,299 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242602.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:16,971 INFO [train.py:904] (5/8) Epoch 24, batch 9150, loss[loss=0.1727, simple_loss=0.2628, pruned_loss=0.04134, over 16485.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2654, pruned_loss=0.03668, over 3080430.64 frames. ], batch size: 147, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:50:47,190 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:51:21,273 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.227e+02 2.578e+02 3.148e+02 6.071e+02, threshold=5.155e+02, percent-clipped=3.0 2023-05-01 21:52:01,747 INFO [train.py:904] (5/8) Epoch 24, batch 9200, loss[loss=0.1534, simple_loss=0.2392, pruned_loss=0.03386, over 12347.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2608, pruned_loss=0.03588, over 3073582.32 frames. ], batch size: 249, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:52:22,359 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:52:24,969 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:53:38,107 INFO [train.py:904] (5/8) Epoch 24, batch 9250, loss[loss=0.1386, simple_loss=0.2236, pruned_loss=0.02679, over 12105.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2606, pruned_loss=0.03598, over 3062125.33 frames. ], batch size: 249, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:54:40,873 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.148e+02 2.469e+02 2.996e+02 5.681e+02, threshold=4.938e+02, percent-clipped=3.0 2023-05-01 21:54:50,849 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 21:55:27,531 INFO [train.py:904] (5/8) Epoch 24, batch 9300, loss[loss=0.1555, simple_loss=0.2474, pruned_loss=0.03175, over 16769.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2584, pruned_loss=0.03519, over 3058534.79 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:56:13,869 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:57:14,584 INFO [train.py:904] (5/8) Epoch 24, batch 9350, loss[loss=0.1579, simple_loss=0.2526, pruned_loss=0.03159, over 16836.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2583, pruned_loss=0.03497, over 3072364.68 frames. ], batch size: 83, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:57:27,694 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242809.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:13,365 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.321e+02 1.952e+02 2.311e+02 3.086e+02 6.417e+02, threshold=4.623e+02, percent-clipped=1.0 2023-05-01 21:58:14,236 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:25,515 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242839.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:54,555 INFO [train.py:904] (5/8) Epoch 24, batch 9400, loss[loss=0.1595, simple_loss=0.2616, pruned_loss=0.02877, over 15398.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2587, pruned_loss=0.035, over 3069191.27 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:59:03,720 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:59:55,004 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:00:35,999 INFO [train.py:904] (5/8) Epoch 24, batch 9450, loss[loss=0.1498, simple_loss=0.2455, pruned_loss=0.02708, over 17156.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.26, pruned_loss=0.03511, over 3053800.19 frames. ], batch size: 48, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:00:51,522 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:08,948 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8522, 1.3602, 1.7634, 1.7777, 1.8873, 1.9433, 1.6626, 1.8823], device='cuda:5'), covar=tensor([0.0303, 0.0489, 0.0255, 0.0341, 0.0352, 0.0248, 0.0498, 0.0166], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0190, 0.0177, 0.0179, 0.0195, 0.0154, 0.0193, 0.0152], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:01:14,133 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:32,437 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:37,831 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 2.255e+02 2.567e+02 3.168e+02 7.490e+02, threshold=5.133e+02, percent-clipped=6.0 2023-05-01 22:01:45,311 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7206, 4.9261, 5.0357, 4.8031, 4.8816, 5.3855, 4.8582, 4.5887], device='cuda:5'), covar=tensor([0.1038, 0.1582, 0.2172, 0.1838, 0.2290, 0.0874, 0.1572, 0.2245], device='cuda:5'), in_proj_covar=tensor([0.0394, 0.0586, 0.0646, 0.0480, 0.0633, 0.0673, 0.0501, 0.0637], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 22:01:54,015 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 22:02:16,745 INFO [train.py:904] (5/8) Epoch 24, batch 9500, loss[loss=0.1848, simple_loss=0.277, pruned_loss=0.0463, over 16962.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2595, pruned_loss=0.0349, over 3076851.58 frames. ], batch size: 109, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:02:18,022 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8805, 3.8397, 4.0362, 3.7286, 3.9712, 4.3612, 4.0055, 3.6229], device='cuda:5'), covar=tensor([0.1924, 0.2163, 0.2125, 0.2506, 0.2407, 0.1509, 0.1579, 0.2723], device='cuda:5'), in_proj_covar=tensor([0.0393, 0.0585, 0.0645, 0.0480, 0.0633, 0.0672, 0.0501, 0.0637], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 22:02:30,103 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:02:43,838 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:03:05,402 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1009, 2.7749, 2.7896, 1.9147, 2.6154, 1.9746, 2.8079, 2.9243], device='cuda:5'), covar=tensor([0.0295, 0.0976, 0.0617, 0.2317, 0.1032, 0.1221, 0.0655, 0.0865], device='cuda:5'), in_proj_covar=tensor([0.0151, 0.0157, 0.0162, 0.0149, 0.0140, 0.0126, 0.0138, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:03:16,241 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242982.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 22:04:01,179 INFO [train.py:904] (5/8) Epoch 24, batch 9550, loss[loss=0.1783, simple_loss=0.2635, pruned_loss=0.04658, over 12456.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2591, pruned_loss=0.03529, over 3082549.02 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:04:24,733 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243013.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:05:06,081 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.055e+02 2.389e+02 2.872e+02 5.971e+02, threshold=4.777e+02, percent-clipped=1.0 2023-05-01 22:05:43,259 INFO [train.py:904] (5/8) Epoch 24, batch 9600, loss[loss=0.1861, simple_loss=0.2692, pruned_loss=0.05146, over 12251.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2609, pruned_loss=0.03603, over 3062828.32 frames. ], batch size: 247, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:06:12,760 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9167, 2.2924, 2.3331, 3.0886, 1.7265, 3.2291, 1.7349, 2.7288], device='cuda:5'), covar=tensor([0.1272, 0.0673, 0.1075, 0.0158, 0.0078, 0.0363, 0.1596, 0.0744], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0173, 0.0193, 0.0187, 0.0197, 0.0211, 0.0202, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:06:16,778 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2951, 3.5174, 3.8671, 2.1688, 3.2947, 2.4467, 3.6924, 3.6342], device='cuda:5'), covar=tensor([0.0223, 0.0797, 0.0483, 0.1993, 0.0722, 0.0970, 0.0514, 0.0947], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0158, 0.0163, 0.0150, 0.0141, 0.0127, 0.0139, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:07:31,369 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6406, 4.4265, 4.6786, 4.8321, 4.9764, 4.4758, 4.9990, 4.9982], device='cuda:5'), covar=tensor([0.1695, 0.1365, 0.1732, 0.0761, 0.0633, 0.0936, 0.0559, 0.0742], device='cuda:5'), in_proj_covar=tensor([0.0617, 0.0762, 0.0874, 0.0769, 0.0590, 0.0607, 0.0640, 0.0741], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:07:32,102 INFO [train.py:904] (5/8) Epoch 24, batch 9650, loss[loss=0.1542, simple_loss=0.2507, pruned_loss=0.02889, over 16461.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2624, pruned_loss=0.03597, over 3069065.92 frames. ], batch size: 75, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:08:29,798 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243128.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:08:37,737 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.193e+02 2.679e+02 3.134e+02 6.217e+02, threshold=5.359e+02, percent-clipped=4.0 2023-05-01 22:08:50,202 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243139.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:09:19,555 INFO [train.py:904] (5/8) Epoch 24, batch 9700, loss[loss=0.1733, simple_loss=0.269, pruned_loss=0.03881, over 16958.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2618, pruned_loss=0.03584, over 3075294.78 frames. ], batch size: 109, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:09:24,352 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7813, 1.3378, 1.7162, 1.6678, 1.7656, 1.9076, 1.7186, 1.7679], device='cuda:5'), covar=tensor([0.0322, 0.0431, 0.0248, 0.0316, 0.0326, 0.0220, 0.0464, 0.0152], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0188, 0.0176, 0.0177, 0.0193, 0.0153, 0.0192, 0.0151], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:09:40,362 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-01 22:10:08,765 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7345, 2.6467, 2.5032, 3.9823, 2.4000, 3.8715, 1.4756, 2.9472], device='cuda:5'), covar=tensor([0.1372, 0.0778, 0.1136, 0.0168, 0.0168, 0.0377, 0.1773, 0.0724], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0173, 0.0192, 0.0186, 0.0196, 0.0210, 0.0202, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:10:30,927 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:11:01,238 INFO [train.py:904] (5/8) Epoch 24, batch 9750, loss[loss=0.1738, simple_loss=0.2686, pruned_loss=0.03947, over 16806.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2604, pruned_loss=0.03567, over 3069082.88 frames. ], batch size: 124, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:11:17,040 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:12:03,361 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.977e+02 2.392e+02 3.038e+02 5.323e+02, threshold=4.783e+02, percent-clipped=0.0 2023-05-01 22:12:37,405 INFO [train.py:904] (5/8) Epoch 24, batch 9800, loss[loss=0.1703, simple_loss=0.2757, pruned_loss=0.03239, over 16340.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2603, pruned_loss=0.03466, over 3088042.81 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:12:49,020 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:12:50,826 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:13:22,209 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243277.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:13:58,667 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=243293.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:14:21,704 INFO [train.py:904] (5/8) Epoch 24, batch 9850, loss[loss=0.1641, simple_loss=0.2626, pruned_loss=0.03277, over 16194.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2606, pruned_loss=0.03406, over 3086117.80 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:14:28,949 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243306.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:14:38,007 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4660, 3.6381, 3.3624, 3.0684, 3.0791, 3.5604, 3.2393, 3.3333], device='cuda:5'), covar=tensor([0.0878, 0.0825, 0.0491, 0.0402, 0.0989, 0.0564, 0.2002, 0.0638], device='cuda:5'), in_proj_covar=tensor([0.0285, 0.0422, 0.0330, 0.0331, 0.0331, 0.0382, 0.0228, 0.0396], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:5') 2023-05-01 22:15:22,940 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.939e+02 2.274e+02 2.876e+02 6.181e+02, threshold=4.547e+02, percent-clipped=2.0 2023-05-01 22:16:11,638 INFO [train.py:904] (5/8) Epoch 24, batch 9900, loss[loss=0.1692, simple_loss=0.2533, pruned_loss=0.04259, over 12488.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2606, pruned_loss=0.03398, over 3067268.65 frames. ], batch size: 247, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:16:15,214 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=243354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:18:09,544 INFO [train.py:904] (5/8) Epoch 24, batch 9950, loss[loss=0.1637, simple_loss=0.2666, pruned_loss=0.03036, over 16160.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2631, pruned_loss=0.03455, over 3070461.36 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:19:11,954 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:19:24,683 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.110e+02 2.470e+02 3.179e+02 5.008e+02, threshold=4.941e+02, percent-clipped=4.0 2023-05-01 22:20:10,940 INFO [train.py:904] (5/8) Epoch 24, batch 10000, loss[loss=0.1553, simple_loss=0.2547, pruned_loss=0.028, over 16792.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2615, pruned_loss=0.034, over 3094483.10 frames. ], batch size: 83, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:20:26,014 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4704, 2.4035, 2.8390, 3.4069, 3.0667, 3.7848, 2.8374, 3.8738], device='cuda:5'), covar=tensor([0.0185, 0.0479, 0.0384, 0.0226, 0.0312, 0.0169, 0.0390, 0.0147], device='cuda:5'), in_proj_covar=tensor([0.0184, 0.0187, 0.0175, 0.0177, 0.0192, 0.0152, 0.0191, 0.0150], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:20:55,590 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:21:51,126 INFO [train.py:904] (5/8) Epoch 24, batch 10050, loss[loss=0.167, simple_loss=0.2606, pruned_loss=0.03665, over 11870.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2617, pruned_loss=0.03426, over 3086798.74 frames. ], batch size: 250, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:22:03,434 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3522, 4.3500, 4.1918, 3.6894, 4.3004, 1.6186, 4.0782, 3.9308], device='cuda:5'), covar=tensor([0.0110, 0.0114, 0.0218, 0.0270, 0.0105, 0.2862, 0.0143, 0.0250], device='cuda:5'), in_proj_covar=tensor([0.0165, 0.0157, 0.0195, 0.0170, 0.0173, 0.0204, 0.0184, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:22:52,103 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.109e+02 2.585e+02 3.087e+02 5.366e+02, threshold=5.170e+02, percent-clipped=3.0 2023-05-01 22:23:17,292 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8608, 5.1898, 5.3304, 5.0958, 5.1882, 5.6823, 5.1475, 4.8370], device='cuda:5'), covar=tensor([0.0954, 0.1562, 0.1845, 0.1856, 0.2212, 0.0822, 0.1485, 0.2377], device='cuda:5'), in_proj_covar=tensor([0.0388, 0.0577, 0.0638, 0.0474, 0.0624, 0.0660, 0.0494, 0.0627], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 22:23:25,160 INFO [train.py:904] (5/8) Epoch 24, batch 10100, loss[loss=0.1532, simple_loss=0.2426, pruned_loss=0.03187, over 12673.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2619, pruned_loss=0.03411, over 3100451.01 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:24:02,834 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-05-01 22:24:16,128 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243577.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:25:09,206 INFO [train.py:904] (5/8) Epoch 25, batch 0, loss[loss=0.2521, simple_loss=0.3066, pruned_loss=0.09879, over 16888.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3066, pruned_loss=0.09879, over 16888.00 frames. ], batch size: 116, lr: 2.72e-03, grad_scale: 8.0 2023-05-01 22:25:09,206 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 22:25:16,827 INFO [train.py:938] (5/8) Epoch 25, validation: loss=0.1443, simple_loss=0.2477, pruned_loss=0.02048, over 944034.00 frames. 2023-05-01 22:25:16,828 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-01 22:25:48,588 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243625.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:26:03,176 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.517e+02 2.877e+02 3.521e+02 6.755e+02, threshold=5.755e+02, percent-clipped=6.0 2023-05-01 22:26:12,145 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4391, 5.7778, 5.4867, 5.5482, 5.2140, 5.2059, 5.1709, 5.8697], device='cuda:5'), covar=tensor([0.1322, 0.0897, 0.1418, 0.0978, 0.0796, 0.0705, 0.1327, 0.0883], device='cuda:5'), in_proj_covar=tensor([0.0671, 0.0808, 0.0665, 0.0625, 0.0512, 0.0519, 0.0676, 0.0634], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:26:21,458 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243649.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:26:26,920 INFO [train.py:904] (5/8) Epoch 25, batch 50, loss[loss=0.1642, simple_loss=0.2622, pruned_loss=0.03306, over 17272.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2666, pruned_loss=0.04735, over 753855.37 frames. ], batch size: 52, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:27:19,704 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2561, 2.9168, 3.2036, 1.8665, 3.2774, 3.2806, 2.7423, 2.5432], device='cuda:5'), covar=tensor([0.0766, 0.0274, 0.0223, 0.1122, 0.0122, 0.0232, 0.0488, 0.0470], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0106, 0.0095, 0.0135, 0.0079, 0.0123, 0.0124, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-01 22:27:35,806 INFO [train.py:904] (5/8) Epoch 25, batch 100, loss[loss=0.1944, simple_loss=0.2818, pruned_loss=0.05355, over 12331.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2631, pruned_loss=0.04627, over 1311432.42 frames. ], batch size: 247, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:28:22,084 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.579e+02 2.266e+02 2.864e+02 3.484e+02 1.313e+03, threshold=5.728e+02, percent-clipped=5.0 2023-05-01 22:28:35,884 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6258, 3.6274, 4.2915, 2.3707, 3.4213, 2.5264, 4.2158, 3.9234], device='cuda:5'), covar=tensor([0.0247, 0.0996, 0.0431, 0.2051, 0.0775, 0.1075, 0.0505, 0.1169], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0160, 0.0165, 0.0152, 0.0142, 0.0128, 0.0141, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:28:41,042 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6356, 2.7294, 2.4804, 2.5421, 3.0299, 2.7404, 3.2712, 3.1377], device='cuda:5'), covar=tensor([0.0167, 0.0456, 0.0523, 0.0504, 0.0315, 0.0449, 0.0292, 0.0334], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0238, 0.0228, 0.0228, 0.0238, 0.0236, 0.0234, 0.0232], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:28:45,149 INFO [train.py:904] (5/8) Epoch 25, batch 150, loss[loss=0.1807, simple_loss=0.262, pruned_loss=0.04971, over 16012.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2616, pruned_loss=0.04385, over 1752148.18 frames. ], batch size: 35, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:29:26,416 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 22:29:53,826 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8445, 1.8996, 2.4437, 2.7892, 2.6895, 3.2647, 2.3144, 3.3662], device='cuda:5'), covar=tensor([0.0317, 0.0636, 0.0423, 0.0417, 0.0434, 0.0251, 0.0586, 0.0199], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0190, 0.0178, 0.0181, 0.0196, 0.0155, 0.0194, 0.0153], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:29:55,582 INFO [train.py:904] (5/8) Epoch 25, batch 200, loss[loss=0.1599, simple_loss=0.2641, pruned_loss=0.02789, over 17049.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2607, pruned_loss=0.04344, over 2102726.52 frames. ], batch size: 50, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:30:25,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6104, 4.9276, 4.7232, 4.6704, 4.4883, 4.3910, 4.4150, 5.0087], device='cuda:5'), covar=tensor([0.1419, 0.1036, 0.1206, 0.1007, 0.0832, 0.1374, 0.1386, 0.1054], device='cuda:5'), in_proj_covar=tensor([0.0685, 0.0824, 0.0679, 0.0638, 0.0522, 0.0529, 0.0691, 0.0646], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:30:40,585 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.064e+02 2.455e+02 3.172e+02 7.590e+02, threshold=4.909e+02, percent-clipped=1.0 2023-05-01 22:30:57,320 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 22:31:04,179 INFO [train.py:904] (5/8) Epoch 25, batch 250, loss[loss=0.157, simple_loss=0.2548, pruned_loss=0.02965, over 16848.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2579, pruned_loss=0.04235, over 2382047.87 frames. ], batch size: 42, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:31:25,426 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 22:32:14,688 INFO [train.py:904] (5/8) Epoch 25, batch 300, loss[loss=0.1673, simple_loss=0.2488, pruned_loss=0.04287, over 16206.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2569, pruned_loss=0.04238, over 2590748.19 frames. ], batch size: 165, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:32:27,369 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0316, 4.0262, 4.0033, 3.4806, 4.0268, 1.8158, 3.8223, 3.5030], device='cuda:5'), covar=tensor([0.0157, 0.0116, 0.0186, 0.0258, 0.0099, 0.2829, 0.0143, 0.0273], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0162, 0.0200, 0.0175, 0.0177, 0.0209, 0.0189, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:33:00,866 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.231e+02 2.519e+02 2.986e+02 7.202e+02, threshold=5.038e+02, percent-clipped=2.0 2023-05-01 22:33:19,566 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243949.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:33:24,548 INFO [train.py:904] (5/8) Epoch 25, batch 350, loss[loss=0.167, simple_loss=0.2436, pruned_loss=0.04523, over 16780.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2542, pruned_loss=0.04107, over 2754668.57 frames. ], batch size: 124, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:34:25,842 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243997.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:34:38,364 INFO [train.py:904] (5/8) Epoch 25, batch 400, loss[loss=0.1603, simple_loss=0.2423, pruned_loss=0.0391, over 12163.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2535, pruned_loss=0.04077, over 2879234.63 frames. ], batch size: 246, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:34:50,913 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0687, 2.9356, 2.7058, 5.1310, 4.1703, 4.2812, 1.8279, 3.0909], device='cuda:5'), covar=tensor([0.1227, 0.0811, 0.1255, 0.0235, 0.0288, 0.0600, 0.1576, 0.0839], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0191, 0.0200, 0.0213, 0.0204, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:34:54,315 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8721, 4.2193, 3.0522, 2.4300, 2.6966, 2.6222, 4.5384, 3.5356], device='cuda:5'), covar=tensor([0.2822, 0.0576, 0.1851, 0.3001, 0.2879, 0.2155, 0.0345, 0.1399], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0268, 0.0305, 0.0315, 0.0294, 0.0265, 0.0295, 0.0339], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 22:35:02,348 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-01 22:35:07,292 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0469, 4.7750, 5.1179, 5.2938, 5.5208, 4.7861, 5.4880, 5.4693], device='cuda:5'), covar=tensor([0.2389, 0.1540, 0.2121, 0.0958, 0.0705, 0.1007, 0.0629, 0.0799], device='cuda:5'), in_proj_covar=tensor([0.0647, 0.0796, 0.0915, 0.0805, 0.0616, 0.0634, 0.0668, 0.0774], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:35:22,992 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.193e+02 2.643e+02 3.297e+02 6.329e+02, threshold=5.285e+02, percent-clipped=2.0 2023-05-01 22:35:47,043 INFO [train.py:904] (5/8) Epoch 25, batch 450, loss[loss=0.1763, simple_loss=0.2736, pruned_loss=0.03953, over 17031.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2519, pruned_loss=0.03991, over 2976286.08 frames. ], batch size: 55, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:36:55,204 INFO [train.py:904] (5/8) Epoch 25, batch 500, loss[loss=0.173, simple_loss=0.2513, pruned_loss=0.04733, over 16936.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2511, pruned_loss=0.03964, over 3043417.05 frames. ], batch size: 90, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:37:42,045 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 2.128e+02 2.486e+02 2.973e+02 4.659e+02, threshold=4.972e+02, percent-clipped=0.0 2023-05-01 22:38:05,724 INFO [train.py:904] (5/8) Epoch 25, batch 550, loss[loss=0.1914, simple_loss=0.2683, pruned_loss=0.05719, over 16730.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2499, pruned_loss=0.03943, over 3110857.21 frames. ], batch size: 124, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:39:15,781 INFO [train.py:904] (5/8) Epoch 25, batch 600, loss[loss=0.1572, simple_loss=0.2548, pruned_loss=0.02983, over 17123.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.25, pruned_loss=0.03977, over 3156439.66 frames. ], batch size: 49, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:39:48,712 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8807, 4.3720, 2.9893, 2.3658, 2.6384, 2.5495, 4.6252, 3.5265], device='cuda:5'), covar=tensor([0.2689, 0.0516, 0.1860, 0.3002, 0.3022, 0.2159, 0.0350, 0.1484], device='cuda:5'), in_proj_covar=tensor([0.0328, 0.0268, 0.0306, 0.0317, 0.0296, 0.0266, 0.0297, 0.0341], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 22:40:02,988 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.116e+02 2.478e+02 2.900e+02 1.597e+03, threshold=4.957e+02, percent-clipped=3.0 2023-05-01 22:40:24,953 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-05-01 22:40:25,356 INFO [train.py:904] (5/8) Epoch 25, batch 650, loss[loss=0.1754, simple_loss=0.2522, pruned_loss=0.0493, over 16848.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2487, pruned_loss=0.03892, over 3201609.30 frames. ], batch size: 116, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:41:33,628 INFO [train.py:904] (5/8) Epoch 25, batch 700, loss[loss=0.1796, simple_loss=0.2632, pruned_loss=0.04804, over 16299.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2487, pruned_loss=0.0388, over 3224205.80 frames. ], batch size: 165, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:41:54,165 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3067, 3.4447, 3.9394, 2.2060, 3.2142, 2.5346, 3.7532, 3.5818], device='cuda:5'), covar=tensor([0.0313, 0.1020, 0.0486, 0.2035, 0.0808, 0.0960, 0.0630, 0.1202], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0165, 0.0169, 0.0155, 0.0145, 0.0131, 0.0144, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:42:16,538 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9545, 4.9418, 5.3967, 5.3599, 5.3901, 5.0643, 4.9869, 4.8439], device='cuda:5'), covar=tensor([0.0360, 0.0586, 0.0413, 0.0424, 0.0469, 0.0438, 0.0992, 0.0461], device='cuda:5'), in_proj_covar=tensor([0.0420, 0.0468, 0.0456, 0.0420, 0.0501, 0.0480, 0.0557, 0.0382], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 22:42:20,954 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 2.146e+02 2.447e+02 2.875e+02 6.652e+02, threshold=4.894e+02, percent-clipped=4.0 2023-05-01 22:42:42,419 INFO [train.py:904] (5/8) Epoch 25, batch 750, loss[loss=0.1579, simple_loss=0.2323, pruned_loss=0.0418, over 16826.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.25, pruned_loss=0.03933, over 3246996.65 frames. ], batch size: 83, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:42:42,812 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:43:52,160 INFO [train.py:904] (5/8) Epoch 25, batch 800, loss[loss=0.145, simple_loss=0.2437, pruned_loss=0.0232, over 17131.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2494, pruned_loss=0.03851, over 3270812.11 frames. ], batch size: 48, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:43:55,008 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6004, 5.9896, 5.7223, 5.7731, 5.3586, 5.3477, 5.3324, 6.1131], device='cuda:5'), covar=tensor([0.1586, 0.1084, 0.1173, 0.0917, 0.1004, 0.0712, 0.1255, 0.0922], device='cuda:5'), in_proj_covar=tensor([0.0706, 0.0853, 0.0698, 0.0657, 0.0540, 0.0546, 0.0717, 0.0667], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:44:08,255 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:44:39,330 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.216e+02 2.518e+02 3.114e+02 7.971e+02, threshold=5.036e+02, percent-clipped=2.0 2023-05-01 22:45:03,363 INFO [train.py:904] (5/8) Epoch 25, batch 850, loss[loss=0.1345, simple_loss=0.2211, pruned_loss=0.02399, over 17212.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2492, pruned_loss=0.03805, over 3283239.49 frames. ], batch size: 43, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:46:12,677 INFO [train.py:904] (5/8) Epoch 25, batch 900, loss[loss=0.1686, simple_loss=0.2525, pruned_loss=0.04231, over 12220.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2484, pruned_loss=0.03821, over 3285974.73 frames. ], batch size: 246, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:47:00,725 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 1.999e+02 2.403e+02 2.874e+02 4.869e+02, threshold=4.806e+02, percent-clipped=0.0 2023-05-01 22:47:23,397 INFO [train.py:904] (5/8) Epoch 25, batch 950, loss[loss=0.1642, simple_loss=0.2695, pruned_loss=0.02948, over 17130.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2485, pruned_loss=0.03864, over 3298793.98 frames. ], batch size: 47, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:48:33,906 INFO [train.py:904] (5/8) Epoch 25, batch 1000, loss[loss=0.1629, simple_loss=0.2567, pruned_loss=0.03455, over 17142.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2476, pruned_loss=0.03892, over 3296385.17 frames. ], batch size: 48, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:48:44,922 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-05-01 22:49:21,002 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.078e+02 2.468e+02 2.868e+02 8.684e+02, threshold=4.936e+02, percent-clipped=5.0 2023-05-01 22:49:25,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0577, 4.5774, 4.5061, 3.3777, 3.7604, 4.4840, 4.0318, 2.6316], device='cuda:5'), covar=tensor([0.0485, 0.0059, 0.0054, 0.0349, 0.0139, 0.0107, 0.0096, 0.0495], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0136, 0.0100, 0.0111, 0.0096, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 22:49:40,128 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9969, 3.1005, 3.3095, 2.0636, 2.8695, 2.2743, 3.4678, 3.3578], device='cuda:5'), covar=tensor([0.0235, 0.0961, 0.0624, 0.1976, 0.0884, 0.1006, 0.0527, 0.0967], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0165, 0.0169, 0.0155, 0.0146, 0.0131, 0.0145, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:49:42,534 INFO [train.py:904] (5/8) Epoch 25, batch 1050, loss[loss=0.1683, simple_loss=0.2488, pruned_loss=0.04395, over 16496.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.247, pruned_loss=0.03804, over 3311916.30 frames. ], batch size: 75, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:50:45,876 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2765, 5.9918, 6.0972, 5.7001, 5.9538, 6.4213, 5.9188, 5.5680], device='cuda:5'), covar=tensor([0.0982, 0.1890, 0.2642, 0.2033, 0.2261, 0.0867, 0.1630, 0.2039], device='cuda:5'), in_proj_covar=tensor([0.0416, 0.0618, 0.0681, 0.0505, 0.0672, 0.0703, 0.0527, 0.0670], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 22:50:53,099 INFO [train.py:904] (5/8) Epoch 25, batch 1100, loss[loss=0.1754, simple_loss=0.246, pruned_loss=0.05242, over 16937.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2461, pruned_loss=0.03749, over 3312518.48 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:51:01,272 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=244709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:51:13,154 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244717.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:51:22,290 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1813, 3.2068, 3.4474, 2.1851, 2.9574, 2.3972, 3.6522, 3.5204], device='cuda:5'), covar=tensor([0.0245, 0.0935, 0.0620, 0.1998, 0.0902, 0.1032, 0.0491, 0.1003], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0166, 0.0170, 0.0156, 0.0146, 0.0131, 0.0145, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 22:51:40,296 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 2.068e+02 2.379e+02 2.713e+02 6.650e+02, threshold=4.759e+02, percent-clipped=1.0 2023-05-01 22:51:43,769 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8763, 4.6022, 4.8963, 5.0536, 5.2450, 4.6264, 5.2534, 5.2412], device='cuda:5'), covar=tensor([0.1906, 0.1384, 0.1665, 0.0778, 0.0585, 0.1001, 0.0636, 0.0660], device='cuda:5'), in_proj_covar=tensor([0.0670, 0.0826, 0.0950, 0.0836, 0.0637, 0.0655, 0.0687, 0.0800], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:52:02,017 INFO [train.py:904] (5/8) Epoch 25, batch 1150, loss[loss=0.1762, simple_loss=0.253, pruned_loss=0.04968, over 16531.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2457, pruned_loss=0.0369, over 3312767.24 frames. ], batch size: 75, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:52:37,194 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244778.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 22:53:11,759 INFO [train.py:904] (5/8) Epoch 25, batch 1200, loss[loss=0.1743, simple_loss=0.273, pruned_loss=0.03778, over 17069.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2453, pruned_loss=0.03655, over 3311006.84 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:53:30,398 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0204, 4.7311, 5.0351, 5.2205, 5.4027, 4.7151, 5.3932, 5.3880], device='cuda:5'), covar=tensor([0.1871, 0.1399, 0.1635, 0.0773, 0.0544, 0.0982, 0.0542, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0668, 0.0824, 0.0949, 0.0834, 0.0636, 0.0654, 0.0686, 0.0799], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:53:57,425 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.058e+02 2.429e+02 2.907e+02 5.779e+02, threshold=4.859e+02, percent-clipped=1.0 2023-05-01 22:54:19,241 INFO [train.py:904] (5/8) Epoch 25, batch 1250, loss[loss=0.1857, simple_loss=0.2511, pruned_loss=0.06014, over 16754.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2462, pruned_loss=0.03793, over 3313170.36 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:54:22,878 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 22:55:06,328 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244887.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:55:21,528 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3056, 4.1278, 4.3514, 4.4744, 4.5578, 4.1172, 4.3773, 4.5670], device='cuda:5'), covar=tensor([0.1575, 0.1185, 0.1205, 0.0634, 0.0544, 0.1319, 0.2744, 0.0728], device='cuda:5'), in_proj_covar=tensor([0.0669, 0.0826, 0.0951, 0.0836, 0.0638, 0.0657, 0.0687, 0.0801], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:55:26,317 INFO [train.py:904] (5/8) Epoch 25, batch 1300, loss[loss=0.1547, simple_loss=0.2342, pruned_loss=0.03761, over 12115.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.246, pruned_loss=0.03781, over 3310182.05 frames. ], batch size: 247, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:55:44,640 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8870, 4.6336, 4.8930, 5.1009, 5.2782, 4.6710, 5.2883, 5.2911], device='cuda:5'), covar=tensor([0.1956, 0.1416, 0.1872, 0.0790, 0.0636, 0.0937, 0.0642, 0.0652], device='cuda:5'), in_proj_covar=tensor([0.0669, 0.0826, 0.0951, 0.0836, 0.0638, 0.0657, 0.0687, 0.0801], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 22:56:12,140 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.186e+02 2.539e+02 3.062e+02 9.189e+02, threshold=5.078e+02, percent-clipped=4.0 2023-05-01 22:56:28,346 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244948.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:56:35,083 INFO [train.py:904] (5/8) Epoch 25, batch 1350, loss[loss=0.1798, simple_loss=0.2752, pruned_loss=0.04217, over 17030.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2462, pruned_loss=0.03788, over 3321272.25 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:56:36,714 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:57:17,902 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1391, 4.1897, 4.5108, 4.5038, 4.5294, 4.2520, 4.2965, 4.2022], device='cuda:5'), covar=tensor([0.0473, 0.0797, 0.0457, 0.0448, 0.0501, 0.0514, 0.0787, 0.0652], device='cuda:5'), in_proj_covar=tensor([0.0430, 0.0478, 0.0465, 0.0429, 0.0510, 0.0491, 0.0568, 0.0389], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 22:57:43,144 INFO [train.py:904] (5/8) Epoch 25, batch 1400, loss[loss=0.1495, simple_loss=0.2446, pruned_loss=0.02717, over 16626.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2466, pruned_loss=0.03834, over 3316474.32 frames. ], batch size: 62, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:57:52,387 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:58:00,049 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245015.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:58:28,652 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.075e+02 2.428e+02 3.013e+02 6.709e+02, threshold=4.856e+02, percent-clipped=2.0 2023-05-01 22:58:38,868 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-01 22:58:51,283 INFO [train.py:904] (5/8) Epoch 25, batch 1450, loss[loss=0.1607, simple_loss=0.2593, pruned_loss=0.03106, over 17112.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2459, pruned_loss=0.03789, over 3306946.43 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:58:56,922 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:59:18,776 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245073.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 22:59:45,279 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-05-01 23:00:00,090 INFO [train.py:904] (5/8) Epoch 25, batch 1500, loss[loss=0.1814, simple_loss=0.263, pruned_loss=0.04988, over 15610.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2458, pruned_loss=0.03785, over 3315066.92 frames. ], batch size: 190, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:00:15,454 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 23:00:46,574 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.188e+02 2.473e+02 2.927e+02 4.606e+02, threshold=4.945e+02, percent-clipped=0.0 2023-05-01 23:00:47,963 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 23:01:08,947 INFO [train.py:904] (5/8) Epoch 25, batch 1550, loss[loss=0.1664, simple_loss=0.2582, pruned_loss=0.03733, over 17128.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2471, pruned_loss=0.03884, over 3314569.03 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:02:07,392 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2221, 4.0729, 4.2729, 4.4070, 4.4849, 4.0633, 4.2561, 4.4862], device='cuda:5'), covar=tensor([0.1696, 0.1271, 0.1295, 0.0723, 0.0595, 0.1325, 0.2876, 0.0849], device='cuda:5'), in_proj_covar=tensor([0.0679, 0.0838, 0.0966, 0.0847, 0.0646, 0.0664, 0.0697, 0.0814], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:02:19,886 INFO [train.py:904] (5/8) Epoch 25, batch 1600, loss[loss=0.1813, simple_loss=0.2662, pruned_loss=0.04821, over 16856.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2496, pruned_loss=0.03972, over 3318372.66 frames. ], batch size: 96, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:03:06,949 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.282e+02 2.634e+02 3.263e+02 7.681e+02, threshold=5.268e+02, percent-clipped=4.0 2023-05-01 23:03:16,808 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245243.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:03:29,746 INFO [train.py:904] (5/8) Epoch 25, batch 1650, loss[loss=0.166, simple_loss=0.2458, pruned_loss=0.04315, over 16757.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.25, pruned_loss=0.03968, over 3312059.12 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:03:34,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5564, 1.8413, 2.2749, 2.4856, 2.5722, 2.5079, 1.9312, 2.6842], device='cuda:5'), covar=tensor([0.0216, 0.0511, 0.0333, 0.0311, 0.0311, 0.0360, 0.0592, 0.0199], device='cuda:5'), in_proj_covar=tensor([0.0195, 0.0196, 0.0183, 0.0187, 0.0203, 0.0161, 0.0199, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:04:33,555 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4228, 2.8401, 3.0760, 2.1596, 2.7411, 2.1381, 3.0678, 3.1348], device='cuda:5'), covar=tensor([0.0263, 0.0924, 0.0573, 0.1815, 0.0871, 0.1042, 0.0620, 0.0888], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0166, 0.0169, 0.0155, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:04:41,938 INFO [train.py:904] (5/8) Epoch 25, batch 1700, loss[loss=0.1662, simple_loss=0.2469, pruned_loss=0.0427, over 16772.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2516, pruned_loss=0.04009, over 3314974.73 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:04:51,859 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:05:30,606 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.070e+02 2.579e+02 3.059e+02 5.844e+02, threshold=5.158e+02, percent-clipped=3.0 2023-05-01 23:05:46,545 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3604, 5.7058, 5.4842, 5.5572, 5.2205, 5.1428, 5.1450, 5.8644], device='cuda:5'), covar=tensor([0.1453, 0.1084, 0.1112, 0.0921, 0.0904, 0.0819, 0.1226, 0.0998], device='cuda:5'), in_proj_covar=tensor([0.0710, 0.0861, 0.0704, 0.0665, 0.0546, 0.0549, 0.0724, 0.0673], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:05:52,660 INFO [train.py:904] (5/8) Epoch 25, batch 1750, loss[loss=0.1755, simple_loss=0.2702, pruned_loss=0.04046, over 17084.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2523, pruned_loss=0.03968, over 3324525.36 frames. ], batch size: 53, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:06:19,993 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6295, 3.5992, 2.3061, 3.8537, 2.8306, 3.8367, 2.3160, 2.9336], device='cuda:5'), covar=tensor([0.0251, 0.0355, 0.1485, 0.0389, 0.0716, 0.0700, 0.1418, 0.0688], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0173, 0.0180, 0.0222, 0.0206, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:06:20,053 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0654, 3.0924, 3.3163, 2.1740, 2.8608, 2.2511, 3.5729, 3.4832], device='cuda:5'), covar=tensor([0.0215, 0.0922, 0.0630, 0.1901, 0.0891, 0.1079, 0.0508, 0.0951], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0166, 0.0169, 0.0155, 0.0146, 0.0131, 0.0145, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:06:21,002 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245373.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:07:03,547 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-05-01 23:07:03,958 INFO [train.py:904] (5/8) Epoch 25, batch 1800, loss[loss=0.1714, simple_loss=0.2598, pruned_loss=0.0415, over 16219.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.253, pruned_loss=0.03965, over 3333342.36 frames. ], batch size: 165, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:07:29,358 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245421.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:07:51,082 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.502e+02 2.188e+02 2.630e+02 2.978e+02 6.380e+02, threshold=5.260e+02, percent-clipped=2.0 2023-05-01 23:07:56,356 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2009, 3.3180, 3.5448, 2.4063, 3.2775, 3.6500, 3.3556, 2.0772], device='cuda:5'), covar=tensor([0.0553, 0.0151, 0.0067, 0.0430, 0.0129, 0.0109, 0.0109, 0.0503], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0087, 0.0087, 0.0135, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 23:08:08,750 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8919, 2.6398, 2.5186, 4.0616, 3.3685, 3.9962, 1.7669, 2.7752], device='cuda:5'), covar=tensor([0.1289, 0.0671, 0.1179, 0.0174, 0.0138, 0.0497, 0.1411, 0.0877], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0178, 0.0197, 0.0196, 0.0203, 0.0217, 0.0205, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:08:12,919 INFO [train.py:904] (5/8) Epoch 25, batch 1850, loss[loss=0.1813, simple_loss=0.2522, pruned_loss=0.05522, over 16870.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2531, pruned_loss=0.03941, over 3337578.42 frames. ], batch size: 116, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:08:23,519 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:09:23,476 INFO [train.py:904] (5/8) Epoch 25, batch 1900, loss[loss=0.1594, simple_loss=0.2413, pruned_loss=0.03873, over 16791.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2526, pruned_loss=0.03876, over 3337121.45 frames. ], batch size: 83, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:09:49,782 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245521.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:09:50,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1465, 3.2957, 3.4547, 2.3447, 3.2227, 3.5786, 3.2764, 2.1821], device='cuda:5'), covar=tensor([0.0580, 0.0132, 0.0067, 0.0442, 0.0131, 0.0108, 0.0108, 0.0451], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0099, 0.0111, 0.0097, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 23:10:05,000 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2939, 5.7713, 5.9154, 5.6259, 5.7132, 6.2838, 5.7788, 5.4226], device='cuda:5'), covar=tensor([0.0901, 0.1988, 0.2565, 0.2135, 0.2815, 0.0937, 0.1455, 0.2290], device='cuda:5'), in_proj_covar=tensor([0.0426, 0.0634, 0.0694, 0.0513, 0.0686, 0.0716, 0.0535, 0.0683], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 23:10:11,813 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.101e+02 2.496e+02 2.952e+02 1.304e+03, threshold=4.992e+02, percent-clipped=2.0 2023-05-01 23:10:12,267 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2376, 5.1430, 5.0729, 4.5185, 4.7058, 5.1134, 5.0751, 4.6980], device='cuda:5'), covar=tensor([0.0595, 0.0589, 0.0342, 0.0426, 0.1181, 0.0495, 0.0332, 0.0886], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0462, 0.0359, 0.0361, 0.0363, 0.0417, 0.0246, 0.0434], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:10:20,938 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245543.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:25,716 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:33,942 INFO [train.py:904] (5/8) Epoch 25, batch 1950, loss[loss=0.1475, simple_loss=0.2382, pruned_loss=0.02837, over 17196.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2522, pruned_loss=0.03825, over 3331090.95 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:11:00,597 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 23:11:04,607 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0851, 5.0995, 4.8555, 4.3140, 4.9135, 1.7273, 4.7065, 4.6162], device='cuda:5'), covar=tensor([0.0106, 0.0101, 0.0254, 0.0434, 0.0115, 0.3272, 0.0165, 0.0297], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0170, 0.0209, 0.0183, 0.0187, 0.0216, 0.0199, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:11:26,899 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:11:41,396 INFO [train.py:904] (5/8) Epoch 25, batch 2000, loss[loss=0.1876, simple_loss=0.2609, pruned_loss=0.05711, over 16688.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2523, pruned_loss=0.03816, over 3332298.09 frames. ], batch size: 134, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:11:49,075 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245607.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:11:53,168 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245610.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:12:31,551 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.177e+02 2.547e+02 3.013e+02 5.081e+02, threshold=5.093e+02, percent-clipped=1.0 2023-05-01 23:12:50,396 INFO [train.py:904] (5/8) Epoch 25, batch 2050, loss[loss=0.1912, simple_loss=0.2681, pruned_loss=0.05714, over 16804.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2518, pruned_loss=0.03819, over 3329012.83 frames. ], batch size: 96, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:12:57,318 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245658.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:13:33,001 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 23:13:34,884 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245686.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:13:58,108 INFO [train.py:904] (5/8) Epoch 25, batch 2100, loss[loss=0.1873, simple_loss=0.2726, pruned_loss=0.05105, over 15648.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2526, pruned_loss=0.03877, over 3330226.41 frames. ], batch size: 191, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:14:49,018 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.201e+02 2.533e+02 2.997e+02 6.005e+02, threshold=5.066e+02, percent-clipped=2.0 2023-05-01 23:14:59,044 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245747.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:15:08,434 INFO [train.py:904] (5/8) Epoch 25, batch 2150, loss[loss=0.2116, simple_loss=0.2853, pruned_loss=0.06892, over 16699.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2542, pruned_loss=0.03989, over 3321949.30 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:15:14,680 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3764, 2.9977, 3.3438, 1.9066, 3.3902, 3.3776, 2.8716, 2.6189], device='cuda:5'), covar=tensor([0.0809, 0.0285, 0.0239, 0.1103, 0.0158, 0.0261, 0.0482, 0.0460], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0111, 0.0100, 0.0139, 0.0084, 0.0131, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:15:20,243 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 23:15:22,411 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9162, 2.0461, 2.5128, 2.7479, 2.6724, 3.3311, 2.3075, 3.3343], device='cuda:5'), covar=tensor([0.0264, 0.0570, 0.0364, 0.0422, 0.0387, 0.0219, 0.0548, 0.0184], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0189, 0.0205, 0.0162, 0.0201, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:16:00,551 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245792.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:16:15,989 INFO [train.py:904] (5/8) Epoch 25, batch 2200, loss[loss=0.1511, simple_loss=0.2368, pruned_loss=0.0327, over 17003.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2543, pruned_loss=0.04009, over 3319193.83 frames. ], batch size: 41, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:16:34,764 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245816.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:17:06,934 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.208e+02 2.555e+02 2.978e+02 8.197e+02, threshold=5.110e+02, percent-clipped=1.0 2023-05-01 23:17:10,970 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7280, 3.8076, 2.9630, 2.2780, 2.4570, 2.4202, 3.8908, 3.2879], device='cuda:5'), covar=tensor([0.2803, 0.0590, 0.1755, 0.3188, 0.2840, 0.2193, 0.0558, 0.1617], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0273, 0.0311, 0.0320, 0.0303, 0.0271, 0.0302, 0.0348], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 23:17:23,657 INFO [train.py:904] (5/8) Epoch 25, batch 2250, loss[loss=0.1701, simple_loss=0.2711, pruned_loss=0.03456, over 16742.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2566, pruned_loss=0.04114, over 3308770.34 frames. ], batch size: 57, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:17:24,786 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:18:16,632 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245891.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:18:32,179 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245902.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 23:18:32,953 INFO [train.py:904] (5/8) Epoch 25, batch 2300, loss[loss=0.1895, simple_loss=0.2804, pruned_loss=0.04936, over 16683.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2569, pruned_loss=0.04098, over 3313243.92 frames. ], batch size: 62, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:19:24,139 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.182e+02 2.492e+02 2.872e+02 5.312e+02, threshold=4.984e+02, percent-clipped=2.0 2023-05-01 23:19:42,195 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245952.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:19:42,921 INFO [train.py:904] (5/8) Epoch 25, batch 2350, loss[loss=0.1675, simple_loss=0.2643, pruned_loss=0.03533, over 16667.00 frames. ], tot_loss[loss=0.17, simple_loss=0.257, pruned_loss=0.04152, over 3321494.99 frames. ], batch size: 57, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:20:13,975 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 23:20:54,782 INFO [train.py:904] (5/8) Epoch 25, batch 2400, loss[loss=0.17, simple_loss=0.2524, pruned_loss=0.04378, over 16822.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2573, pruned_loss=0.04141, over 3330014.63 frames. ], batch size: 83, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:20:57,978 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2339, 3.9509, 4.4711, 2.4669, 4.6805, 4.7967, 3.4829, 3.6087], device='cuda:5'), covar=tensor([0.0693, 0.0296, 0.0264, 0.1084, 0.0086, 0.0193, 0.0412, 0.0437], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0111, 0.0101, 0.0140, 0.0084, 0.0131, 0.0131, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:21:17,996 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-01 23:21:24,620 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0951, 3.7703, 4.3103, 2.2322, 4.4532, 4.6019, 3.3180, 3.5543], device='cuda:5'), covar=tensor([0.0731, 0.0333, 0.0255, 0.1191, 0.0100, 0.0178, 0.0491, 0.0406], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0111, 0.0101, 0.0141, 0.0084, 0.0131, 0.0131, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:21:46,456 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.068e+02 2.493e+02 3.230e+02 1.350e+03, threshold=4.987e+02, percent-clipped=5.0 2023-05-01 23:21:49,244 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246042.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:22:04,206 INFO [train.py:904] (5/8) Epoch 25, batch 2450, loss[loss=0.1551, simple_loss=0.2505, pruned_loss=0.02984, over 17186.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2569, pruned_loss=0.04093, over 3322268.08 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:23:10,972 INFO [train.py:904] (5/8) Epoch 25, batch 2500, loss[loss=0.1467, simple_loss=0.24, pruned_loss=0.02669, over 16835.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.257, pruned_loss=0.04036, over 3318999.81 frames. ], batch size: 42, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:23:21,567 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246111.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:23:28,051 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:01,583 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.119e+02 2.579e+02 3.197e+02 6.008e+02, threshold=5.158e+02, percent-clipped=3.0 2023-05-01 23:24:12,521 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:18,472 INFO [train.py:904] (5/8) Epoch 25, batch 2550, loss[loss=0.1492, simple_loss=0.2347, pruned_loss=0.03186, over 16809.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2577, pruned_loss=0.04056, over 3315565.59 frames. ], batch size: 39, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:24:35,188 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:45,447 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:25:01,931 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8542, 5.0997, 5.2582, 4.9939, 5.0400, 5.7056, 5.1988, 4.8727], device='cuda:5'), covar=tensor([0.1300, 0.2194, 0.2623, 0.2090, 0.2703, 0.1047, 0.1652, 0.2364], device='cuda:5'), in_proj_covar=tensor([0.0429, 0.0637, 0.0699, 0.0517, 0.0690, 0.0720, 0.0541, 0.0687], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 23:25:17,907 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246196.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:25:21,816 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6400, 6.0128, 5.7968, 5.8492, 5.4250, 5.4511, 5.3931, 6.1682], device='cuda:5'), covar=tensor([0.1554, 0.1119, 0.1116, 0.0916, 0.0901, 0.0736, 0.1272, 0.0880], device='cuda:5'), in_proj_covar=tensor([0.0718, 0.0872, 0.0711, 0.0673, 0.0553, 0.0553, 0.0732, 0.0684], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:25:26,241 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246202.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:25:27,004 INFO [train.py:904] (5/8) Epoch 25, batch 2600, loss[loss=0.1634, simple_loss=0.2466, pruned_loss=0.04011, over 16765.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.257, pruned_loss=0.04037, over 3317505.22 frames. ], batch size: 83, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:25:51,230 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0426, 4.5811, 3.2288, 2.6736, 3.0984, 2.9866, 4.9802, 3.8359], device='cuda:5'), covar=tensor([0.2693, 0.0512, 0.1749, 0.2672, 0.2686, 0.1848, 0.0335, 0.1253], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0274, 0.0312, 0.0321, 0.0303, 0.0272, 0.0302, 0.0349], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-01 23:25:52,193 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246221.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:18,784 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.097e+02 2.510e+02 3.006e+02 6.140e+02, threshold=5.021e+02, percent-clipped=2.0 2023-05-01 23:26:28,341 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:33,158 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246250.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:37,212 INFO [train.py:904] (5/8) Epoch 25, batch 2650, loss[loss=0.1887, simple_loss=0.2664, pruned_loss=0.05546, over 16757.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2577, pruned_loss=0.04015, over 3313952.54 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:26:42,303 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246257.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:27:17,661 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:27:38,630 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6309, 3.6479, 2.3852, 3.9231, 2.9135, 3.8543, 2.3110, 2.9418], device='cuda:5'), covar=tensor([0.0263, 0.0414, 0.1440, 0.0396, 0.0791, 0.0784, 0.1520, 0.0743], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0182, 0.0197, 0.0174, 0.0181, 0.0224, 0.0206, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:27:44,616 INFO [train.py:904] (5/8) Epoch 25, batch 2700, loss[loss=0.176, simple_loss=0.2707, pruned_loss=0.04071, over 16743.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2577, pruned_loss=0.04002, over 3304087.86 frames. ], batch size: 62, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:28:34,093 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 1.968e+02 2.327e+02 2.625e+02 5.550e+02, threshold=4.654e+02, percent-clipped=1.0 2023-05-01 23:28:36,947 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246342.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:28:46,154 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7065, 2.6909, 2.4528, 4.2001, 3.4661, 4.0444, 1.5898, 2.8993], device='cuda:5'), covar=tensor([0.1430, 0.0758, 0.1272, 0.0173, 0.0137, 0.0384, 0.1649, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0179, 0.0198, 0.0197, 0.0205, 0.0218, 0.0206, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:28:52,096 INFO [train.py:904] (5/8) Epoch 25, batch 2750, loss[loss=0.1561, simple_loss=0.2429, pruned_loss=0.03469, over 16849.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2586, pruned_loss=0.03929, over 3319082.82 frames. ], batch size: 102, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:29:32,247 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7914, 1.9203, 2.4055, 2.7164, 2.7765, 2.7656, 1.9850, 2.9680], device='cuda:5'), covar=tensor([0.0193, 0.0523, 0.0393, 0.0312, 0.0292, 0.0296, 0.0590, 0.0194], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0197, 0.0185, 0.0189, 0.0204, 0.0163, 0.0201, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:29:44,293 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246390.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:30:01,771 INFO [train.py:904] (5/8) Epoch 25, batch 2800, loss[loss=0.1394, simple_loss=0.2364, pruned_loss=0.0212, over 17139.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2575, pruned_loss=0.03956, over 3318951.35 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:30:14,575 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9999, 2.1570, 2.2652, 3.5673, 2.1355, 2.4189, 2.2506, 2.2737], device='cuda:5'), covar=tensor([0.1673, 0.4039, 0.3227, 0.0768, 0.4151, 0.2803, 0.4026, 0.3359], device='cuda:5'), in_proj_covar=tensor([0.0417, 0.0466, 0.0384, 0.0338, 0.0445, 0.0534, 0.0438, 0.0546], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:30:54,341 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 1.989e+02 2.330e+02 2.808e+02 5.234e+02, threshold=4.661e+02, percent-clipped=4.0 2023-05-01 23:31:04,274 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:31:10,752 INFO [train.py:904] (5/8) Epoch 25, batch 2850, loss[loss=0.184, simple_loss=0.2748, pruned_loss=0.04659, over 16471.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2572, pruned_loss=0.03914, over 3323330.00 frames. ], batch size: 68, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:31:22,996 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5457, 2.4531, 2.4658, 4.2698, 2.3426, 2.8175, 2.5222, 2.5716], device='cuda:5'), covar=tensor([0.1323, 0.3690, 0.3098, 0.0584, 0.4220, 0.2662, 0.3647, 0.3703], device='cuda:5'), in_proj_covar=tensor([0.0416, 0.0465, 0.0383, 0.0338, 0.0444, 0.0533, 0.0438, 0.0545], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:31:31,947 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:05,537 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 23:32:10,939 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:20,908 INFO [train.py:904] (5/8) Epoch 25, batch 2900, loss[loss=0.1632, simple_loss=0.2586, pruned_loss=0.03394, over 17109.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2566, pruned_loss=0.03946, over 3320850.18 frames. ], batch size: 48, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:32:21,669 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 23:33:02,883 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9911, 4.9622, 5.3902, 5.3900, 5.4329, 5.1049, 5.0677, 4.8683], device='cuda:5'), covar=tensor([0.0360, 0.0676, 0.0454, 0.0398, 0.0474, 0.0425, 0.0948, 0.0453], device='cuda:5'), in_proj_covar=tensor([0.0434, 0.0486, 0.0473, 0.0434, 0.0517, 0.0498, 0.0577, 0.0393], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 23:33:14,958 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.258e+02 2.678e+02 3.242e+02 4.768e+02, threshold=5.357e+02, percent-clipped=1.0 2023-05-01 23:33:25,328 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246547.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:33:32,678 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246552.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:33:33,513 INFO [train.py:904] (5/8) Epoch 25, batch 2950, loss[loss=0.1474, simple_loss=0.2333, pruned_loss=0.03074, over 17011.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2553, pruned_loss=0.03961, over 3317118.14 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:06,285 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:34:08,283 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5144, 3.5970, 3.9321, 2.2381, 3.1923, 2.5725, 3.9862, 3.7935], device='cuda:5'), covar=tensor([0.0227, 0.0877, 0.0503, 0.1945, 0.0825, 0.0930, 0.0490, 0.1043], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0169, 0.0171, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-01 23:34:31,858 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-01 23:34:32,311 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246595.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:34:42,716 INFO [train.py:904] (5/8) Epoch 25, batch 3000, loss[loss=0.1314, simple_loss=0.2153, pruned_loss=0.02371, over 16782.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2552, pruned_loss=0.04005, over 3317782.88 frames. ], batch size: 39, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:42,716 INFO [train.py:929] (5/8) Computing validation loss 2023-05-01 23:34:52,580 INFO [train.py:938] (5/8) Epoch 25, validation: loss=0.1341, simple_loss=0.239, pruned_loss=0.01457, over 944034.00 frames. 2023-05-01 23:34:52,580 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-01 23:35:46,637 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.065e+02 2.466e+02 2.875e+02 5.514e+02, threshold=4.932e+02, percent-clipped=1.0 2023-05-01 23:36:05,622 INFO [train.py:904] (5/8) Epoch 25, batch 3050, loss[loss=0.1937, simple_loss=0.27, pruned_loss=0.0587, over 16438.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2555, pruned_loss=0.04057, over 3318026.78 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:36:38,473 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:36:44,913 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9981, 4.7759, 5.0306, 5.2149, 5.4108, 4.7438, 5.4184, 5.4012], device='cuda:5'), covar=tensor([0.1972, 0.1343, 0.1658, 0.0760, 0.0556, 0.0997, 0.0552, 0.0624], device='cuda:5'), in_proj_covar=tensor([0.0689, 0.0850, 0.0979, 0.0858, 0.0655, 0.0677, 0.0704, 0.0824], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:37:01,175 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-05-01 23:37:13,565 INFO [train.py:904] (5/8) Epoch 25, batch 3100, loss[loss=0.1718, simple_loss=0.2544, pruned_loss=0.04458, over 16339.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2557, pruned_loss=0.04055, over 3322317.80 frames. ], batch size: 165, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:38:05,107 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246738.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:38:08,782 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 2.159e+02 2.563e+02 3.074e+02 4.728e+02, threshold=5.126e+02, percent-clipped=0.0 2023-05-01 23:38:25,016 INFO [train.py:904] (5/8) Epoch 25, batch 3150, loss[loss=0.1568, simple_loss=0.2406, pruned_loss=0.03653, over 16722.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2543, pruned_loss=0.04065, over 3321660.56 frames. ], batch size: 89, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:38:44,302 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:39:34,228 INFO [train.py:904] (5/8) Epoch 25, batch 3200, loss[loss=0.1687, simple_loss=0.258, pruned_loss=0.03973, over 16539.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2532, pruned_loss=0.03994, over 3322678.53 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:39:51,750 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246815.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:40:13,496 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5386, 2.6430, 2.2896, 2.4664, 2.9629, 2.6450, 3.1065, 3.1404], device='cuda:5'), covar=tensor([0.0201, 0.0448, 0.0564, 0.0504, 0.0303, 0.0425, 0.0309, 0.0308], device='cuda:5'), in_proj_covar=tensor([0.0229, 0.0247, 0.0234, 0.0237, 0.0249, 0.0247, 0.0250, 0.0246], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:40:27,531 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.068e+02 2.379e+02 2.953e+02 6.702e+02, threshold=4.759e+02, percent-clipped=1.0 2023-05-01 23:40:42,183 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246852.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:40:42,941 INFO [train.py:904] (5/8) Epoch 25, batch 3250, loss[loss=0.1727, simple_loss=0.2686, pruned_loss=0.03841, over 17033.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2536, pruned_loss=0.03995, over 3313115.96 frames. ], batch size: 55, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:41:16,922 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:41:49,849 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246900.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:41:53,758 INFO [train.py:904] (5/8) Epoch 25, batch 3300, loss[loss=0.1615, simple_loss=0.2538, pruned_loss=0.03464, over 16586.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2532, pruned_loss=0.03985, over 3322235.86 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:42:24,637 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:35,888 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5726, 2.5027, 2.5065, 4.3527, 2.3671, 2.8826, 2.5735, 2.6304], device='cuda:5'), covar=tensor([0.1327, 0.3591, 0.3156, 0.0571, 0.4190, 0.2621, 0.3620, 0.3676], device='cuda:5'), in_proj_covar=tensor([0.0416, 0.0464, 0.0383, 0.0337, 0.0444, 0.0533, 0.0437, 0.0544], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:42:36,969 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:46,635 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.089e+02 2.401e+02 2.822e+02 3.775e+02, threshold=4.802e+02, percent-clipped=0.0 2023-05-01 23:43:02,692 INFO [train.py:904] (5/8) Epoch 25, batch 3350, loss[loss=0.2052, simple_loss=0.2908, pruned_loss=0.05974, over 12331.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2539, pruned_loss=0.03961, over 3327711.82 frames. ], batch size: 247, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:44:01,874 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246995.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:44:13,550 INFO [train.py:904] (5/8) Epoch 25, batch 3400, loss[loss=0.167, simple_loss=0.2435, pruned_loss=0.04525, over 16870.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2531, pruned_loss=0.03943, over 3323127.69 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:44:56,888 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247033.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:45:08,921 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.118e+02 2.456e+02 2.904e+02 5.980e+02, threshold=4.912e+02, percent-clipped=3.0 2023-05-01 23:45:26,159 INFO [train.py:904] (5/8) Epoch 25, batch 3450, loss[loss=0.1708, simple_loss=0.2525, pruned_loss=0.0445, over 11977.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2521, pruned_loss=0.03921, over 3308646.61 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:46:35,594 INFO [train.py:904] (5/8) Epoch 25, batch 3500, loss[loss=0.1776, simple_loss=0.2705, pruned_loss=0.04238, over 17055.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2511, pruned_loss=0.03904, over 3313183.88 frames. ], batch size: 53, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:46:45,079 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9974, 5.0840, 5.4706, 5.4574, 5.4793, 5.1552, 5.0918, 4.9434], device='cuda:5'), covar=tensor([0.0419, 0.0523, 0.0429, 0.0420, 0.0437, 0.0411, 0.0989, 0.0438], device='cuda:5'), in_proj_covar=tensor([0.0437, 0.0491, 0.0477, 0.0438, 0.0521, 0.0502, 0.0581, 0.0397], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 23:47:28,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5713, 3.6319, 2.3633, 3.8465, 2.9122, 3.7659, 2.3423, 2.9155], device='cuda:5'), covar=tensor([0.0262, 0.0415, 0.1375, 0.0305, 0.0741, 0.0844, 0.1346, 0.0688], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0183, 0.0198, 0.0176, 0.0181, 0.0226, 0.0206, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:47:30,018 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.124e+02 2.460e+02 2.748e+02 9.578e+02, threshold=4.920e+02, percent-clipped=1.0 2023-05-01 23:47:45,787 INFO [train.py:904] (5/8) Epoch 25, batch 3550, loss[loss=0.1872, simple_loss=0.2625, pruned_loss=0.05593, over 16777.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2507, pruned_loss=0.03872, over 3321863.14 frames. ], batch size: 134, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:47:47,974 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8964, 4.3025, 4.3192, 3.1757, 3.6190, 4.2421, 3.9103, 2.5179], device='cuda:5'), covar=tensor([0.0505, 0.0065, 0.0051, 0.0375, 0.0138, 0.0118, 0.0094, 0.0525], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0089, 0.0090, 0.0137, 0.0102, 0.0114, 0.0099, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-01 23:48:55,169 INFO [train.py:904] (5/8) Epoch 25, batch 3600, loss[loss=0.1479, simple_loss=0.2419, pruned_loss=0.02696, over 16998.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2498, pruned_loss=0.03867, over 3324319.94 frames. ], batch size: 50, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:49:00,562 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9440, 5.2224, 5.4643, 5.1807, 5.2570, 5.8476, 5.3890, 5.0429], device='cuda:5'), covar=tensor([0.1200, 0.2207, 0.2553, 0.2240, 0.2980, 0.1097, 0.1785, 0.2750], device='cuda:5'), in_proj_covar=tensor([0.0430, 0.0638, 0.0698, 0.0518, 0.0693, 0.0727, 0.0541, 0.0691], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-01 23:49:49,110 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.145e+02 2.569e+02 3.333e+02 7.279e+02, threshold=5.139e+02, percent-clipped=3.0 2023-05-01 23:49:59,360 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 23:50:05,497 INFO [train.py:904] (5/8) Epoch 25, batch 3650, loss[loss=0.1475, simple_loss=0.2243, pruned_loss=0.03535, over 16892.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2494, pruned_loss=0.03891, over 3308474.50 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:50:59,441 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:51:17,360 INFO [train.py:904] (5/8) Epoch 25, batch 3700, loss[loss=0.1608, simple_loss=0.2356, pruned_loss=0.04305, over 16499.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.249, pruned_loss=0.0409, over 3302044.12 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:51:51,826 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4924, 4.4144, 4.4566, 3.4688, 4.4679, 1.7350, 4.1350, 4.0145], device='cuda:5'), covar=tensor([0.0219, 0.0175, 0.0261, 0.0726, 0.0164, 0.3526, 0.0250, 0.0370], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0171, 0.0212, 0.0187, 0.0189, 0.0217, 0.0201, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-01 23:52:01,268 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:52:12,948 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.292e+02 2.615e+02 3.085e+02 5.082e+02, threshold=5.229e+02, percent-clipped=0.0 2023-05-01 23:52:28,855 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 23:52:30,275 INFO [train.py:904] (5/8) Epoch 25, batch 3750, loss[loss=0.1782, simple_loss=0.2607, pruned_loss=0.04786, over 16414.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2499, pruned_loss=0.04234, over 3278607.29 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:53:06,946 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:10,260 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:42,916 INFO [train.py:904] (5/8) Epoch 25, batch 3800, loss[loss=0.1826, simple_loss=0.2772, pruned_loss=0.04395, over 16625.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2517, pruned_loss=0.04384, over 3274213.65 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:54:07,998 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8704, 2.4849, 2.4061, 3.8655, 3.0540, 3.8995, 1.6100, 2.8047], device='cuda:5'), covar=tensor([0.1403, 0.0824, 0.1342, 0.0225, 0.0175, 0.0456, 0.1727, 0.0945], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0179, 0.0197, 0.0198, 0.0206, 0.0219, 0.0207, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-01 23:54:16,057 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247426.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:36,418 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247440.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:38,776 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.181e+02 2.707e+02 3.148e+02 5.686e+02, threshold=5.414e+02, percent-clipped=1.0 2023-05-01 23:54:54,713 INFO [train.py:904] (5/8) Epoch 25, batch 3850, loss[loss=0.1507, simple_loss=0.2264, pruned_loss=0.03747, over 16808.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2519, pruned_loss=0.04458, over 3277834.35 frames. ], batch size: 83, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:55:32,186 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 23:55:43,386 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247487.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:02,900 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:06,095 INFO [train.py:904] (5/8) Epoch 25, batch 3900, loss[loss=0.1623, simple_loss=0.2428, pruned_loss=0.04085, over 16674.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2505, pruned_loss=0.04438, over 3275774.41 frames. ], batch size: 76, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:56:39,815 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:03,109 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.139e+02 2.490e+02 2.933e+02 5.052e+02, threshold=4.980e+02, percent-clipped=0.0 2023-05-01 23:57:04,295 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1129, 4.1613, 4.4217, 4.4038, 4.4569, 4.1907, 4.2224, 4.1190], device='cuda:5'), covar=tensor([0.0448, 0.0674, 0.0401, 0.0417, 0.0512, 0.0470, 0.0731, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0486, 0.0472, 0.0433, 0.0516, 0.0497, 0.0573, 0.0394], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-01 23:57:17,843 INFO [train.py:904] (5/8) Epoch 25, batch 3950, loss[loss=0.1779, simple_loss=0.2675, pruned_loss=0.04411, over 16753.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2493, pruned_loss=0.04432, over 3281237.00 frames. ], batch size: 39, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:57:20,006 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:30,980 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:06,033 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:12,183 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247590.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:30,193 INFO [train.py:904] (5/8) Epoch 25, batch 4000, loss[loss=0.1714, simple_loss=0.2544, pruned_loss=0.04422, over 16932.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2498, pruned_loss=0.0448, over 3284319.85 frames. ], batch size: 90, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:58:49,603 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:21,892 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247638.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:27,476 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 1.992e+02 2.395e+02 2.922e+02 5.946e+02, threshold=4.790e+02, percent-clipped=1.0 2023-05-01 23:59:44,169 INFO [train.py:904] (5/8) Epoch 25, batch 4050, loss[loss=0.1753, simple_loss=0.2612, pruned_loss=0.04465, over 12069.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2504, pruned_loss=0.04414, over 3286999.56 frames. ], batch size: 248, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:00:04,150 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 00:00:26,416 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3836, 3.4251, 2.1189, 3.7044, 2.5540, 3.8211, 2.2928, 2.7514], device='cuda:5'), covar=tensor([0.0267, 0.0368, 0.1620, 0.0152, 0.0820, 0.0352, 0.1475, 0.0764], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0180, 0.0195, 0.0172, 0.0177, 0.0222, 0.0203, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:00:59,197 INFO [train.py:904] (5/8) Epoch 25, batch 4100, loss[loss=0.1946, simple_loss=0.2873, pruned_loss=0.05093, over 16417.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2526, pruned_loss=0.044, over 3272981.16 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:01:16,255 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6387, 2.6364, 1.8124, 2.7284, 2.0232, 2.8042, 2.0920, 2.3360], device='cuda:5'), covar=tensor([0.0284, 0.0347, 0.1348, 0.0218, 0.0692, 0.0413, 0.1220, 0.0598], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0181, 0.0196, 0.0172, 0.0178, 0.0222, 0.0204, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:01:30,788 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9936, 4.9756, 4.7403, 4.1125, 4.9355, 1.8136, 4.7126, 4.2914], device='cuda:5'), covar=tensor([0.0066, 0.0059, 0.0176, 0.0339, 0.0067, 0.3180, 0.0091, 0.0298], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0170, 0.0211, 0.0187, 0.0188, 0.0216, 0.0200, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:01:48,429 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247735.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:01:57,930 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 2.110e+02 2.475e+02 2.947e+02 8.069e+02, threshold=4.949e+02, percent-clipped=1.0 2023-05-02 00:02:14,953 INFO [train.py:904] (5/8) Epoch 25, batch 4150, loss[loss=0.1907, simple_loss=0.2827, pruned_loss=0.0493, over 17039.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2597, pruned_loss=0.04616, over 3261101.31 frames. ], batch size: 53, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:03:01,853 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:05,022 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247784.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:11,364 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1816, 5.4629, 5.2150, 5.2732, 4.9809, 4.8663, 4.8854, 5.6013], device='cuda:5'), covar=tensor([0.1230, 0.0882, 0.1001, 0.0887, 0.0788, 0.0914, 0.1237, 0.0862], device='cuda:5'), in_proj_covar=tensor([0.0715, 0.0864, 0.0712, 0.0669, 0.0548, 0.0551, 0.0729, 0.0677], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:03:27,323 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-05-02 00:03:32,732 INFO [train.py:904] (5/8) Epoch 25, batch 4200, loss[loss=0.2184, simple_loss=0.3031, pruned_loss=0.06682, over 16907.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2667, pruned_loss=0.04753, over 3244113.37 frames. ], batch size: 109, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:03:43,816 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0440, 2.2671, 2.2716, 3.7153, 2.0691, 2.5597, 2.3013, 2.4138], device='cuda:5'), covar=tensor([0.1515, 0.3375, 0.2862, 0.0630, 0.4274, 0.2366, 0.3435, 0.3318], device='cuda:5'), in_proj_covar=tensor([0.0414, 0.0464, 0.0381, 0.0335, 0.0442, 0.0532, 0.0436, 0.0543], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:04:06,457 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6963, 4.0082, 3.9073, 2.1278, 3.2459, 2.4720, 3.9637, 4.1608], device='cuda:5'), covar=tensor([0.0232, 0.0669, 0.0644, 0.2283, 0.0937, 0.1112, 0.0620, 0.0901], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0168, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:04:20,535 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8648, 2.6903, 2.6265, 1.9153, 2.5880, 2.7242, 2.6199, 1.9427], device='cuda:5'), covar=tensor([0.0450, 0.0102, 0.0114, 0.0384, 0.0127, 0.0145, 0.0127, 0.0409], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0100, 0.0111, 0.0097, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 00:04:25,401 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9587, 2.7524, 2.6051, 4.4613, 3.0082, 4.0151, 1.6859, 3.0706], device='cuda:5'), covar=tensor([0.1234, 0.0904, 0.1244, 0.0196, 0.0252, 0.0521, 0.1733, 0.0850], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0197, 0.0205, 0.0217, 0.0206, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:04:30,296 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.155e+02 2.612e+02 3.147e+02 5.112e+02, threshold=5.224e+02, percent-clipped=1.0 2023-05-02 00:04:35,403 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:04:42,970 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 00:04:46,255 INFO [train.py:904] (5/8) Epoch 25, batch 4250, loss[loss=0.1716, simple_loss=0.2673, pruned_loss=0.03798, over 16467.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2702, pruned_loss=0.04739, over 3222584.48 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:52,107 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247856.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:05:28,794 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:06:02,547 INFO [train.py:904] (5/8) Epoch 25, batch 4300, loss[loss=0.1913, simple_loss=0.2855, pruned_loss=0.0486, over 16701.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2713, pruned_loss=0.04673, over 3214191.16 frames. ], batch size: 76, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:06:13,459 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:06:21,402 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7856, 6.0813, 5.7428, 5.8882, 5.5179, 5.3582, 5.4675, 6.2200], device='cuda:5'), covar=tensor([0.1157, 0.0779, 0.1109, 0.0889, 0.0838, 0.0695, 0.1138, 0.0815], device='cuda:5'), in_proj_covar=tensor([0.0713, 0.0861, 0.0708, 0.0667, 0.0546, 0.0549, 0.0725, 0.0675], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:06:25,201 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2201, 4.3309, 4.1136, 3.7867, 3.8582, 4.2320, 3.8861, 3.9605], device='cuda:5'), covar=tensor([0.0560, 0.0444, 0.0285, 0.0303, 0.0789, 0.0407, 0.0915, 0.0649], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0463, 0.0361, 0.0366, 0.0366, 0.0421, 0.0247, 0.0436], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:06:42,004 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0649, 4.0437, 2.8884, 5.0330, 3.3122, 4.8636, 2.9388, 3.4340], device='cuda:5'), covar=tensor([0.0266, 0.0374, 0.1448, 0.0134, 0.0806, 0.0403, 0.1394, 0.0766], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0181, 0.0196, 0.0172, 0.0179, 0.0223, 0.0204, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:07:01,392 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.163e+02 2.477e+02 3.039e+02 5.024e+02, threshold=4.955e+02, percent-clipped=0.0 2023-05-02 00:07:17,420 INFO [train.py:904] (5/8) Epoch 25, batch 4350, loss[loss=0.1763, simple_loss=0.2661, pruned_loss=0.04327, over 17000.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.274, pruned_loss=0.04756, over 3206917.30 frames. ], batch size: 50, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:08:05,108 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247984.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:08:24,029 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247996.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:08:36,586 INFO [train.py:904] (5/8) Epoch 25, batch 4400, loss[loss=0.2167, simple_loss=0.3023, pruned_loss=0.06557, over 15515.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2761, pruned_loss=0.04843, over 3212761.92 frames. ], batch size: 190, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:24,863 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:09:34,457 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 1.997e+02 2.337e+02 2.699e+02 4.708e+02, threshold=4.674e+02, percent-clipped=0.0 2023-05-02 00:09:39,345 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248045.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:09:50,317 INFO [train.py:904] (5/8) Epoch 25, batch 4450, loss[loss=0.2059, simple_loss=0.2976, pruned_loss=0.05707, over 15397.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2793, pruned_loss=0.04978, over 3209389.13 frames. ], batch size: 190, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:56,583 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248057.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,125 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,198 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:35,247 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248083.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:04,516 INFO [train.py:904] (5/8) Epoch 25, batch 4500, loss[loss=0.1947, simple_loss=0.2765, pruned_loss=0.05648, over 17024.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2799, pruned_loss=0.05062, over 3214365.63 frames. ], batch size: 53, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:11:42,771 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 00:11:44,573 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:12:00,008 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:12:01,977 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.873e+02 2.071e+02 2.399e+02 5.095e+02, threshold=4.142e+02, percent-clipped=1.0 2023-05-02 00:12:04,393 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:12:17,509 INFO [train.py:904] (5/8) Epoch 25, batch 4550, loss[loss=0.1826, simple_loss=0.2742, pruned_loss=0.04548, over 16817.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2807, pruned_loss=0.0517, over 3210914.81 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:12:22,886 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:00,139 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248181.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:06,091 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.75 vs. limit=5.0 2023-05-02 00:13:32,549 INFO [train.py:904] (5/8) Epoch 25, batch 4600, loss[loss=0.1887, simple_loss=0.2838, pruned_loss=0.04686, over 16414.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2819, pruned_loss=0.05233, over 3219453.50 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:13:34,234 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248204.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:42,818 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:11,939 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248229.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:25,110 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 00:14:29,189 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.740e+02 1.903e+02 2.276e+02 6.899e+02, threshold=3.806e+02, percent-clipped=0.0 2023-05-02 00:14:36,319 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2018, 1.6236, 1.9677, 2.1341, 2.1956, 2.4302, 1.7492, 2.2591], device='cuda:5'), covar=tensor([0.0259, 0.0534, 0.0301, 0.0398, 0.0326, 0.0250, 0.0572, 0.0177], device='cuda:5'), in_proj_covar=tensor([0.0195, 0.0196, 0.0183, 0.0189, 0.0204, 0.0162, 0.0200, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:14:46,963 INFO [train.py:904] (5/8) Epoch 25, batch 4650, loss[loss=0.1921, simple_loss=0.2746, pruned_loss=0.05484, over 16659.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2821, pruned_loss=0.05301, over 3226960.73 frames. ], batch size: 57, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:14:54,077 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:01,797 INFO [train.py:904] (5/8) Epoch 25, batch 4700, loss[loss=0.1799, simple_loss=0.2626, pruned_loss=0.04861, over 16986.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2795, pruned_loss=0.05194, over 3228123.83 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:16:38,384 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7088, 3.6939, 2.3316, 4.4068, 2.8547, 4.2726, 2.6236, 3.1017], device='cuda:5'), covar=tensor([0.0292, 0.0392, 0.1768, 0.0148, 0.0798, 0.0513, 0.1428, 0.0797], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0171, 0.0178, 0.0222, 0.0204, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:16:44,306 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:56,011 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248340.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:57,977 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 1.897e+02 2.230e+02 2.532e+02 4.236e+02, threshold=4.460e+02, percent-clipped=2.0 2023-05-02 00:17:13,842 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:17:14,736 INFO [train.py:904] (5/8) Epoch 25, batch 4750, loss[loss=0.1512, simple_loss=0.2449, pruned_loss=0.02875, over 16430.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2754, pruned_loss=0.04983, over 3225080.94 frames. ], batch size: 146, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:17:59,900 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:04,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9536, 2.7253, 2.8669, 2.0781, 2.7077, 2.1799, 2.7251, 2.8595], device='cuda:5'), covar=tensor([0.0283, 0.0794, 0.0534, 0.1849, 0.0824, 0.0870, 0.0632, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0168, 0.0170, 0.0155, 0.0147, 0.0132, 0.0145, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 00:18:13,870 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:15,330 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2886, 3.4669, 3.7197, 2.0881, 3.1779, 2.5024, 3.6907, 3.6884], device='cuda:5'), covar=tensor([0.0231, 0.0798, 0.0551, 0.2137, 0.0808, 0.0907, 0.0633, 0.1041], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0145, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 00:18:29,685 INFO [train.py:904] (5/8) Epoch 25, batch 4800, loss[loss=0.2041, simple_loss=0.2816, pruned_loss=0.06329, over 11997.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2718, pruned_loss=0.04794, over 3214115.34 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:19:22,193 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:25,571 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248440.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:27,484 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 1.792e+02 2.090e+02 2.399e+02 3.533e+02, threshold=4.179e+02, percent-clipped=0.0 2023-05-02 00:19:31,592 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:43,842 INFO [train.py:904] (5/8) Epoch 25, batch 4850, loss[loss=0.2385, simple_loss=0.3102, pruned_loss=0.0834, over 12128.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2726, pruned_loss=0.04675, over 3209675.77 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:20:37,397 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:20:37,528 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6809, 4.7927, 4.5793, 4.2060, 4.2155, 4.6792, 4.4779, 4.3828], device='cuda:5'), covar=tensor([0.0629, 0.0441, 0.0336, 0.0345, 0.1055, 0.0535, 0.0486, 0.0688], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0452, 0.0353, 0.0358, 0.0357, 0.0411, 0.0242, 0.0425], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:20:58,439 INFO [train.py:904] (5/8) Epoch 25, batch 4900, loss[loss=0.1865, simple_loss=0.2845, pruned_loss=0.04425, over 16696.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2707, pruned_loss=0.04484, over 3209764.12 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:22:03,446 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 1.924e+02 2.287e+02 2.776e+02 8.047e+02, threshold=4.573e+02, percent-clipped=3.0 2023-05-02 00:22:06,095 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6594, 3.7330, 2.2698, 4.3838, 2.7474, 4.2301, 2.4600, 2.9833], device='cuda:5'), covar=tensor([0.0313, 0.0401, 0.1794, 0.0175, 0.0921, 0.0621, 0.1643, 0.0909], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0169, 0.0177, 0.0220, 0.0203, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:22:20,192 INFO [train.py:904] (5/8) Epoch 25, batch 4950, loss[loss=0.1927, simple_loss=0.2858, pruned_loss=0.04983, over 16819.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2702, pruned_loss=0.04409, over 3216408.68 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:31,272 INFO [train.py:904] (5/8) Epoch 25, batch 5000, loss[loss=0.1842, simple_loss=0.2845, pruned_loss=0.04193, over 16491.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2722, pruned_loss=0.04456, over 3205958.05 frames. ], batch size: 146, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:40,244 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7660, 2.6476, 2.1256, 2.5760, 3.0354, 2.7436, 3.1626, 3.2573], device='cuda:5'), covar=tensor([0.0091, 0.0480, 0.0647, 0.0477, 0.0294, 0.0449, 0.0226, 0.0296], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0242, 0.0230, 0.0233, 0.0244, 0.0242, 0.0243, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:24:09,190 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9802, 3.9046, 4.0352, 4.1756, 4.2683, 3.8750, 4.2321, 4.2908], device='cuda:5'), covar=tensor([0.1581, 0.1129, 0.1352, 0.0652, 0.0514, 0.1424, 0.0817, 0.0677], device='cuda:5'), in_proj_covar=tensor([0.0657, 0.0808, 0.0930, 0.0817, 0.0625, 0.0643, 0.0671, 0.0784], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:24:22,318 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9927, 5.0127, 5.3074, 5.3132, 5.3385, 5.0181, 4.9394, 4.7268], device='cuda:5'), covar=tensor([0.0304, 0.0511, 0.0389, 0.0361, 0.0502, 0.0363, 0.1053, 0.0477], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0464, 0.0455, 0.0415, 0.0499, 0.0475, 0.0553, 0.0378], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 00:24:25,332 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248640.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:27,929 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.126e+02 2.416e+02 2.928e+02 6.487e+02, threshold=4.832e+02, percent-clipped=1.0 2023-05-02 00:24:42,936 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:43,841 INFO [train.py:904] (5/8) Epoch 25, batch 5050, loss[loss=0.1617, simple_loss=0.2526, pruned_loss=0.03545, over 17128.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2723, pruned_loss=0.04411, over 3234272.32 frames. ], batch size: 47, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:25:35,138 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:35,224 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:39,670 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:53,428 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:57,289 INFO [train.py:904] (5/8) Epoch 25, batch 5100, loss[loss=0.1777, simple_loss=0.2698, pruned_loss=0.04283, over 16750.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2706, pruned_loss=0.04362, over 3221224.29 frames. ], batch size: 124, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:26:15,828 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:50,106 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:52,087 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248739.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:56,589 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.974e+02 2.293e+02 2.738e+02 8.128e+02, threshold=4.585e+02, percent-clipped=1.0 2023-05-02 00:27:10,591 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6669, 4.6552, 4.4988, 3.6717, 4.5599, 1.6948, 4.3211, 4.1862], device='cuda:5'), covar=tensor([0.0082, 0.0089, 0.0174, 0.0425, 0.0106, 0.2957, 0.0142, 0.0279], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0167, 0.0206, 0.0184, 0.0184, 0.0213, 0.0196, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:27:12,746 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:27:13,432 INFO [train.py:904] (5/8) Epoch 25, batch 5150, loss[loss=0.1794, simple_loss=0.269, pruned_loss=0.0449, over 12137.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2706, pruned_loss=0.04282, over 3224090.70 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:27:47,478 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:28:03,431 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:28:28,608 INFO [train.py:904] (5/8) Epoch 25, batch 5200, loss[loss=0.1723, simple_loss=0.2629, pruned_loss=0.04087, over 16804.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2693, pruned_loss=0.0424, over 3206482.76 frames. ], batch size: 124, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:29:25,195 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.065e+02 2.411e+02 2.935e+02 6.100e+02, threshold=4.822e+02, percent-clipped=2.0 2023-05-02 00:29:27,560 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8495, 5.1252, 4.8617, 4.9462, 4.6756, 4.6617, 4.5504, 5.2174], device='cuda:5'), covar=tensor([0.1355, 0.0848, 0.1044, 0.0858, 0.0828, 0.1026, 0.1202, 0.0895], device='cuda:5'), in_proj_covar=tensor([0.0696, 0.0836, 0.0692, 0.0646, 0.0534, 0.0534, 0.0704, 0.0656], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:29:41,117 INFO [train.py:904] (5/8) Epoch 25, batch 5250, loss[loss=0.202, simple_loss=0.2792, pruned_loss=0.06237, over 12215.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2666, pruned_loss=0.04215, over 3211785.05 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:29:49,078 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3438, 5.3768, 5.7189, 5.7286, 5.7318, 5.4157, 5.3255, 5.0731], device='cuda:5'), covar=tensor([0.0291, 0.0461, 0.0327, 0.0308, 0.0468, 0.0336, 0.0940, 0.0385], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0465, 0.0455, 0.0416, 0.0499, 0.0476, 0.0553, 0.0378], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 00:30:53,692 INFO [train.py:904] (5/8) Epoch 25, batch 5300, loss[loss=0.1636, simple_loss=0.2569, pruned_loss=0.03515, over 16206.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2632, pruned_loss=0.04128, over 3210374.37 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:31:51,349 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 1.862e+02 2.129e+02 2.555e+02 4.739e+02, threshold=4.259e+02, percent-clipped=0.0 2023-05-02 00:32:08,012 INFO [train.py:904] (5/8) Epoch 25, batch 5350, loss[loss=0.1671, simple_loss=0.2587, pruned_loss=0.03773, over 17108.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.262, pruned_loss=0.04125, over 3200030.43 frames. ], batch size: 47, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:33:01,213 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248988.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:33:22,814 INFO [train.py:904] (5/8) Epoch 25, batch 5400, loss[loss=0.2042, simple_loss=0.2897, pruned_loss=0.05931, over 12237.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2642, pruned_loss=0.04204, over 3182289.19 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:34:11,347 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:17,362 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249039.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:20,473 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 1.925e+02 2.196e+02 2.578e+02 3.732e+02, threshold=4.392e+02, percent-clipped=0.0 2023-05-02 00:34:29,107 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249047.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:40,347 INFO [train.py:904] (5/8) Epoch 25, batch 5450, loss[loss=0.2092, simple_loss=0.3027, pruned_loss=0.05783, over 16432.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2676, pruned_loss=0.04348, over 3185722.63 frames. ], batch size: 146, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:34:42,730 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9925, 3.1062, 3.1740, 2.1067, 2.9421, 3.2129, 3.0157, 1.9407], device='cuda:5'), covar=tensor([0.0544, 0.0079, 0.0074, 0.0451, 0.0131, 0.0120, 0.0123, 0.0499], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0100, 0.0111, 0.0097, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 00:35:09,497 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:34,926 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249087.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:58,374 INFO [train.py:904] (5/8) Epoch 25, batch 5500, loss[loss=0.2056, simple_loss=0.2941, pruned_loss=0.05857, over 16316.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2744, pruned_loss=0.04701, over 3178283.62 frames. ], batch size: 146, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:37:00,872 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 3.023e+02 3.674e+02 4.693e+02 9.094e+02, threshold=7.348e+02, percent-clipped=33.0 2023-05-02 00:37:17,263 INFO [train.py:904] (5/8) Epoch 25, batch 5550, loss[loss=0.1847, simple_loss=0.2776, pruned_loss=0.04592, over 17186.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2817, pruned_loss=0.05223, over 3126626.51 frames. ], batch size: 46, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:37:40,590 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 00:37:46,914 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 00:38:26,969 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5951, 2.5196, 2.3727, 3.8028, 2.6802, 3.8167, 1.3747, 2.7200], device='cuda:5'), covar=tensor([0.1464, 0.0856, 0.1351, 0.0193, 0.0232, 0.0471, 0.1875, 0.0923], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0205, 0.0215, 0.0206, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:38:41,406 INFO [train.py:904] (5/8) Epoch 25, batch 5600, loss[loss=0.1911, simple_loss=0.2717, pruned_loss=0.05522, over 16832.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2864, pruned_loss=0.05644, over 3091856.57 frames. ], batch size: 42, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:39:02,168 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:39:21,243 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9667, 4.0633, 4.3075, 4.2839, 4.2994, 4.0616, 4.0618, 4.0761], device='cuda:5'), covar=tensor([0.0416, 0.0698, 0.0432, 0.0465, 0.0457, 0.0509, 0.0890, 0.0566], device='cuda:5'), in_proj_covar=tensor([0.0416, 0.0469, 0.0457, 0.0417, 0.0500, 0.0478, 0.0558, 0.0381], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 00:39:47,611 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 3.336e+02 4.056e+02 4.975e+02 8.511e+02, threshold=8.112e+02, percent-clipped=2.0 2023-05-02 00:39:54,831 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9416, 5.2124, 4.9812, 5.0039, 4.7345, 4.6666, 4.6550, 5.3293], device='cuda:5'), covar=tensor([0.1187, 0.0903, 0.1049, 0.0989, 0.0863, 0.1069, 0.1213, 0.0921], device='cuda:5'), in_proj_covar=tensor([0.0691, 0.0834, 0.0690, 0.0644, 0.0530, 0.0531, 0.0702, 0.0653], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:40:04,586 INFO [train.py:904] (5/8) Epoch 25, batch 5650, loss[loss=0.2735, simple_loss=0.3281, pruned_loss=0.1094, over 11365.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2915, pruned_loss=0.06049, over 3056651.61 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:40:41,206 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:40:43,005 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1949, 4.2873, 4.0960, 3.8135, 3.8041, 4.1963, 3.9470, 3.9683], device='cuda:5'), covar=tensor([0.0616, 0.0543, 0.0304, 0.0296, 0.0810, 0.0512, 0.0749, 0.0641], device='cuda:5'), in_proj_covar=tensor([0.0300, 0.0451, 0.0351, 0.0356, 0.0356, 0.0412, 0.0240, 0.0424], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:41:22,856 INFO [train.py:904] (5/8) Epoch 25, batch 5700, loss[loss=0.1795, simple_loss=0.2808, pruned_loss=0.03908, over 16743.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.292, pruned_loss=0.06121, over 3053637.04 frames. ], batch size: 83, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:41:42,441 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7169, 2.3342, 2.2990, 3.2024, 1.9807, 3.4804, 1.6275, 2.5835], device='cuda:5'), covar=tensor([0.1518, 0.0947, 0.1403, 0.0235, 0.0202, 0.0453, 0.1805, 0.0996], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0205, 0.0216, 0.0206, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:42:13,557 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 00:42:25,335 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.887e+02 2.916e+02 3.392e+02 3.948e+02 9.464e+02, threshold=6.785e+02, percent-clipped=1.0 2023-05-02 00:42:35,266 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:42:43,871 INFO [train.py:904] (5/8) Epoch 25, batch 5750, loss[loss=0.1864, simple_loss=0.2804, pruned_loss=0.04621, over 16507.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2943, pruned_loss=0.06198, over 3073879.11 frames. ], batch size: 75, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:43:13,846 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:43:53,993 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249395.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:44:07,479 INFO [train.py:904] (5/8) Epoch 25, batch 5800, loss[loss=0.1771, simple_loss=0.265, pruned_loss=0.04462, over 16637.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2931, pruned_loss=0.06045, over 3073435.89 frames. ], batch size: 57, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:44:24,847 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5584, 3.4541, 2.5963, 2.2963, 2.4104, 2.3694, 3.7244, 3.2662], device='cuda:5'), covar=tensor([0.3095, 0.0799, 0.2100, 0.2800, 0.2429, 0.2191, 0.0529, 0.1400], device='cuda:5'), in_proj_covar=tensor([0.0331, 0.0272, 0.0310, 0.0319, 0.0302, 0.0268, 0.0301, 0.0344], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 00:44:32,818 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:45:09,431 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.976e+02 3.356e+02 4.382e+02 8.536e+02, threshold=6.712e+02, percent-clipped=3.0 2023-05-02 00:45:26,298 INFO [train.py:904] (5/8) Epoch 25, batch 5850, loss[loss=0.209, simple_loss=0.2966, pruned_loss=0.06068, over 16256.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.292, pruned_loss=0.0593, over 3079517.11 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:45:38,545 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9649, 5.3379, 5.4679, 5.2519, 5.2939, 5.8250, 5.2582, 5.0473], device='cuda:5'), covar=tensor([0.0970, 0.1560, 0.2079, 0.1726, 0.2218, 0.0822, 0.1517, 0.2262], device='cuda:5'), in_proj_covar=tensor([0.0419, 0.0617, 0.0675, 0.0502, 0.0667, 0.0705, 0.0524, 0.0672], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 00:46:03,311 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-02 00:46:46,949 INFO [train.py:904] (5/8) Epoch 25, batch 5900, loss[loss=0.1921, simple_loss=0.2783, pruned_loss=0.05295, over 15413.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2917, pruned_loss=0.05954, over 3083821.78 frames. ], batch size: 191, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:47:10,880 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0403, 2.3204, 2.3571, 2.6912, 1.9126, 3.1598, 1.9084, 2.7499], device='cuda:5'), covar=tensor([0.1114, 0.0651, 0.1042, 0.0162, 0.0105, 0.0320, 0.1351, 0.0662], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0178, 0.0197, 0.0195, 0.0206, 0.0216, 0.0206, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:47:52,236 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.624e+02 3.247e+02 4.027e+02 8.250e+02, threshold=6.495e+02, percent-clipped=2.0 2023-05-02 00:48:08,076 INFO [train.py:904] (5/8) Epoch 25, batch 5950, loss[loss=0.1756, simple_loss=0.2712, pruned_loss=0.04002, over 16792.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2926, pruned_loss=0.05835, over 3082338.26 frames. ], batch size: 102, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:48:37,013 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:49:28,154 INFO [train.py:904] (5/8) Epoch 25, batch 6000, loss[loss=0.2322, simple_loss=0.3033, pruned_loss=0.08052, over 11710.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2914, pruned_loss=0.05762, over 3093823.18 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:49:28,154 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 00:49:38,599 INFO [train.py:938] (5/8) Epoch 25, validation: loss=0.1487, simple_loss=0.2613, pruned_loss=0.01811, over 944034.00 frames. 2023-05-02 00:49:38,600 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 00:50:36,539 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.903e+02 3.391e+02 3.955e+02 5.249e+02, threshold=6.782e+02, percent-clipped=0.0 2023-05-02 00:50:54,729 INFO [train.py:904] (5/8) Epoch 25, batch 6050, loss[loss=0.1813, simple_loss=0.2688, pruned_loss=0.04692, over 16295.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.29, pruned_loss=0.05721, over 3103982.32 frames. ], batch size: 35, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:51:44,121 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 00:52:08,858 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-05-02 00:52:12,401 INFO [train.py:904] (5/8) Epoch 25, batch 6100, loss[loss=0.184, simple_loss=0.2897, pruned_loss=0.03916, over 16770.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2888, pruned_loss=0.0561, over 3107929.65 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:52:27,257 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3784, 1.7330, 2.0552, 2.3191, 2.4361, 2.6176, 1.8075, 2.4987], device='cuda:5'), covar=tensor([0.0253, 0.0530, 0.0335, 0.0354, 0.0322, 0.0226, 0.0576, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0194, 0.0182, 0.0186, 0.0201, 0.0160, 0.0199, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 00:53:15,559 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.711e+02 3.338e+02 3.699e+02 9.166e+02, threshold=6.676e+02, percent-clipped=2.0 2023-05-02 00:53:29,954 INFO [train.py:904] (5/8) Epoch 25, batch 6150, loss[loss=0.1798, simple_loss=0.276, pruned_loss=0.04176, over 16895.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2868, pruned_loss=0.05541, over 3111287.59 frames. ], batch size: 96, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:54:51,561 INFO [train.py:904] (5/8) Epoch 25, batch 6200, loss[loss=0.1671, simple_loss=0.2546, pruned_loss=0.03983, over 16289.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2852, pruned_loss=0.0552, over 3112446.57 frames. ], batch size: 35, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:55:10,525 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:55:55,233 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.970e+02 2.750e+02 3.284e+02 4.072e+02 6.772e+02, threshold=6.569e+02, percent-clipped=1.0 2023-05-02 00:56:10,145 INFO [train.py:904] (5/8) Epoch 25, batch 6250, loss[loss=0.1988, simple_loss=0.2955, pruned_loss=0.05105, over 16799.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2856, pruned_loss=0.05572, over 3101444.46 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:56:38,959 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249871.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:56:40,186 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7231, 3.5094, 4.0339, 2.0700, 4.1553, 4.1912, 3.1233, 3.1550], device='cuda:5'), covar=tensor([0.0761, 0.0335, 0.0191, 0.1160, 0.0079, 0.0158, 0.0431, 0.0446], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0140, 0.0085, 0.0130, 0.0130, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 00:56:44,493 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:57:26,275 INFO [train.py:904] (5/8) Epoch 25, batch 6300, loss[loss=0.1757, simple_loss=0.2657, pruned_loss=0.04289, over 16559.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2861, pruned_loss=0.05488, over 3121148.31 frames. ], batch size: 57, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:57:52,595 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249919.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:58:29,095 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.746e+02 3.246e+02 3.923e+02 7.422e+02, threshold=6.492e+02, percent-clipped=1.0 2023-05-02 00:58:45,049 INFO [train.py:904] (5/8) Epoch 25, batch 6350, loss[loss=0.2029, simple_loss=0.2849, pruned_loss=0.06047, over 16653.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2871, pruned_loss=0.05645, over 3089018.07 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:58:48,596 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249955.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 00:58:54,532 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 00:58:59,134 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9496, 4.0096, 4.3010, 4.2725, 4.2947, 4.0363, 4.0533, 4.0569], device='cuda:5'), covar=tensor([0.0416, 0.0828, 0.0533, 0.0556, 0.0594, 0.0613, 0.0978, 0.0647], device='cuda:5'), in_proj_covar=tensor([0.0418, 0.0471, 0.0458, 0.0418, 0.0503, 0.0480, 0.0558, 0.0383], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 00:59:01,730 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 00:59:20,620 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249976.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:00:04,135 INFO [train.py:904] (5/8) Epoch 25, batch 6400, loss[loss=0.251, simple_loss=0.3216, pruned_loss=0.09022, over 11414.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2866, pruned_loss=0.05693, over 3095549.90 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:00:13,462 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 01:00:23,782 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250016.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 01:00:46,704 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7013, 3.6107, 4.0779, 1.9856, 4.1998, 4.2479, 3.1948, 3.1520], device='cuda:5'), covar=tensor([0.0780, 0.0281, 0.0176, 0.1256, 0.0074, 0.0174, 0.0397, 0.0473], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0101, 0.0141, 0.0085, 0.0131, 0.0131, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:00:56,515 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:01:01,997 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9549, 4.2075, 4.0196, 4.0818, 3.7419, 3.8293, 3.8996, 4.2127], device='cuda:5'), covar=tensor([0.1160, 0.0922, 0.1082, 0.0912, 0.0891, 0.1619, 0.1020, 0.1017], device='cuda:5'), in_proj_covar=tensor([0.0694, 0.0838, 0.0692, 0.0647, 0.0533, 0.0534, 0.0705, 0.0656], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:01:05,943 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 2.850e+02 3.461e+02 4.247e+02 7.668e+02, threshold=6.921e+02, percent-clipped=1.0 2023-05-02 01:01:20,201 INFO [train.py:904] (5/8) Epoch 25, batch 6450, loss[loss=0.2334, simple_loss=0.3198, pruned_loss=0.07352, over 16417.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2867, pruned_loss=0.0563, over 3099272.88 frames. ], batch size: 35, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:02:36,761 INFO [train.py:904] (5/8) Epoch 25, batch 6500, loss[loss=0.2427, simple_loss=0.3105, pruned_loss=0.08746, over 11271.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2843, pruned_loss=0.05551, over 3109632.00 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:03:30,073 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 01:03:40,565 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.844e+02 2.643e+02 3.230e+02 3.830e+02 1.038e+03, threshold=6.461e+02, percent-clipped=2.0 2023-05-02 01:03:52,689 INFO [train.py:904] (5/8) Epoch 25, batch 6550, loss[loss=0.197, simple_loss=0.3018, pruned_loss=0.04614, over 16387.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2869, pruned_loss=0.05584, over 3113084.42 frames. ], batch size: 146, lr: 2.69e-03, grad_scale: 4.0 2023-05-02 01:04:01,388 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2409, 2.3882, 2.2993, 4.0341, 2.2254, 2.7591, 2.4190, 2.5260], device='cuda:5'), covar=tensor([0.1375, 0.3612, 0.2996, 0.0517, 0.4034, 0.2405, 0.3704, 0.3249], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0462, 0.0380, 0.0332, 0.0442, 0.0530, 0.0433, 0.0541], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:04:15,558 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 01:04:19,109 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:04:30,994 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-02 01:04:40,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9725, 2.8836, 2.4196, 4.5693, 3.1794, 4.0337, 1.7542, 3.0246], device='cuda:5'), covar=tensor([0.1269, 0.0758, 0.1402, 0.0146, 0.0281, 0.0494, 0.1601, 0.0876], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0195, 0.0205, 0.0215, 0.0206, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:05:05,606 INFO [train.py:904] (5/8) Epoch 25, batch 6600, loss[loss=0.2535, simple_loss=0.3216, pruned_loss=0.09272, over 11539.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2889, pruned_loss=0.05606, over 3118217.26 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:05:50,342 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 01:06:08,500 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 2.823e+02 3.325e+02 3.947e+02 6.947e+02, threshold=6.650e+02, percent-clipped=1.0 2023-05-02 01:06:11,853 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 01:06:21,910 INFO [train.py:904] (5/8) Epoch 25, batch 6650, loss[loss=0.1787, simple_loss=0.2711, pruned_loss=0.0431, over 16893.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2892, pruned_loss=0.05706, over 3100768.86 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:06:25,570 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0881, 3.2203, 3.4286, 2.0689, 2.9801, 2.1782, 3.5332, 3.5365], device='cuda:5'), covar=tensor([0.0256, 0.0889, 0.0654, 0.2202, 0.0924, 0.1104, 0.0629, 0.1073], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0167, 0.0169, 0.0154, 0.0147, 0.0131, 0.0145, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:07:11,077 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3688, 2.4715, 2.3696, 4.0931, 2.4062, 2.8282, 2.5057, 2.5622], device='cuda:5'), covar=tensor([0.1345, 0.3324, 0.3022, 0.0596, 0.3897, 0.2265, 0.3259, 0.3268], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0462, 0.0379, 0.0332, 0.0442, 0.0530, 0.0434, 0.0540], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:07:26,257 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-05-02 01:07:37,545 INFO [train.py:904] (5/8) Epoch 25, batch 6700, loss[loss=0.2358, simple_loss=0.3057, pruned_loss=0.08296, over 11433.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2878, pruned_loss=0.05702, over 3110228.58 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:07:39,792 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4761, 3.4954, 3.4535, 2.6689, 3.2532, 2.1528, 3.0732, 2.7752], device='cuda:5'), covar=tensor([0.0160, 0.0141, 0.0207, 0.0239, 0.0110, 0.2299, 0.0140, 0.0272], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0167, 0.0206, 0.0183, 0.0183, 0.0213, 0.0195, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:07:50,666 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250311.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 01:08:02,203 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-05-02 01:08:13,436 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-05-02 01:08:23,208 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:08:41,138 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.721e+02 3.128e+02 3.707e+02 8.776e+02, threshold=6.256e+02, percent-clipped=1.0 2023-05-02 01:08:51,975 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1576, 5.4517, 5.1690, 5.2186, 4.9524, 4.9214, 4.8258, 5.5802], device='cuda:5'), covar=tensor([0.1282, 0.0865, 0.1096, 0.0892, 0.0848, 0.0908, 0.1187, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0698, 0.0844, 0.0698, 0.0652, 0.0536, 0.0539, 0.0710, 0.0659], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:08:54,001 INFO [train.py:904] (5/8) Epoch 25, batch 6750, loss[loss=0.1974, simple_loss=0.2864, pruned_loss=0.0542, over 16538.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2872, pruned_loss=0.05669, over 3119354.68 frames. ], batch size: 75, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:09:33,532 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9936, 2.2401, 2.3363, 2.7657, 1.9899, 3.1883, 1.7839, 2.7332], device='cuda:5'), covar=tensor([0.1217, 0.0705, 0.1132, 0.0178, 0.0130, 0.0349, 0.1516, 0.0738], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0179, 0.0197, 0.0196, 0.0206, 0.0217, 0.0207, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:09:43,550 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-02 01:10:10,579 INFO [train.py:904] (5/8) Epoch 25, batch 6800, loss[loss=0.2174, simple_loss=0.2987, pruned_loss=0.06805, over 15205.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2877, pruned_loss=0.05688, over 3124838.84 frames. ], batch size: 190, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:16,512 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.847e+02 3.369e+02 4.013e+02 7.021e+02, threshold=6.738e+02, percent-clipped=2.0 2023-05-02 01:11:27,494 INFO [train.py:904] (5/8) Epoch 25, batch 6850, loss[loss=0.2068, simple_loss=0.3117, pruned_loss=0.05088, over 16704.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2888, pruned_loss=0.05753, over 3108789.81 frames. ], batch size: 76, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:28,064 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4351, 2.1711, 1.8022, 1.9913, 2.4987, 2.1766, 2.2114, 2.6251], device='cuda:5'), covar=tensor([0.0267, 0.0507, 0.0665, 0.0546, 0.0296, 0.0448, 0.0261, 0.0334], device='cuda:5'), in_proj_covar=tensor([0.0218, 0.0238, 0.0227, 0.0228, 0.0238, 0.0237, 0.0237, 0.0236], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:11:43,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9599, 2.8066, 2.9011, 2.1573, 2.7418, 2.2515, 2.7981, 2.9832], device='cuda:5'), covar=tensor([0.0250, 0.0703, 0.0494, 0.1699, 0.0749, 0.0866, 0.0526, 0.0638], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0168, 0.0170, 0.0155, 0.0148, 0.0131, 0.0145, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:11:52,838 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:12:43,358 INFO [train.py:904] (5/8) Epoch 25, batch 6900, loss[loss=0.1973, simple_loss=0.2921, pruned_loss=0.05131, over 16895.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2917, pruned_loss=0.05797, over 3099294.49 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:13:06,142 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250518.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:13:47,068 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.786e+02 3.277e+02 4.153e+02 1.055e+03, threshold=6.554e+02, percent-clipped=5.0 2023-05-02 01:14:00,200 INFO [train.py:904] (5/8) Epoch 25, batch 6950, loss[loss=0.2024, simple_loss=0.2913, pruned_loss=0.05672, over 16515.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2923, pruned_loss=0.05863, over 3124971.26 frames. ], batch size: 35, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:14:43,061 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3022, 3.4196, 3.5851, 3.5799, 3.5811, 3.3977, 3.4358, 3.4731], device='cuda:5'), covar=tensor([0.0419, 0.0772, 0.0537, 0.0471, 0.0570, 0.0640, 0.0847, 0.0545], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0468, 0.0454, 0.0416, 0.0499, 0.0476, 0.0553, 0.0380], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 01:14:47,864 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6393, 4.8526, 4.9863, 4.7820, 4.8194, 5.3667, 4.8139, 4.5634], device='cuda:5'), covar=tensor([0.1155, 0.1833, 0.2308, 0.1999, 0.2440, 0.0874, 0.1608, 0.2317], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0617, 0.0678, 0.0505, 0.0670, 0.0704, 0.0525, 0.0675], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 01:15:18,383 INFO [train.py:904] (5/8) Epoch 25, batch 7000, loss[loss=0.2034, simple_loss=0.3026, pruned_loss=0.05211, over 16788.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2929, pruned_loss=0.05884, over 3095959.80 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:15:21,941 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1184, 2.4009, 2.5675, 2.0063, 2.7044, 2.7944, 2.4091, 2.4046], device='cuda:5'), covar=tensor([0.0664, 0.0273, 0.0249, 0.0890, 0.0141, 0.0278, 0.0471, 0.0417], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0139, 0.0084, 0.0129, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:15:32,224 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250611.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 01:16:03,597 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:16:22,010 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.716e+02 3.192e+02 3.909e+02 7.863e+02, threshold=6.384e+02, percent-clipped=2.0 2023-05-02 01:16:35,096 INFO [train.py:904] (5/8) Epoch 25, batch 7050, loss[loss=0.1937, simple_loss=0.2809, pruned_loss=0.05322, over 16690.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2933, pruned_loss=0.0587, over 3094385.82 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:16:44,699 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250659.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:16:49,722 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250662.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:14,817 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9033, 4.1436, 3.9620, 4.0327, 3.7291, 3.7976, 3.8223, 4.1537], device='cuda:5'), covar=tensor([0.1306, 0.0968, 0.1100, 0.0871, 0.0882, 0.1777, 0.1006, 0.1042], device='cuda:5'), in_proj_covar=tensor([0.0692, 0.0836, 0.0692, 0.0645, 0.0531, 0.0536, 0.0702, 0.0653], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:17:17,450 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250680.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:20,126 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6532, 3.7845, 2.8243, 2.2279, 2.4344, 2.4099, 4.1168, 3.2978], device='cuda:5'), covar=tensor([0.3076, 0.0656, 0.1951, 0.2871, 0.2902, 0.2206, 0.0448, 0.1426], device='cuda:5'), in_proj_covar=tensor([0.0333, 0.0273, 0.0311, 0.0320, 0.0302, 0.0269, 0.0301, 0.0345], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 01:17:51,107 INFO [train.py:904] (5/8) Epoch 25, batch 7100, loss[loss=0.2475, simple_loss=0.3131, pruned_loss=0.09096, over 11498.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.292, pruned_loss=0.05873, over 3081979.36 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:18:23,440 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250723.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:18:56,843 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.742e+02 3.283e+02 4.161e+02 6.949e+02, threshold=6.565e+02, percent-clipped=2.0 2023-05-02 01:19:09,296 INFO [train.py:904] (5/8) Epoch 25, batch 7150, loss[loss=0.1826, simple_loss=0.2727, pruned_loss=0.04632, over 16252.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2897, pruned_loss=0.05806, over 3086137.95 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:19:31,603 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 01:20:22,686 INFO [train.py:904] (5/8) Epoch 25, batch 7200, loss[loss=0.1906, simple_loss=0.2764, pruned_loss=0.05242, over 12015.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2873, pruned_loss=0.05642, over 3082648.16 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:20:38,799 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8874, 2.1743, 2.4210, 3.0995, 2.2311, 2.3401, 2.3332, 2.2546], device='cuda:5'), covar=tensor([0.1519, 0.3622, 0.2713, 0.0775, 0.4177, 0.2766, 0.3621, 0.3583], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0461, 0.0378, 0.0331, 0.0441, 0.0528, 0.0433, 0.0539], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:21:08,888 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:28,105 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:28,841 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.397e+02 2.821e+02 3.422e+02 6.552e+02, threshold=5.642e+02, percent-clipped=0.0 2023-05-02 01:21:41,039 INFO [train.py:904] (5/8) Epoch 25, batch 7250, loss[loss=0.1986, simple_loss=0.2812, pruned_loss=0.05801, over 16744.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2843, pruned_loss=0.05449, over 3097110.72 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:22:17,018 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8459, 5.1827, 5.4195, 5.1372, 5.1783, 5.7273, 5.2207, 4.9912], device='cuda:5'), covar=tensor([0.1032, 0.1841, 0.2307, 0.1851, 0.2206, 0.0903, 0.1635, 0.2390], device='cuda:5'), in_proj_covar=tensor([0.0422, 0.0617, 0.0679, 0.0506, 0.0671, 0.0705, 0.0528, 0.0677], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 01:22:42,131 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:22:55,048 INFO [train.py:904] (5/8) Epoch 25, batch 7300, loss[loss=0.2295, simple_loss=0.2989, pruned_loss=0.08003, over 11664.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2841, pruned_loss=0.05495, over 3087817.03 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:23:00,809 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250906.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 01:23:59,651 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.893e+02 3.388e+02 3.918e+02 7.629e+02, threshold=6.776e+02, percent-clipped=7.0 2023-05-02 01:24:09,702 INFO [train.py:904] (5/8) Epoch 25, batch 7350, loss[loss=0.2011, simple_loss=0.2877, pruned_loss=0.05727, over 16839.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2852, pruned_loss=0.05579, over 3079590.33 frames. ], batch size: 102, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:24:36,882 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8105, 3.5506, 4.0830, 2.1061, 4.1712, 4.2579, 3.0937, 3.1447], device='cuda:5'), covar=tensor([0.0727, 0.0300, 0.0179, 0.1181, 0.0077, 0.0141, 0.0434, 0.0438], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0139, 0.0084, 0.0129, 0.0129, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:25:20,180 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4962, 3.4823, 3.4272, 2.5895, 3.3473, 2.0616, 3.1124, 2.7673], device='cuda:5'), covar=tensor([0.0196, 0.0163, 0.0245, 0.0326, 0.0135, 0.2757, 0.0178, 0.0325], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0165, 0.0204, 0.0181, 0.0181, 0.0211, 0.0193, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:25:27,065 INFO [train.py:904] (5/8) Epoch 25, batch 7400, loss[loss=0.2031, simple_loss=0.3073, pruned_loss=0.04951, over 16699.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2874, pruned_loss=0.05702, over 3061380.26 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:50,978 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:26:35,985 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.828e+02 2.726e+02 3.311e+02 3.902e+02 7.761e+02, threshold=6.622e+02, percent-clipped=1.0 2023-05-02 01:26:46,737 INFO [train.py:904] (5/8) Epoch 25, batch 7450, loss[loss=0.226, simple_loss=0.293, pruned_loss=0.07945, over 11201.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2882, pruned_loss=0.05814, over 3052700.10 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:27:49,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8507, 1.9830, 2.4670, 2.8119, 2.7398, 3.2392, 2.0728, 3.1665], device='cuda:5'), covar=tensor([0.0222, 0.0527, 0.0355, 0.0340, 0.0314, 0.0179, 0.0573, 0.0145], device='cuda:5'), in_proj_covar=tensor([0.0192, 0.0195, 0.0182, 0.0185, 0.0201, 0.0160, 0.0198, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:28:06,273 INFO [train.py:904] (5/8) Epoch 25, batch 7500, loss[loss=0.1827, simple_loss=0.273, pruned_loss=0.04615, over 16846.00 frames. ], tot_loss[loss=0.201, simple_loss=0.288, pruned_loss=0.05696, over 3065822.16 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:28:12,219 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4172, 4.4857, 4.2977, 3.9928, 4.0186, 4.3953, 4.0887, 4.1151], device='cuda:5'), covar=tensor([0.0570, 0.0488, 0.0304, 0.0307, 0.0791, 0.0449, 0.0710, 0.0631], device='cuda:5'), in_proj_covar=tensor([0.0295, 0.0444, 0.0344, 0.0350, 0.0350, 0.0402, 0.0238, 0.0417], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:29:04,994 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2542, 2.4049, 2.5415, 3.9516, 2.3319, 2.7409, 2.4815, 2.5722], device='cuda:5'), covar=tensor([0.1391, 0.3342, 0.2696, 0.0567, 0.3818, 0.2290, 0.3249, 0.3182], device='cuda:5'), in_proj_covar=tensor([0.0412, 0.0462, 0.0378, 0.0332, 0.0441, 0.0529, 0.0435, 0.0540], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:29:12,783 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.911e+02 3.474e+02 4.178e+02 7.715e+02, threshold=6.948e+02, percent-clipped=3.0 2023-05-02 01:29:23,951 INFO [train.py:904] (5/8) Epoch 25, batch 7550, loss[loss=0.1685, simple_loss=0.2576, pruned_loss=0.03965, over 16716.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2877, pruned_loss=0.05766, over 3059968.01 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:30:19,185 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:32,714 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:37,931 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251201.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 01:30:39,959 INFO [train.py:904] (5/8) Epoch 25, batch 7600, loss[loss=0.2191, simple_loss=0.3019, pruned_loss=0.06811, over 15242.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2878, pruned_loss=0.05824, over 3058511.57 frames. ], batch size: 190, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:30:40,849 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 01:31:43,658 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.097e+02 2.965e+02 3.524e+02 4.371e+02 1.071e+03, threshold=7.047e+02, percent-clipped=4.0 2023-05-02 01:31:53,107 INFO [train.py:904] (5/8) Epoch 25, batch 7650, loss[loss=0.1872, simple_loss=0.2784, pruned_loss=0.04803, over 16936.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2883, pruned_loss=0.05908, over 3041583.39 frames. ], batch size: 90, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:32:02,630 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251259.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:32:17,861 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-02 01:33:00,790 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 01:33:06,990 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251301.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:33:08,942 INFO [train.py:904] (5/8) Epoch 25, batch 7700, loss[loss=0.2294, simple_loss=0.2967, pruned_loss=0.08108, over 11571.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2878, pruned_loss=0.05864, over 3058508.49 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:33:32,390 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:33:58,574 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4357, 4.4056, 4.3063, 3.5191, 4.3512, 1.7254, 4.0958, 3.9035], device='cuda:5'), covar=tensor([0.0135, 0.0116, 0.0208, 0.0372, 0.0113, 0.2852, 0.0169, 0.0306], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0165, 0.0204, 0.0181, 0.0181, 0.0211, 0.0193, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:34:08,406 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7582, 1.4263, 1.6903, 1.6612, 1.7550, 1.8807, 1.6493, 1.7989], device='cuda:5'), covar=tensor([0.0256, 0.0389, 0.0202, 0.0296, 0.0271, 0.0192, 0.0397, 0.0151], device='cuda:5'), in_proj_covar=tensor([0.0192, 0.0195, 0.0183, 0.0186, 0.0201, 0.0160, 0.0198, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:34:14,482 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.953e+02 3.490e+02 4.215e+02 8.631e+02, threshold=6.980e+02, percent-clipped=3.0 2023-05-02 01:34:25,533 INFO [train.py:904] (5/8) Epoch 25, batch 7750, loss[loss=0.1809, simple_loss=0.2677, pruned_loss=0.04708, over 17204.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2873, pruned_loss=0.05791, over 3062799.03 frames. ], batch size: 46, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:34:39,755 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:34:45,644 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:35:40,758 INFO [train.py:904] (5/8) Epoch 25, batch 7800, loss[loss=0.1824, simple_loss=0.269, pruned_loss=0.04787, over 16634.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2881, pruned_loss=0.05821, over 3064597.10 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:35:42,506 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251404.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:36:45,155 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.864e+02 3.394e+02 4.001e+02 5.962e+02, threshold=6.789e+02, percent-clipped=0.0 2023-05-02 01:36:55,307 INFO [train.py:904] (5/8) Epoch 25, batch 7850, loss[loss=0.2003, simple_loss=0.2927, pruned_loss=0.05393, over 16942.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.289, pruned_loss=0.05788, over 3069730.57 frames. ], batch size: 109, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:37:14,504 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251465.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:37:50,409 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:38:07,514 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251501.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:38:09,241 INFO [train.py:904] (5/8) Epoch 25, batch 7900, loss[loss=0.2199, simple_loss=0.3041, pruned_loss=0.06788, over 16724.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2887, pruned_loss=0.05797, over 3074914.29 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:38:11,698 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0640, 3.4049, 3.4722, 2.2243, 3.1909, 3.5190, 3.2677, 2.0222], device='cuda:5'), covar=tensor([0.0637, 0.0072, 0.0070, 0.0493, 0.0118, 0.0133, 0.0113, 0.0520], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0086, 0.0087, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 01:38:46,228 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 01:39:04,228 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:17,294 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 2.635e+02 3.162e+02 3.936e+02 7.872e+02, threshold=6.324e+02, percent-clipped=1.0 2023-05-02 01:39:22,132 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:28,560 INFO [train.py:904] (5/8) Epoch 25, batch 7950, loss[loss=0.1944, simple_loss=0.286, pruned_loss=0.05141, over 16220.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2885, pruned_loss=0.05756, over 3091977.96 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:39:30,142 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:40:27,558 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6495, 1.7382, 1.4725, 1.3912, 1.8097, 1.4854, 1.5555, 1.8552], device='cuda:5'), covar=tensor([0.0269, 0.0375, 0.0524, 0.0459, 0.0284, 0.0353, 0.0211, 0.0276], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0235, 0.0225, 0.0227, 0.0236, 0.0235, 0.0234, 0.0233], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:40:46,603 INFO [train.py:904] (5/8) Epoch 25, batch 8000, loss[loss=0.1966, simple_loss=0.2822, pruned_loss=0.05547, over 16554.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2887, pruned_loss=0.05775, over 3096385.93 frames. ], batch size: 57, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:41:43,468 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2159, 4.9818, 4.9395, 5.3921, 5.5864, 4.9360, 5.4819, 5.5576], device='cuda:5'), covar=tensor([0.1837, 0.1415, 0.2373, 0.0838, 0.0767, 0.1102, 0.1097, 0.0959], device='cuda:5'), in_proj_covar=tensor([0.0645, 0.0794, 0.0915, 0.0802, 0.0616, 0.0635, 0.0669, 0.0774], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:41:44,853 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251641.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:41:51,618 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.813e+02 2.777e+02 3.202e+02 3.994e+02 6.409e+02, threshold=6.404e+02, percent-clipped=2.0 2023-05-02 01:42:01,903 INFO [train.py:904] (5/8) Epoch 25, batch 8050, loss[loss=0.2219, simple_loss=0.3036, pruned_loss=0.07011, over 11433.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2885, pruned_loss=0.05709, over 3113680.57 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:42:09,023 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:42:09,119 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4368, 4.3190, 4.5095, 4.6360, 4.8059, 4.3275, 4.7447, 4.8119], device='cuda:5'), covar=tensor([0.1928, 0.1261, 0.1502, 0.0719, 0.0625, 0.1042, 0.0768, 0.0776], device='cuda:5'), in_proj_covar=tensor([0.0644, 0.0793, 0.0914, 0.0802, 0.0616, 0.0634, 0.0669, 0.0774], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:42:22,410 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0699, 2.0599, 2.6804, 3.0142, 2.8721, 3.5777, 2.3875, 3.5307], device='cuda:5'), covar=tensor([0.0237, 0.0549, 0.0329, 0.0327, 0.0341, 0.0156, 0.0496, 0.0133], device='cuda:5'), in_proj_covar=tensor([0.0191, 0.0195, 0.0182, 0.0185, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:43:16,980 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:43:17,641 INFO [train.py:904] (5/8) Epoch 25, batch 8100, loss[loss=0.1876, simple_loss=0.2767, pruned_loss=0.04931, over 16657.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2878, pruned_loss=0.05631, over 3116758.14 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:43:27,359 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251709.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:44:16,637 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1194, 2.4184, 2.5930, 1.9375, 2.6973, 2.8049, 2.4324, 2.3851], device='cuda:5'), covar=tensor([0.0644, 0.0270, 0.0228, 0.0898, 0.0129, 0.0284, 0.0415, 0.0422], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0084, 0.0129, 0.0129, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:44:22,958 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.808e+02 3.404e+02 4.395e+02 1.324e+03, threshold=6.808e+02, percent-clipped=5.0 2023-05-02 01:44:33,798 INFO [train.py:904] (5/8) Epoch 25, batch 8150, loss[loss=0.2131, simple_loss=0.289, pruned_loss=0.06856, over 11568.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2855, pruned_loss=0.05583, over 3093307.85 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:44:45,089 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251760.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:00,606 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:50,918 INFO [train.py:904] (5/8) Epoch 25, batch 8200, loss[loss=0.1982, simple_loss=0.2792, pruned_loss=0.05859, over 17021.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2827, pruned_loss=0.05529, over 3087357.68 frames. ], batch size: 50, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:46:59,194 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 2.618e+02 3.043e+02 3.668e+02 8.118e+02, threshold=6.086e+02, percent-clipped=2.0 2023-05-02 01:47:04,902 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1281, 1.6928, 1.9945, 2.1907, 2.3439, 2.4920, 1.8557, 2.4463], device='cuda:5'), covar=tensor([0.0257, 0.0530, 0.0345, 0.0366, 0.0351, 0.0238, 0.0520, 0.0204], device='cuda:5'), in_proj_covar=tensor([0.0191, 0.0194, 0.0181, 0.0184, 0.0200, 0.0159, 0.0197, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:47:10,671 INFO [train.py:904] (5/8) Epoch 25, batch 8250, loss[loss=0.1936, simple_loss=0.2867, pruned_loss=0.05024, over 15351.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2818, pruned_loss=0.05313, over 3075618.09 frames. ], batch size: 191, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:47:12,510 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:48:29,882 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251902.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:48:30,717 INFO [train.py:904] (5/8) Epoch 25, batch 8300, loss[loss=0.1718, simple_loss=0.2499, pruned_loss=0.04683, over 11896.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2794, pruned_loss=0.05052, over 3058980.25 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:41,168 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 2.151e+02 2.560e+02 3.132e+02 5.384e+02, threshold=5.121e+02, percent-clipped=0.0 2023-05-02 01:49:52,646 INFO [train.py:904] (5/8) Epoch 25, batch 8350, loss[loss=0.1911, simple_loss=0.2851, pruned_loss=0.04852, over 15342.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2796, pruned_loss=0.04944, over 3052560.19 frames. ], batch size: 191, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:59,496 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:50:55,726 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4574, 4.0416, 4.5631, 2.5584, 4.7788, 4.8083, 3.6288, 3.8264], device='cuda:5'), covar=tensor([0.0536, 0.0222, 0.0160, 0.0982, 0.0048, 0.0126, 0.0297, 0.0312], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0109, 0.0098, 0.0137, 0.0083, 0.0127, 0.0127, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 01:51:05,356 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251997.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:51:17,319 INFO [train.py:904] (5/8) Epoch 25, batch 8400, loss[loss=0.1721, simple_loss=0.2658, pruned_loss=0.03921, over 16748.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2765, pruned_loss=0.04733, over 3033639.70 frames. ], batch size: 124, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:51:22,059 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252005.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:51:49,466 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2949, 3.8392, 4.2780, 2.3844, 4.4550, 4.4766, 3.4880, 3.6208], device='cuda:5'), covar=tensor([0.0518, 0.0214, 0.0187, 0.1022, 0.0062, 0.0145, 0.0312, 0.0318], device='cuda:5'), in_proj_covar=tensor([0.0146, 0.0109, 0.0098, 0.0137, 0.0083, 0.0127, 0.0127, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 01:51:54,265 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 01:52:04,301 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0486, 3.3497, 3.6801, 2.1515, 3.1639, 2.2915, 3.6250, 3.5322], device='cuda:5'), covar=tensor([0.0258, 0.0924, 0.0483, 0.2180, 0.0735, 0.1073, 0.0593, 0.0981], device='cuda:5'), in_proj_covar=tensor([0.0154, 0.0164, 0.0166, 0.0152, 0.0144, 0.0128, 0.0142, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:52:27,819 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.253e+02 2.668e+02 3.204e+02 6.662e+02, threshold=5.337e+02, percent-clipped=3.0 2023-05-02 01:52:40,176 INFO [train.py:904] (5/8) Epoch 25, batch 8450, loss[loss=0.169, simple_loss=0.2627, pruned_loss=0.03764, over 15404.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2744, pruned_loss=0.04552, over 3048233.60 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:52:48,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1960, 2.4748, 2.6296, 1.9828, 2.7593, 2.7978, 2.5121, 2.5318], device='cuda:5'), covar=tensor([0.0562, 0.0269, 0.0246, 0.0925, 0.0123, 0.0306, 0.0430, 0.0378], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0108, 0.0097, 0.0136, 0.0082, 0.0126, 0.0126, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 01:52:51,832 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252060.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:52:59,913 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:54:01,930 INFO [train.py:904] (5/8) Epoch 25, batch 8500, loss[loss=0.1692, simple_loss=0.2485, pruned_loss=0.04498, over 11706.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2713, pruned_loss=0.04354, over 3051550.61 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:54:11,580 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252108.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:54:39,163 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0154, 5.4517, 5.5573, 5.3021, 5.3567, 5.9134, 5.3867, 5.0984], device='cuda:5'), covar=tensor([0.1019, 0.1723, 0.2147, 0.1862, 0.2215, 0.0837, 0.1440, 0.2274], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0600, 0.0665, 0.0493, 0.0654, 0.0689, 0.0516, 0.0661], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 01:54:47,333 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9770, 2.7734, 2.5868, 2.0349, 2.5049, 2.7552, 2.6380, 2.0007], device='cuda:5'), covar=tensor([0.0396, 0.0097, 0.0080, 0.0342, 0.0139, 0.0124, 0.0110, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0085, 0.0086, 0.0133, 0.0098, 0.0110, 0.0095, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 01:55:15,324 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.299e+02 2.680e+02 3.253e+02 6.153e+02, threshold=5.360e+02, percent-clipped=1.0 2023-05-02 01:55:28,838 INFO [train.py:904] (5/8) Epoch 25, batch 8550, loss[loss=0.1737, simple_loss=0.2756, pruned_loss=0.03593, over 16263.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2689, pruned_loss=0.04242, over 3025452.94 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:56:18,167 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7642, 3.0638, 3.4465, 1.9504, 2.8457, 2.1125, 3.2911, 3.3109], device='cuda:5'), covar=tensor([0.0307, 0.0942, 0.0565, 0.2280, 0.0880, 0.1124, 0.0739, 0.1076], device='cuda:5'), in_proj_covar=tensor([0.0155, 0.0164, 0.0166, 0.0152, 0.0144, 0.0129, 0.0142, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 01:57:09,044 INFO [train.py:904] (5/8) Epoch 25, batch 8600, loss[loss=0.1776, simple_loss=0.2738, pruned_loss=0.04072, over 15386.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2692, pruned_loss=0.04145, over 3034434.31 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:57:24,013 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9557, 2.3215, 2.0295, 2.1424, 2.6534, 2.2838, 2.4447, 2.7925], device='cuda:5'), covar=tensor([0.0207, 0.0503, 0.0582, 0.0535, 0.0319, 0.0480, 0.0262, 0.0318], device='cuda:5'), in_proj_covar=tensor([0.0212, 0.0233, 0.0223, 0.0224, 0.0233, 0.0232, 0.0230, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:57:30,059 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5906, 3.5770, 3.5424, 2.7448, 3.4561, 2.0638, 3.2866, 2.9698], device='cuda:5'), covar=tensor([0.0131, 0.0118, 0.0160, 0.0189, 0.0092, 0.2426, 0.0117, 0.0272], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0163, 0.0202, 0.0179, 0.0179, 0.0210, 0.0191, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 01:58:34,362 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.200e+02 2.678e+02 3.278e+02 9.763e+02, threshold=5.356e+02, percent-clipped=4.0 2023-05-02 01:58:37,316 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5358, 3.7313, 2.8307, 2.1239, 2.2369, 2.4340, 3.9763, 3.1902], device='cuda:5'), covar=tensor([0.3234, 0.0594, 0.1994, 0.3280, 0.3133, 0.2244, 0.0394, 0.1468], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0267, 0.0305, 0.0314, 0.0297, 0.0266, 0.0295, 0.0338], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 01:58:42,126 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 01:58:48,825 INFO [train.py:904] (5/8) Epoch 25, batch 8650, loss[loss=0.1672, simple_loss=0.2567, pruned_loss=0.03888, over 12142.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2671, pruned_loss=0.03999, over 3027689.85 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:00:24,119 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252297.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:00:33,692 INFO [train.py:904] (5/8) Epoch 25, batch 8700, loss[loss=0.1727, simple_loss=0.2573, pruned_loss=0.04404, over 12216.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2643, pruned_loss=0.03896, over 3032582.57 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:01:02,661 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5021, 3.5383, 2.0606, 3.9526, 2.6493, 3.8857, 2.2540, 2.8968], device='cuda:5'), covar=tensor([0.0305, 0.0436, 0.1862, 0.0232, 0.0935, 0.0630, 0.1761, 0.0805], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0175, 0.0192, 0.0163, 0.0174, 0.0213, 0.0201, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 02:01:53,637 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:01:54,401 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.271e+02 2.533e+02 3.010e+02 6.763e+02, threshold=5.067e+02, percent-clipped=2.0 2023-05-02 02:02:09,371 INFO [train.py:904] (5/8) Epoch 25, batch 8750, loss[loss=0.2052, simple_loss=0.2988, pruned_loss=0.05581, over 16425.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2637, pruned_loss=0.03835, over 3026255.92 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:02:41,284 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252365.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:01,519 INFO [train.py:904] (5/8) Epoch 25, batch 8800, loss[loss=0.17, simple_loss=0.2634, pruned_loss=0.0383, over 16191.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2616, pruned_loss=0.03695, over 3029798.52 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:04:13,662 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2531, 1.5633, 1.9869, 2.2267, 2.3020, 2.4988, 1.7367, 2.4813], device='cuda:5'), covar=tensor([0.0307, 0.0672, 0.0428, 0.0383, 0.0402, 0.0261, 0.0620, 0.0191], device='cuda:5'), in_proj_covar=tensor([0.0189, 0.0192, 0.0181, 0.0182, 0.0199, 0.0158, 0.0196, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:04:22,164 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:30,833 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252417.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:54,405 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3637, 4.6648, 4.5309, 4.5190, 4.2200, 4.1979, 4.1643, 4.7380], device='cuda:5'), covar=tensor([0.1151, 0.0856, 0.0832, 0.0760, 0.0733, 0.1610, 0.1083, 0.0830], device='cuda:5'), in_proj_covar=tensor([0.0675, 0.0817, 0.0675, 0.0631, 0.0518, 0.0526, 0.0684, 0.0638], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:05:07,249 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:05:31,486 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.215e+02 2.676e+02 3.061e+02 6.379e+02, threshold=5.352e+02, percent-clipped=4.0 2023-05-02 02:05:46,048 INFO [train.py:904] (5/8) Epoch 25, batch 8850, loss[loss=0.1493, simple_loss=0.2404, pruned_loss=0.0291, over 12489.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2646, pruned_loss=0.03649, over 3031463.81 frames. ], batch size: 246, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:06:39,871 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252478.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:06:48,505 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3948, 3.4277, 3.6462, 3.6484, 3.6535, 3.4783, 3.5023, 3.5215], device='cuda:5'), covar=tensor([0.0402, 0.0754, 0.0562, 0.0469, 0.0500, 0.0556, 0.0826, 0.0538], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0462, 0.0450, 0.0412, 0.0495, 0.0472, 0.0546, 0.0378], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 02:07:16,646 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:07:17,145 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 02:07:31,615 INFO [train.py:904] (5/8) Epoch 25, batch 8900, loss[loss=0.1691, simple_loss=0.2702, pruned_loss=0.03399, over 16272.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2651, pruned_loss=0.03587, over 3044873.24 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:08:24,561 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6000, 3.5837, 3.5312, 2.7988, 3.4759, 1.9736, 3.3416, 2.9958], device='cuda:5'), covar=tensor([0.0146, 0.0132, 0.0170, 0.0231, 0.0115, 0.2476, 0.0145, 0.0283], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0162, 0.0201, 0.0177, 0.0178, 0.0208, 0.0190, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:09:18,931 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.220e+02 2.582e+02 3.143e+02 6.266e+02, threshold=5.163e+02, percent-clipped=2.0 2023-05-02 02:09:34,587 INFO [train.py:904] (5/8) Epoch 25, batch 8950, loss[loss=0.171, simple_loss=0.2606, pruned_loss=0.04072, over 12594.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2648, pruned_loss=0.03591, over 3061778.36 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:10:59,656 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252592.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:11:21,909 INFO [train.py:904] (5/8) Epoch 25, batch 9000, loss[loss=0.1608, simple_loss=0.2494, pruned_loss=0.03607, over 12113.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2617, pruned_loss=0.03461, over 3089547.03 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:11:21,909 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 02:11:31,593 INFO [train.py:938] (5/8) Epoch 25, validation: loss=0.1442, simple_loss=0.248, pruned_loss=0.02014, over 944034.00 frames. 2023-05-02 02:11:31,595 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 02:11:53,539 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:12:30,441 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3720, 4.3474, 4.2429, 3.5149, 4.2726, 1.7312, 4.0156, 3.9741], device='cuda:5'), covar=tensor([0.0120, 0.0123, 0.0187, 0.0325, 0.0120, 0.2821, 0.0151, 0.0253], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0163, 0.0201, 0.0177, 0.0178, 0.0209, 0.0190, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:13:01,732 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.125e+02 2.594e+02 3.127e+02 6.449e+02, threshold=5.187e+02, percent-clipped=3.0 2023-05-02 02:13:14,926 INFO [train.py:904] (5/8) Epoch 25, batch 9050, loss[loss=0.142, simple_loss=0.2345, pruned_loss=0.02476, over 16885.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2628, pruned_loss=0.03504, over 3078426.24 frames. ], batch size: 102, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:13:15,986 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252653.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:47,192 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:56,978 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:14:23,835 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7280, 2.3926, 2.2071, 3.5172, 1.6963, 3.5701, 1.4748, 2.7174], device='cuda:5'), covar=tensor([0.1554, 0.0962, 0.1537, 0.0201, 0.0095, 0.0476, 0.2016, 0.0871], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0175, 0.0193, 0.0191, 0.0200, 0.0212, 0.0204, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 02:14:58,393 INFO [train.py:904] (5/8) Epoch 25, batch 9100, loss[loss=0.1783, simple_loss=0.2731, pruned_loss=0.04178, over 16717.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2627, pruned_loss=0.03545, over 3085871.48 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:16:01,692 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:16:41,928 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.125e+02 2.555e+02 2.905e+02 5.383e+02, threshold=5.110e+02, percent-clipped=1.0 2023-05-02 02:16:54,987 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 02:16:57,725 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 02:16:57,794 INFO [train.py:904] (5/8) Epoch 25, batch 9150, loss[loss=0.1569, simple_loss=0.2507, pruned_loss=0.03157, over 16499.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.263, pruned_loss=0.03518, over 3080991.65 frames. ], batch size: 68, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:17:42,776 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:18:20,261 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:18:20,661 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 02:18:43,281 INFO [train.py:904] (5/8) Epoch 25, batch 9200, loss[loss=0.1728, simple_loss=0.2666, pruned_loss=0.03951, over 16222.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2586, pruned_loss=0.03433, over 3088017.38 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:20:05,207 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.337e+02 2.810e+02 3.410e+02 9.767e+02, threshold=5.620e+02, percent-clipped=4.0 2023-05-02 02:20:19,117 INFO [train.py:904] (5/8) Epoch 25, batch 9250, loss[loss=0.1434, simple_loss=0.24, pruned_loss=0.02339, over 16572.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2578, pruned_loss=0.03428, over 3082352.48 frames. ], batch size: 62, lr: 2.67e-03, grad_scale: 16.0 2023-05-02 02:20:28,379 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252857.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:20:38,183 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 02:22:11,619 INFO [train.py:904] (5/8) Epoch 25, batch 9300, loss[loss=0.1671, simple_loss=0.2486, pruned_loss=0.04283, over 12403.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2565, pruned_loss=0.03356, over 3097502.23 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:22:47,858 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9568, 3.8859, 4.0458, 4.1700, 4.2608, 3.8671, 4.2477, 4.3105], device='cuda:5'), covar=tensor([0.1905, 0.1118, 0.1454, 0.0735, 0.0676, 0.1379, 0.0769, 0.0678], device='cuda:5'), in_proj_covar=tensor([0.0622, 0.0764, 0.0879, 0.0777, 0.0595, 0.0614, 0.0644, 0.0747], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:22:47,947 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:22:56,679 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 02:23:45,077 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.953e+02 2.173e+02 2.787e+02 5.118e+02, threshold=4.346e+02, percent-clipped=0.0 2023-05-02 02:23:47,881 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252948.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:23:55,323 INFO [train.py:904] (5/8) Epoch 25, batch 9350, loss[loss=0.1749, simple_loss=0.2691, pruned_loss=0.04034, over 16859.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2567, pruned_loss=0.03352, over 3095534.79 frames. ], batch size: 116, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:24:30,032 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:24:50,085 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4949, 3.5010, 2.6003, 2.0957, 2.1374, 2.2774, 3.6644, 2.9583], device='cuda:5'), covar=tensor([0.3256, 0.0663, 0.2125, 0.3323, 0.3304, 0.2358, 0.0460, 0.1584], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0264, 0.0303, 0.0312, 0.0292, 0.0263, 0.0293, 0.0335], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 02:25:16,674 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252993.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:25:36,667 INFO [train.py:904] (5/8) Epoch 25, batch 9400, loss[loss=0.1803, simple_loss=0.2858, pruned_loss=0.03739, over 16633.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2563, pruned_loss=0.03309, over 3095937.70 frames. ], batch size: 83, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:25:42,798 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0047, 3.8827, 4.0532, 4.1705, 4.2553, 3.8970, 4.2492, 4.3089], device='cuda:5'), covar=tensor([0.1709, 0.1174, 0.1519, 0.0805, 0.0627, 0.1567, 0.0803, 0.0810], device='cuda:5'), in_proj_covar=tensor([0.0622, 0.0764, 0.0880, 0.0777, 0.0596, 0.0616, 0.0644, 0.0747], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:26:19,764 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:26:44,085 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7340, 1.4002, 1.7403, 1.6983, 1.8240, 1.9458, 1.6965, 1.8436], device='cuda:5'), covar=tensor([0.0347, 0.0481, 0.0262, 0.0400, 0.0376, 0.0249, 0.0488, 0.0166], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0192, 0.0180, 0.0182, 0.0198, 0.0157, 0.0195, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:27:02,624 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-02 02:27:05,790 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.088e+02 2.613e+02 3.190e+02 8.435e+02, threshold=5.225e+02, percent-clipped=5.0 2023-05-02 02:27:17,028 INFO [train.py:904] (5/8) Epoch 25, batch 9450, loss[loss=0.1812, simple_loss=0.2746, pruned_loss=0.04395, over 16710.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2582, pruned_loss=0.03332, over 3090454.36 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:27:19,788 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:27:57,818 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:20,125 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9711, 4.2372, 4.0441, 4.0982, 3.7560, 3.8538, 3.8512, 4.2289], device='cuda:5'), covar=tensor([0.1161, 0.0990, 0.1032, 0.0856, 0.0817, 0.1760, 0.1043, 0.1034], device='cuda:5'), in_proj_covar=tensor([0.0676, 0.0815, 0.0673, 0.0631, 0.0518, 0.0526, 0.0683, 0.0640], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:28:31,285 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253090.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:54,660 INFO [train.py:904] (5/8) Epoch 25, batch 9500, loss[loss=0.1543, simple_loss=0.2393, pruned_loss=0.03462, over 12606.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2577, pruned_loss=0.03337, over 3087970.90 frames. ], batch size: 246, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:29:33,433 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253121.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:30:06,534 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:30:26,757 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.108e+02 2.493e+02 2.917e+02 7.333e+02, threshold=4.986e+02, percent-clipped=1.0 2023-05-02 02:30:40,908 INFO [train.py:904] (5/8) Epoch 25, batch 9550, loss[loss=0.1588, simple_loss=0.251, pruned_loss=0.0333, over 12627.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2577, pruned_loss=0.03372, over 3090212.23 frames. ], batch size: 246, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:30:48,750 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8103, 5.1333, 4.9443, 4.9048, 4.5957, 4.6505, 4.5011, 5.2116], device='cuda:5'), covar=tensor([0.1329, 0.0917, 0.1004, 0.0892, 0.0816, 0.1086, 0.1294, 0.0866], device='cuda:5'), in_proj_covar=tensor([0.0676, 0.0814, 0.0672, 0.0631, 0.0518, 0.0525, 0.0682, 0.0639], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:32:22,177 INFO [train.py:904] (5/8) Epoch 25, batch 9600, loss[loss=0.1702, simple_loss=0.257, pruned_loss=0.04164, over 12353.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2589, pruned_loss=0.0343, over 3083984.47 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:32:42,652 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253213.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:33:29,394 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 02:33:55,688 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.388e+02 2.773e+02 3.482e+02 6.416e+02, threshold=5.546e+02, percent-clipped=8.0 2023-05-02 02:33:59,810 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253248.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:34:07,829 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9245, 4.5774, 4.4086, 3.2627, 3.8903, 4.5258, 3.9816, 2.6292], device='cuda:5'), covar=tensor([0.0517, 0.0035, 0.0040, 0.0364, 0.0095, 0.0074, 0.0071, 0.0492], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0084, 0.0085, 0.0132, 0.0098, 0.0108, 0.0094, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 02:34:10,806 INFO [train.py:904] (5/8) Epoch 25, batch 9650, loss[loss=0.1589, simple_loss=0.247, pruned_loss=0.03544, over 12519.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2606, pruned_loss=0.0348, over 3057628.12 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:34:52,674 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:35:46,089 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253296.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:00,320 INFO [train.py:904] (5/8) Epoch 25, batch 9700, loss[loss=0.1605, simple_loss=0.2579, pruned_loss=0.03149, over 15232.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2596, pruned_loss=0.03443, over 3076720.69 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:36:27,592 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:42,990 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:56,160 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253329.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:37:34,482 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 2.159e+02 2.493e+02 2.931e+02 5.551e+02, threshold=4.985e+02, percent-clipped=1.0 2023-05-02 02:37:37,195 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253349.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:37:44,454 INFO [train.py:904] (5/8) Epoch 25, batch 9750, loss[loss=0.1628, simple_loss=0.2587, pruned_loss=0.03345, over 16282.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2586, pruned_loss=0.03453, over 3074539.59 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:38:21,447 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:38:38,998 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253380.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:01,192 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253390.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:23,686 INFO [train.py:904] (5/8) Epoch 25, batch 9800, loss[loss=0.1702, simple_loss=0.2747, pruned_loss=0.0328, over 16976.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2588, pruned_loss=0.0341, over 3063730.10 frames. ], batch size: 109, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:39:31,998 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9307, 3.2699, 3.1624, 2.1075, 2.9559, 3.2488, 3.0802, 2.0077], device='cuda:5'), covar=tensor([0.0612, 0.0054, 0.0071, 0.0462, 0.0129, 0.0090, 0.0097, 0.0516], device='cuda:5'), in_proj_covar=tensor([0.0132, 0.0084, 0.0085, 0.0132, 0.0098, 0.0107, 0.0094, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 02:39:53,073 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:58,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6255, 3.7394, 2.2188, 4.2401, 2.7477, 4.1788, 2.6919, 3.0944], device='cuda:5'), covar=tensor([0.0308, 0.0356, 0.1806, 0.0288, 0.0923, 0.0480, 0.1338, 0.0733], device='cuda:5'), in_proj_covar=tensor([0.0167, 0.0172, 0.0191, 0.0160, 0.0172, 0.0208, 0.0198, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 02:40:07,739 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 02:40:38,478 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253441.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:40:55,720 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 1.999e+02 2.298e+02 2.745e+02 1.319e+03, threshold=4.597e+02, percent-clipped=1.0 2023-05-02 02:41:06,303 INFO [train.py:904] (5/8) Epoch 25, batch 9850, loss[loss=0.1762, simple_loss=0.2724, pruned_loss=0.03999, over 16339.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2604, pruned_loss=0.03397, over 3078584.63 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:41:46,393 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-05-02 02:42:04,016 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:43:00,298 INFO [train.py:904] (5/8) Epoch 25, batch 9900, loss[loss=0.1731, simple_loss=0.2704, pruned_loss=0.03787, over 15251.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2607, pruned_loss=0.0338, over 3076965.16 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:43:25,536 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253513.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:43:41,972 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:44:39,445 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4498, 2.8804, 3.1453, 1.8428, 2.8303, 2.1483, 3.1104, 3.1723], device='cuda:5'), covar=tensor([0.0250, 0.0891, 0.0583, 0.2231, 0.0788, 0.1018, 0.0624, 0.0881], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0159, 0.0163, 0.0150, 0.0141, 0.0127, 0.0139, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 02:44:46,260 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 2.031e+02 2.358e+02 2.986e+02 6.654e+02, threshold=4.715e+02, percent-clipped=4.0 2023-05-02 02:44:59,719 INFO [train.py:904] (5/8) Epoch 25, batch 9950, loss[loss=0.1713, simple_loss=0.2612, pruned_loss=0.04068, over 12674.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2627, pruned_loss=0.03422, over 3071107.57 frames. ], batch size: 246, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:45:18,773 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253561.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:45:34,742 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:45:44,637 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8451, 3.1162, 3.3684, 1.9217, 2.9126, 2.2132, 3.4445, 3.3919], device='cuda:5'), covar=tensor([0.0196, 0.0761, 0.0563, 0.2077, 0.0729, 0.0945, 0.0529, 0.0876], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0158, 0.0163, 0.0150, 0.0141, 0.0126, 0.0139, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 02:46:14,355 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:46:27,758 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0023, 3.9060, 4.0749, 4.1759, 4.2843, 3.8676, 4.2712, 4.3089], device='cuda:5'), covar=tensor([0.1635, 0.1131, 0.1435, 0.0812, 0.0626, 0.1509, 0.0726, 0.0879], device='cuda:5'), in_proj_covar=tensor([0.0621, 0.0760, 0.0876, 0.0775, 0.0595, 0.0613, 0.0643, 0.0746], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 02:47:02,506 INFO [train.py:904] (5/8) Epoch 25, batch 10000, loss[loss=0.1517, simple_loss=0.2438, pruned_loss=0.02981, over 17037.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.261, pruned_loss=0.03382, over 3070787.73 frames. ], batch size: 55, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:47:21,357 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 02:47:53,029 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253628.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:10,862 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-05-02 02:48:10,955 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-02 02:48:36,795 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 2.008e+02 2.354e+02 2.749e+02 5.133e+02, threshold=4.708e+02, percent-clipped=1.0 2023-05-02 02:48:39,825 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:41,250 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253650.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:46,090 INFO [train.py:904] (5/8) Epoch 25, batch 10050, loss[loss=0.1841, simple_loss=0.2788, pruned_loss=0.04463, over 16225.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2613, pruned_loss=0.03403, over 3056834.83 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:48:53,118 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 02:49:24,584 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4817, 3.2254, 3.5830, 1.7787, 3.6793, 3.7960, 2.9553, 2.8589], device='cuda:5'), covar=tensor([0.0780, 0.0277, 0.0181, 0.1247, 0.0085, 0.0164, 0.0446, 0.0485], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0107, 0.0093, 0.0134, 0.0080, 0.0122, 0.0124, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 02:49:48,488 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253685.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:11,424 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253697.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:20,881 INFO [train.py:904] (5/8) Epoch 25, batch 10100, loss[loss=0.1526, simple_loss=0.2456, pruned_loss=0.02977, over 16719.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2616, pruned_loss=0.0343, over 3041055.88 frames. ], batch size: 76, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:50:35,645 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253711.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:35,754 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4649, 3.3365, 2.7895, 2.1932, 2.1782, 2.3271, 3.4810, 2.9852], device='cuda:5'), covar=tensor([0.2946, 0.0678, 0.1655, 0.2975, 0.3017, 0.2233, 0.0449, 0.1492], device='cuda:5'), in_proj_covar=tensor([0.0321, 0.0262, 0.0300, 0.0307, 0.0288, 0.0260, 0.0289, 0.0332], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 02:51:21,606 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253736.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:51:33,767 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.529e+02 2.192e+02 2.668e+02 3.131e+02 7.050e+02, threshold=5.336e+02, percent-clipped=1.0 2023-05-02 02:52:07,295 INFO [train.py:904] (5/8) Epoch 26, batch 0, loss[loss=0.2264, simple_loss=0.3094, pruned_loss=0.07169, over 12375.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3094, pruned_loss=0.07169, over 12375.00 frames. ], batch size: 247, lr: 2.61e-03, grad_scale: 8.0 2023-05-02 02:52:07,295 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 02:52:14,677 INFO [train.py:938] (5/8) Epoch 26, validation: loss=0.1437, simple_loss=0.2472, pruned_loss=0.02009, over 944034.00 frames. 2023-05-02 02:52:14,677 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 02:52:44,915 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253775.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:53:23,662 INFO [train.py:904] (5/8) Epoch 26, batch 50, loss[loss=0.1759, simple_loss=0.2624, pruned_loss=0.04467, over 17077.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2678, pruned_loss=0.04643, over 749940.77 frames. ], batch size: 55, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:54:28,067 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.375e+02 2.827e+02 3.414e+02 5.395e+02, threshold=5.654e+02, percent-clipped=1.0 2023-05-02 02:54:30,996 INFO [train.py:904] (5/8) Epoch 26, batch 100, loss[loss=0.1262, simple_loss=0.2117, pruned_loss=0.02038, over 16996.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2636, pruned_loss=0.04381, over 1332006.81 frames. ], batch size: 41, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:55:03,754 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253877.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:55:39,325 INFO [train.py:904] (5/8) Epoch 26, batch 150, loss[loss=0.1724, simple_loss=0.272, pruned_loss=0.03643, over 17051.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2618, pruned_loss=0.04267, over 1775504.15 frames. ], batch size: 53, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:56:08,415 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:56:19,629 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-02 02:56:46,629 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.138e+02 2.556e+02 3.055e+02 5.694e+02, threshold=5.112e+02, percent-clipped=1.0 2023-05-02 02:56:49,075 INFO [train.py:904] (5/8) Epoch 26, batch 200, loss[loss=0.1773, simple_loss=0.2595, pruned_loss=0.04753, over 16742.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.261, pruned_loss=0.04226, over 2107626.26 frames. ], batch size: 83, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:57:34,322 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253985.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:03,920 INFO [train.py:904] (5/8) Epoch 26, batch 250, loss[loss=0.169, simple_loss=0.2619, pruned_loss=0.03805, over 17123.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2591, pruned_loss=0.04251, over 2375998.14 frames. ], batch size: 49, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:58:07,781 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:47,186 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:50,846 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:10,749 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.097e+02 2.455e+02 3.037e+02 5.274e+02, threshold=4.910e+02, percent-clipped=1.0 2023-05-02 02:59:14,080 INFO [train.py:904] (5/8) Epoch 26, batch 300, loss[loss=0.1473, simple_loss=0.243, pruned_loss=0.02577, over 17158.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2563, pruned_loss=0.04124, over 2587513.90 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:59:45,475 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254075.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:57,805 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:00:23,594 INFO [train.py:904] (5/8) Epoch 26, batch 350, loss[loss=0.1712, simple_loss=0.2495, pruned_loss=0.04644, over 16751.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2544, pruned_loss=0.04084, over 2744747.21 frames. ], batch size: 89, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 03:00:52,070 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:00:52,244 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3746, 3.2937, 3.5159, 2.4861, 3.2899, 3.6731, 3.3490, 2.2064], device='cuda:5'), covar=tensor([0.0532, 0.0166, 0.0071, 0.0427, 0.0134, 0.0117, 0.0119, 0.0489], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0101, 0.0111, 0.0096, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 03:01:30,801 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.013e+02 2.457e+02 3.049e+02 5.935e+02, threshold=4.915e+02, percent-clipped=1.0 2023-05-02 03:01:33,888 INFO [train.py:904] (5/8) Epoch 26, batch 400, loss[loss=0.1451, simple_loss=0.2304, pruned_loss=0.0299, over 12528.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.253, pruned_loss=0.04028, over 2866369.91 frames. ], batch size: 247, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:02:07,829 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:02:14,800 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0215, 4.9038, 4.8869, 4.4705, 4.5280, 4.9677, 4.7789, 4.6604], device='cuda:5'), covar=tensor([0.0707, 0.0816, 0.0397, 0.0406, 0.1114, 0.0470, 0.0528, 0.0744], device='cuda:5'), in_proj_covar=tensor([0.0299, 0.0446, 0.0349, 0.0352, 0.0350, 0.0403, 0.0239, 0.0416], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:02:44,717 INFO [train.py:904] (5/8) Epoch 26, batch 450, loss[loss=0.1617, simple_loss=0.2402, pruned_loss=0.04163, over 16889.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2507, pruned_loss=0.03961, over 2963850.16 frames. ], batch size: 96, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:03:12,913 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:03:15,097 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254225.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:03:50,702 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.133e+02 2.439e+02 2.881e+02 4.883e+02, threshold=4.879e+02, percent-clipped=0.0 2023-05-02 03:03:53,001 INFO [train.py:904] (5/8) Epoch 26, batch 500, loss[loss=0.1768, simple_loss=0.2544, pruned_loss=0.04956, over 16856.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2492, pruned_loss=0.03886, over 3040701.26 frames. ], batch size: 116, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:04:19,051 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254271.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:04:57,427 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7145, 3.8131, 2.4847, 4.3161, 3.0368, 4.3004, 2.6902, 3.2300], device='cuda:5'), covar=tensor([0.0371, 0.0444, 0.1687, 0.0368, 0.0863, 0.0619, 0.1463, 0.0783], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0179, 0.0197, 0.0169, 0.0178, 0.0217, 0.0204, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 03:05:01,752 INFO [train.py:904] (5/8) Epoch 26, batch 550, loss[loss=0.1587, simple_loss=0.2547, pruned_loss=0.03135, over 17139.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2491, pruned_loss=0.03831, over 3106190.47 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:05:05,519 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:05:47,907 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 03:06:08,398 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.164e+02 2.429e+02 2.784e+02 6.302e+02, threshold=4.858e+02, percent-clipped=1.0 2023-05-02 03:06:11,652 INFO [train.py:904] (5/8) Epoch 26, batch 600, loss[loss=0.1699, simple_loss=0.2651, pruned_loss=0.0374, over 17012.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2485, pruned_loss=0.03861, over 3156632.31 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:06:13,075 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:07:21,749 INFO [train.py:904] (5/8) Epoch 26, batch 650, loss[loss=0.1495, simple_loss=0.2492, pruned_loss=0.02491, over 17134.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2479, pruned_loss=0.03814, over 3192961.77 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:07:22,577 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 03:07:34,535 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-05-02 03:08:28,769 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.176e+02 2.565e+02 3.138e+02 6.256e+02, threshold=5.131e+02, percent-clipped=2.0 2023-05-02 03:08:30,998 INFO [train.py:904] (5/8) Epoch 26, batch 700, loss[loss=0.1332, simple_loss=0.2269, pruned_loss=0.01979, over 16748.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2471, pruned_loss=0.03759, over 3220019.37 frames. ], batch size: 39, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:08:34,513 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6169, 2.6412, 2.2114, 2.5683, 3.0168, 2.7386, 3.2759, 3.1786], device='cuda:5'), covar=tensor([0.0193, 0.0535, 0.0689, 0.0545, 0.0335, 0.0507, 0.0294, 0.0351], device='cuda:5'), in_proj_covar=tensor([0.0226, 0.0246, 0.0234, 0.0234, 0.0245, 0.0244, 0.0241, 0.0242], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:09:41,480 INFO [train.py:904] (5/8) Epoch 26, batch 750, loss[loss=0.1844, simple_loss=0.259, pruned_loss=0.05492, over 16701.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2482, pruned_loss=0.0382, over 3246648.28 frames. ], batch size: 134, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:10:27,485 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254536.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:10:46,997 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 2.194e+02 2.558e+02 3.127e+02 5.314e+02, threshold=5.115e+02, percent-clipped=1.0 2023-05-02 03:10:50,620 INFO [train.py:904] (5/8) Epoch 26, batch 800, loss[loss=0.1382, simple_loss=0.2287, pruned_loss=0.02388, over 16969.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2477, pruned_loss=0.03792, over 3257235.46 frames. ], batch size: 41, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:10:50,936 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5031, 4.5883, 4.6953, 4.5260, 4.5695, 5.1510, 4.6715, 4.3239], device='cuda:5'), covar=tensor([0.1741, 0.2219, 0.2774, 0.2410, 0.2858, 0.1218, 0.1798, 0.2763], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0622, 0.0691, 0.0512, 0.0677, 0.0713, 0.0535, 0.0681], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:11:20,515 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254574.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:11:51,532 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:11:59,172 INFO [train.py:904] (5/8) Epoch 26, batch 850, loss[loss=0.1691, simple_loss=0.2509, pruned_loss=0.04368, over 16740.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2478, pruned_loss=0.0374, over 3277818.99 frames. ], batch size: 124, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:12:45,162 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:13:08,052 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 2.023e+02 2.316e+02 2.771e+02 4.972e+02, threshold=4.632e+02, percent-clipped=1.0 2023-05-02 03:13:10,208 INFO [train.py:904] (5/8) Epoch 26, batch 900, loss[loss=0.1635, simple_loss=0.2598, pruned_loss=0.03365, over 17144.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2471, pruned_loss=0.03728, over 3273679.83 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:13:44,876 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0075, 4.5983, 4.5160, 3.2317, 3.8057, 4.5359, 4.0244, 2.9764], device='cuda:5'), covar=tensor([0.0511, 0.0058, 0.0049, 0.0387, 0.0146, 0.0096, 0.0095, 0.0425], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0136, 0.0101, 0.0112, 0.0097, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 03:13:58,861 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8318, 2.9293, 3.2368, 2.1298, 2.7478, 2.2449, 3.3103, 3.2991], device='cuda:5'), covar=tensor([0.0264, 0.0995, 0.0613, 0.2040, 0.0958, 0.1024, 0.0604, 0.0956], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0168, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 03:14:19,304 INFO [train.py:904] (5/8) Epoch 26, batch 950, loss[loss=0.1591, simple_loss=0.2499, pruned_loss=0.03414, over 16687.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.247, pruned_loss=0.03743, over 3288662.09 frames. ], batch size: 62, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:14:34,110 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3570, 3.8418, 4.3522, 2.3467, 4.5170, 4.6499, 3.3762, 3.6246], device='cuda:5'), covar=tensor([0.0620, 0.0288, 0.0253, 0.1129, 0.0102, 0.0176, 0.0449, 0.0409], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0111, 0.0099, 0.0140, 0.0085, 0.0129, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 03:14:49,012 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 03:15:24,093 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 2.119e+02 2.483e+02 2.968e+02 5.109e+02, threshold=4.965e+02, percent-clipped=3.0 2023-05-02 03:15:27,110 INFO [train.py:904] (5/8) Epoch 26, batch 1000, loss[loss=0.1352, simple_loss=0.2171, pruned_loss=0.02661, over 16969.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2461, pruned_loss=0.03789, over 3283676.27 frames. ], batch size: 41, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:15:31,165 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:28,391 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:35,899 INFO [train.py:904] (5/8) Epoch 26, batch 1050, loss[loss=0.1766, simple_loss=0.2566, pruned_loss=0.04828, over 15452.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2466, pruned_loss=0.03798, over 3282477.02 frames. ], batch size: 190, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:16:47,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0413, 5.1060, 5.5587, 5.5207, 5.5219, 5.1751, 5.0992, 4.9434], device='cuda:5'), covar=tensor([0.0376, 0.0477, 0.0355, 0.0362, 0.0502, 0.0393, 0.1005, 0.0444], device='cuda:5'), in_proj_covar=tensor([0.0429, 0.0481, 0.0469, 0.0430, 0.0515, 0.0494, 0.0569, 0.0392], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 03:16:55,135 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:42,959 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254851.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:43,594 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 2.171e+02 2.721e+02 3.139e+02 5.089e+02, threshold=5.442e+02, percent-clipped=1.0 2023-05-02 03:17:44,778 INFO [train.py:904] (5/8) Epoch 26, batch 1100, loss[loss=0.1625, simple_loss=0.2629, pruned_loss=0.03107, over 17030.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.246, pruned_loss=0.03743, over 3297797.33 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:17:53,568 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:18:18,018 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0533, 3.9355, 4.1153, 4.2223, 4.2857, 3.8792, 4.1618, 4.3042], device='cuda:5'), covar=tensor([0.1474, 0.1088, 0.1225, 0.0652, 0.0552, 0.1370, 0.1697, 0.0678], device='cuda:5'), in_proj_covar=tensor([0.0673, 0.0824, 0.0953, 0.0839, 0.0639, 0.0661, 0.0696, 0.0806], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:18:36,904 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254892.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:18:53,041 INFO [train.py:904] (5/8) Epoch 26, batch 1150, loss[loss=0.1617, simple_loss=0.2448, pruned_loss=0.0393, over 12314.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2463, pruned_loss=0.03741, over 3306855.03 frames. ], batch size: 246, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:19:04,306 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254912.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:29,842 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:46,592 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-02 03:19:52,828 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9958, 5.0559, 5.4556, 5.4115, 5.4397, 5.1160, 5.0638, 4.9226], device='cuda:5'), covar=tensor([0.0403, 0.0566, 0.0426, 0.0482, 0.0509, 0.0447, 0.0946, 0.0473], device='cuda:5'), in_proj_covar=tensor([0.0428, 0.0482, 0.0470, 0.0430, 0.0516, 0.0494, 0.0569, 0.0393], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 03:19:59,615 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.804e+02 2.093e+02 2.455e+02 3.027e+02 6.541e+02, threshold=4.911e+02, percent-clipped=2.0 2023-05-02 03:20:00,675 INFO [train.py:904] (5/8) Epoch 26, batch 1200, loss[loss=0.1606, simple_loss=0.2374, pruned_loss=0.04194, over 16724.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2452, pruned_loss=0.03708, over 3300226.25 frames. ], batch size: 134, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:10,675 INFO [train.py:904] (5/8) Epoch 26, batch 1250, loss[loss=0.1789, simple_loss=0.2503, pruned_loss=0.0537, over 16833.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2451, pruned_loss=0.03754, over 3309559.73 frames. ], batch size: 102, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:26,826 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 03:22:19,977 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.021e+02 2.372e+02 2.873e+02 4.328e+02, threshold=4.744e+02, percent-clipped=0.0 2023-05-02 03:22:21,133 INFO [train.py:904] (5/8) Epoch 26, batch 1300, loss[loss=0.1412, simple_loss=0.2283, pruned_loss=0.02705, over 15424.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2449, pruned_loss=0.03768, over 3314481.21 frames. ], batch size: 190, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:23:12,974 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 03:23:30,838 INFO [train.py:904] (5/8) Epoch 26, batch 1350, loss[loss=0.147, simple_loss=0.2379, pruned_loss=0.0281, over 17117.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2451, pruned_loss=0.03767, over 3318960.00 frames. ], batch size: 49, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:23:42,888 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:24:38,902 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 2.137e+02 2.466e+02 2.927e+02 7.701e+02, threshold=4.932e+02, percent-clipped=1.0 2023-05-02 03:24:40,852 INFO [train.py:904] (5/8) Epoch 26, batch 1400, loss[loss=0.1543, simple_loss=0.2447, pruned_loss=0.03195, over 17037.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2454, pruned_loss=0.03768, over 3325726.26 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:24:41,768 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:25:12,857 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2881, 4.0393, 4.5122, 2.3674, 4.7686, 4.8340, 3.5611, 3.7312], device='cuda:5'), covar=tensor([0.0729, 0.0263, 0.0236, 0.1235, 0.0067, 0.0168, 0.0418, 0.0430], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0113, 0.0101, 0.0141, 0.0086, 0.0131, 0.0131, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 03:25:29,803 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 03:25:34,149 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:25:37,632 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2735, 3.8076, 4.3207, 2.3563, 4.5581, 4.6693, 3.3843, 3.5760], device='cuda:5'), covar=tensor([0.0644, 0.0282, 0.0251, 0.1136, 0.0088, 0.0175, 0.0469, 0.0406], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0113, 0.0101, 0.0141, 0.0086, 0.0131, 0.0131, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 03:25:43,465 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1239, 5.1398, 5.5565, 5.5362, 5.5761, 5.2251, 5.1306, 4.9803], device='cuda:5'), covar=tensor([0.0355, 0.0625, 0.0438, 0.0433, 0.0464, 0.0423, 0.0979, 0.0514], device='cuda:5'), in_proj_covar=tensor([0.0430, 0.0486, 0.0474, 0.0431, 0.0518, 0.0496, 0.0572, 0.0395], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 03:25:48,428 INFO [train.py:904] (5/8) Epoch 26, batch 1450, loss[loss=0.1603, simple_loss=0.2456, pruned_loss=0.03748, over 16851.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2447, pruned_loss=0.03718, over 3324239.92 frames. ], batch size: 42, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:25:54,181 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:26,916 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255230.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:40,764 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:48,864 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2583, 5.0030, 5.2258, 5.4707, 5.6286, 4.9387, 5.5425, 5.6164], device='cuda:5'), covar=tensor([0.1845, 0.1598, 0.2136, 0.0922, 0.0673, 0.0807, 0.0664, 0.0787], device='cuda:5'), in_proj_covar=tensor([0.0682, 0.0835, 0.0963, 0.0850, 0.0645, 0.0670, 0.0702, 0.0816], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:26:56,778 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.211e+02 2.586e+02 2.983e+02 9.031e+02, threshold=5.172e+02, percent-clipped=4.0 2023-05-02 03:26:57,920 INFO [train.py:904] (5/8) Epoch 26, batch 1500, loss[loss=0.1932, simple_loss=0.2642, pruned_loss=0.06107, over 16769.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.2447, pruned_loss=0.03751, over 3324006.84 frames. ], batch size: 102, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:27:31,537 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:28:04,982 INFO [train.py:904] (5/8) Epoch 26, batch 1550, loss[loss=0.1925, simple_loss=0.2582, pruned_loss=0.06342, over 16535.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2462, pruned_loss=0.03817, over 3315525.16 frames. ], batch size: 68, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:29:12,932 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.260e+02 2.709e+02 3.380e+02 1.198e+03, threshold=5.419e+02, percent-clipped=4.0 2023-05-02 03:29:14,103 INFO [train.py:904] (5/8) Epoch 26, batch 1600, loss[loss=0.1588, simple_loss=0.2592, pruned_loss=0.02924, over 17249.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2483, pruned_loss=0.03914, over 3305419.70 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:29:50,874 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1401, 5.6955, 5.8119, 5.4728, 5.5590, 6.1483, 5.6152, 5.3314], device='cuda:5'), covar=tensor([0.0883, 0.1726, 0.2144, 0.2016, 0.2635, 0.0926, 0.1502, 0.2272], device='cuda:5'), in_proj_covar=tensor([0.0427, 0.0632, 0.0701, 0.0519, 0.0686, 0.0725, 0.0544, 0.0690], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:30:03,509 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:30:23,428 INFO [train.py:904] (5/8) Epoch 26, batch 1650, loss[loss=0.1567, simple_loss=0.2547, pruned_loss=0.02933, over 17203.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2492, pruned_loss=0.03915, over 3305694.02 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:30:35,904 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:29,019 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=255449.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:32,748 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.212e+02 2.578e+02 3.028e+02 5.571e+02, threshold=5.156e+02, percent-clipped=1.0 2023-05-02 03:31:34,038 INFO [train.py:904] (5/8) Epoch 26, batch 1700, loss[loss=0.1841, simple_loss=0.2594, pruned_loss=0.05435, over 16926.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2514, pruned_loss=0.0399, over 3311652.04 frames. ], batch size: 116, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:31:34,492 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:43,234 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255459.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:32:06,770 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7769, 3.7463, 3.8506, 3.6553, 3.8205, 4.2490, 3.8404, 3.5352], device='cuda:5'), covar=tensor([0.2476, 0.2432, 0.2879, 0.2410, 0.2935, 0.1865, 0.1781, 0.2738], device='cuda:5'), in_proj_covar=tensor([0.0428, 0.0632, 0.0699, 0.0517, 0.0684, 0.0723, 0.0543, 0.0688], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:32:29,932 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4372, 3.5984, 3.6090, 1.8260, 2.9427, 2.1448, 3.8295, 3.8071], device='cuda:5'), covar=tensor([0.0240, 0.0914, 0.0688, 0.2583, 0.1054, 0.1255, 0.0622, 0.1014], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0169, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 03:32:40,139 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:32:42,743 INFO [train.py:904] (5/8) Epoch 26, batch 1750, loss[loss=0.1765, simple_loss=0.2718, pruned_loss=0.04063, over 17028.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2528, pruned_loss=0.04036, over 3301623.56 frames. ], batch size: 53, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:32:47,918 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255507.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:33:49,124 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.153e+02 2.652e+02 3.222e+02 7.615e+02, threshold=5.303e+02, percent-clipped=5.0 2023-05-02 03:33:50,397 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-02 03:33:51,390 INFO [train.py:904] (5/8) Epoch 26, batch 1800, loss[loss=0.1481, simple_loss=0.23, pruned_loss=0.03311, over 15839.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2532, pruned_loss=0.03984, over 3303271.21 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:33:54,877 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:34:18,564 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 03:34:28,739 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2550, 2.3772, 2.3388, 3.9694, 2.3190, 2.7071, 2.3887, 2.5224], device='cuda:5'), covar=tensor([0.1578, 0.3805, 0.3303, 0.0692, 0.4095, 0.2675, 0.3952, 0.3436], device='cuda:5'), in_proj_covar=tensor([0.0419, 0.0470, 0.0386, 0.0339, 0.0446, 0.0537, 0.0441, 0.0548], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:34:59,047 INFO [train.py:904] (5/8) Epoch 26, batch 1850, loss[loss=0.1756, simple_loss=0.2576, pruned_loss=0.04685, over 16790.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2545, pruned_loss=0.03996, over 3300726.52 frames. ], batch size: 83, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:36:05,027 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.481e+02 2.077e+02 2.353e+02 2.960e+02 6.152e+02, threshold=4.707e+02, percent-clipped=2.0 2023-05-02 03:36:06,159 INFO [train.py:904] (5/8) Epoch 26, batch 1900, loss[loss=0.1593, simple_loss=0.24, pruned_loss=0.03931, over 16501.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2533, pruned_loss=0.03936, over 3300548.30 frames. ], batch size: 146, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:36:23,467 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-05-02 03:36:42,024 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4448, 4.4398, 4.7808, 4.7757, 4.8301, 4.5046, 4.5022, 4.4089], device='cuda:5'), covar=tensor([0.0393, 0.0674, 0.0439, 0.0459, 0.0550, 0.0491, 0.0948, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0437, 0.0492, 0.0477, 0.0437, 0.0524, 0.0503, 0.0580, 0.0400], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 03:37:15,843 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8651, 2.8073, 2.7087, 4.5117, 3.5104, 4.1662, 1.6834, 3.1785], device='cuda:5'), covar=tensor([0.1367, 0.0787, 0.1175, 0.0227, 0.0230, 0.0447, 0.1637, 0.0758], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0199, 0.0206, 0.0220, 0.0210, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 03:37:16,455 INFO [train.py:904] (5/8) Epoch 26, batch 1950, loss[loss=0.169, simple_loss=0.2613, pruned_loss=0.03839, over 16511.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2528, pruned_loss=0.03878, over 3307602.98 frames. ], batch size: 75, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:37:32,860 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4994, 3.4590, 3.9861, 2.2754, 3.3721, 2.6215, 3.8767, 3.6726], device='cuda:5'), covar=tensor([0.0265, 0.1069, 0.0539, 0.2176, 0.0809, 0.1000, 0.0653, 0.1220], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0170, 0.0171, 0.0158, 0.0148, 0.0133, 0.0147, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 03:37:42,887 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-02 03:37:55,217 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9812, 4.8748, 4.8567, 4.4686, 4.5495, 4.9136, 4.7733, 4.5771], device='cuda:5'), covar=tensor([0.0653, 0.0802, 0.0348, 0.0374, 0.1045, 0.0512, 0.0450, 0.0796], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0472, 0.0368, 0.0372, 0.0371, 0.0426, 0.0253, 0.0445], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:38:12,962 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255744.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:38:24,269 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.052e+02 2.408e+02 2.894e+02 5.578e+02, threshold=4.816e+02, percent-clipped=3.0 2023-05-02 03:38:24,842 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6912, 3.9015, 4.1007, 2.2518, 3.4480, 2.5273, 4.1742, 4.1166], device='cuda:5'), covar=tensor([0.0213, 0.0883, 0.0582, 0.2267, 0.0879, 0.1148, 0.0525, 0.0953], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0170, 0.0171, 0.0158, 0.0149, 0.0133, 0.0147, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 03:38:25,466 INFO [train.py:904] (5/8) Epoch 26, batch 2000, loss[loss=0.144, simple_loss=0.2307, pruned_loss=0.02871, over 15884.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2525, pruned_loss=0.03835, over 3310524.60 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:38:35,742 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7937, 3.8857, 2.8643, 2.3159, 2.6178, 2.3680, 3.9727, 3.3303], device='cuda:5'), covar=tensor([0.2697, 0.0622, 0.1780, 0.2996, 0.2690, 0.2234, 0.0546, 0.1518], device='cuda:5'), in_proj_covar=tensor([0.0333, 0.0274, 0.0312, 0.0323, 0.0304, 0.0273, 0.0303, 0.0349], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:38:59,270 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9876, 2.2233, 2.7490, 3.0034, 2.8874, 3.5237, 2.4686, 3.5320], device='cuda:5'), covar=tensor([0.0347, 0.0590, 0.0365, 0.0406, 0.0396, 0.0228, 0.0520, 0.0197], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0200, 0.0188, 0.0191, 0.0207, 0.0164, 0.0202, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:39:35,310 INFO [train.py:904] (5/8) Epoch 26, batch 2050, loss[loss=0.1727, simple_loss=0.2517, pruned_loss=0.04679, over 16843.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2518, pruned_loss=0.03842, over 3312772.71 frames. ], batch size: 102, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:40:29,774 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 03:40:44,483 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.146e+02 2.478e+02 2.962e+02 5.196e+02, threshold=4.956e+02, percent-clipped=1.0 2023-05-02 03:40:45,707 INFO [train.py:904] (5/8) Epoch 26, batch 2100, loss[loss=0.1888, simple_loss=0.2657, pruned_loss=0.05598, over 16884.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2535, pruned_loss=0.03954, over 3300960.23 frames. ], batch size: 109, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:41:13,067 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6944, 1.8486, 2.3692, 2.5932, 2.6160, 2.6612, 1.8503, 2.8433], device='cuda:5'), covar=tensor([0.0195, 0.0553, 0.0350, 0.0296, 0.0338, 0.0332, 0.0587, 0.0196], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0200, 0.0188, 0.0191, 0.0207, 0.0164, 0.0202, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:41:22,430 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1270, 3.8730, 4.3946, 2.2092, 4.5831, 4.6813, 3.4707, 3.6437], device='cuda:5'), covar=tensor([0.0711, 0.0269, 0.0199, 0.1139, 0.0079, 0.0168, 0.0405, 0.0394], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0112, 0.0101, 0.0140, 0.0086, 0.0131, 0.0130, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 03:41:54,462 INFO [train.py:904] (5/8) Epoch 26, batch 2150, loss[loss=0.1421, simple_loss=0.2325, pruned_loss=0.0259, over 17248.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2536, pruned_loss=0.03982, over 3306187.19 frames. ], batch size: 45, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:41:55,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3209, 5.3115, 5.0376, 4.5552, 5.1800, 1.9698, 4.8845, 4.8227], device='cuda:5'), covar=tensor([0.0109, 0.0091, 0.0242, 0.0400, 0.0101, 0.2861, 0.0153, 0.0264], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0172, 0.0209, 0.0185, 0.0186, 0.0217, 0.0199, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:42:21,318 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0125, 4.5106, 3.1644, 2.4196, 2.6639, 2.7681, 4.8670, 3.5931], device='cuda:5'), covar=tensor([0.2811, 0.0532, 0.1852, 0.2922, 0.3044, 0.2065, 0.0344, 0.1578], device='cuda:5'), in_proj_covar=tensor([0.0334, 0.0275, 0.0312, 0.0323, 0.0305, 0.0274, 0.0303, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:42:44,132 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1214, 2.2536, 2.4731, 3.8945, 2.1916, 2.6257, 2.3098, 2.4389], device='cuda:5'), covar=tensor([0.1601, 0.3821, 0.3096, 0.0764, 0.4018, 0.2599, 0.4081, 0.3222], device='cuda:5'), in_proj_covar=tensor([0.0418, 0.0469, 0.0384, 0.0338, 0.0445, 0.0535, 0.0440, 0.0547], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:43:00,871 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8628, 2.1095, 2.6644, 2.9254, 2.8064, 3.4882, 2.3260, 3.4235], device='cuda:5'), covar=tensor([0.0371, 0.0597, 0.0402, 0.0425, 0.0381, 0.0220, 0.0590, 0.0226], device='cuda:5'), in_proj_covar=tensor([0.0198, 0.0200, 0.0188, 0.0191, 0.0207, 0.0165, 0.0203, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:43:04,642 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.196e+02 2.655e+02 3.354e+02 7.063e+02, threshold=5.310e+02, percent-clipped=7.0 2023-05-02 03:43:05,923 INFO [train.py:904] (5/8) Epoch 26, batch 2200, loss[loss=0.1399, simple_loss=0.2371, pruned_loss=0.02135, over 17133.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2551, pruned_loss=0.04033, over 3306877.02 frames. ], batch size: 48, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:43:10,474 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255956.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:43:12,574 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-05-02 03:43:17,925 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1957, 5.8465, 5.9357, 5.6318, 5.7185, 6.2817, 5.7855, 5.5067], device='cuda:5'), covar=tensor([0.0968, 0.1877, 0.2494, 0.2066, 0.2591, 0.0925, 0.1589, 0.2213], device='cuda:5'), in_proj_covar=tensor([0.0434, 0.0640, 0.0708, 0.0525, 0.0695, 0.0731, 0.0549, 0.0698], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:44:01,834 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-05-02 03:44:18,592 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7401, 4.1643, 4.0959, 2.9295, 3.6368, 4.1941, 3.7980, 2.6119], device='cuda:5'), covar=tensor([0.0543, 0.0120, 0.0065, 0.0415, 0.0149, 0.0101, 0.0111, 0.0467], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0089, 0.0090, 0.0138, 0.0103, 0.0114, 0.0099, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 03:44:19,364 INFO [train.py:904] (5/8) Epoch 26, batch 2250, loss[loss=0.1658, simple_loss=0.257, pruned_loss=0.03726, over 16647.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2555, pruned_loss=0.04018, over 3306580.28 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:44:40,059 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:05,566 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8855, 4.8902, 4.7486, 4.2038, 4.8248, 1.9450, 4.5855, 4.4352], device='cuda:5'), covar=tensor([0.0188, 0.0128, 0.0226, 0.0399, 0.0130, 0.2889, 0.0188, 0.0267], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0172, 0.0210, 0.0185, 0.0186, 0.0218, 0.0199, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:45:09,566 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256038.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:17,863 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256044.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:27,438 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 2.041e+02 2.398e+02 2.796e+02 4.429e+02, threshold=4.795e+02, percent-clipped=0.0 2023-05-02 03:45:29,164 INFO [train.py:904] (5/8) Epoch 26, batch 2300, loss[loss=0.1754, simple_loss=0.2744, pruned_loss=0.03816, over 17106.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2547, pruned_loss=0.03978, over 3313025.94 frames. ], batch size: 49, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:45:38,662 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7208, 3.8321, 2.6828, 4.5219, 3.0693, 4.4657, 2.6350, 3.2172], device='cuda:5'), covar=tensor([0.0377, 0.0478, 0.1543, 0.0362, 0.0862, 0.0580, 0.1626, 0.0808], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0184, 0.0201, 0.0176, 0.0182, 0.0224, 0.0208, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 03:46:09,858 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:14,218 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:18,812 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8591, 2.7561, 2.2190, 2.7213, 3.0400, 2.9297, 3.4187, 3.3756], device='cuda:5'), covar=tensor([0.0209, 0.0533, 0.0735, 0.0558, 0.0373, 0.0452, 0.0339, 0.0343], device='cuda:5'), in_proj_covar=tensor([0.0232, 0.0248, 0.0236, 0.0236, 0.0249, 0.0248, 0.0247, 0.0246], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:46:24,422 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256092.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:34,674 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:39,720 INFO [train.py:904] (5/8) Epoch 26, batch 2350, loss[loss=0.1997, simple_loss=0.287, pruned_loss=0.05624, over 16430.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.256, pruned_loss=0.04044, over 3311154.48 frames. ], batch size: 146, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:47:19,473 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:32,106 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:37,044 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:45,354 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.073e+02 2.391e+02 2.953e+02 7.650e+02, threshold=4.783e+02, percent-clipped=2.0 2023-05-02 03:47:45,722 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6805, 4.7878, 4.9779, 4.7160, 4.7704, 5.3869, 4.8615, 4.5501], device='cuda:5'), covar=tensor([0.1495, 0.1972, 0.2320, 0.2305, 0.2864, 0.1054, 0.1836, 0.2584], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0638, 0.0705, 0.0523, 0.0691, 0.0727, 0.0548, 0.0696], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:47:46,529 INFO [train.py:904] (5/8) Epoch 26, batch 2400, loss[loss=0.1983, simple_loss=0.2699, pruned_loss=0.06333, over 16873.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2561, pruned_loss=0.04, over 3324038.72 frames. ], batch size: 116, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:47:50,008 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 03:48:00,616 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8043, 4.0080, 2.9960, 2.3262, 2.7598, 2.6233, 4.3442, 3.5052], device='cuda:5'), covar=tensor([0.2979, 0.0712, 0.1931, 0.3133, 0.3027, 0.2057, 0.0517, 0.1494], device='cuda:5'), in_proj_covar=tensor([0.0332, 0.0274, 0.0311, 0.0322, 0.0304, 0.0272, 0.0302, 0.0350], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:48:24,885 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6200, 3.3042, 3.7134, 1.9605, 3.8180, 3.8060, 3.1562, 2.8186], device='cuda:5'), covar=tensor([0.0766, 0.0269, 0.0208, 0.1238, 0.0119, 0.0218, 0.0410, 0.0509], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0113, 0.0101, 0.0141, 0.0086, 0.0132, 0.0131, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 03:48:42,798 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256193.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 03:48:55,679 INFO [train.py:904] (5/8) Epoch 26, batch 2450, loss[loss=0.2021, simple_loss=0.2741, pruned_loss=0.06506, over 16904.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2557, pruned_loss=0.03949, over 3325480.53 frames. ], batch size: 109, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:01,729 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.193e+02 2.630e+02 3.165e+02 5.747e+02, threshold=5.261e+02, percent-clipped=2.0 2023-05-02 03:50:03,666 INFO [train.py:904] (5/8) Epoch 26, batch 2500, loss[loss=0.1722, simple_loss=0.2566, pruned_loss=0.04395, over 16936.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2561, pruned_loss=0.03966, over 3321855.51 frames. ], batch size: 116, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:48,274 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256285.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:51:13,210 INFO [train.py:904] (5/8) Epoch 26, batch 2550, loss[loss=0.1706, simple_loss=0.2673, pruned_loss=0.0369, over 17038.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2559, pruned_loss=0.03954, over 3323511.17 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:51:26,261 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256312.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:52:14,137 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256346.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:52:22,560 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.118e+02 2.541e+02 2.890e+02 5.508e+02, threshold=5.081e+02, percent-clipped=1.0 2023-05-02 03:52:23,678 INFO [train.py:904] (5/8) Epoch 26, batch 2600, loss[loss=0.1537, simple_loss=0.249, pruned_loss=0.02914, over 16844.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.255, pruned_loss=0.03894, over 3326577.32 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:52:51,382 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:53:05,469 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9738, 5.4396, 5.6363, 5.2949, 5.4015, 5.9949, 5.4604, 5.1654], device='cuda:5'), covar=tensor([0.1075, 0.1860, 0.2215, 0.2010, 0.2605, 0.1007, 0.1431, 0.2340], device='cuda:5'), in_proj_covar=tensor([0.0434, 0.0638, 0.0705, 0.0522, 0.0692, 0.0727, 0.0547, 0.0697], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 03:53:21,526 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256394.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:53:33,354 INFO [train.py:904] (5/8) Epoch 26, batch 2650, loss[loss=0.1844, simple_loss=0.2709, pruned_loss=0.0489, over 16891.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2565, pruned_loss=0.03933, over 3324484.60 frames. ], batch size: 96, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:54:14,563 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256433.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:20,826 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256437.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:24,144 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256440.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:40,366 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.157e+02 2.407e+02 2.834e+02 4.989e+02, threshold=4.815e+02, percent-clipped=0.0 2023-05-02 03:54:41,410 INFO [train.py:904] (5/8) Epoch 26, batch 2700, loss[loss=0.1738, simple_loss=0.2561, pruned_loss=0.0458, over 16273.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2564, pruned_loss=0.03904, over 3316395.81 frames. ], batch size: 165, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:55:06,519 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:55:28,874 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256488.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 03:55:48,628 INFO [train.py:904] (5/8) Epoch 26, batch 2750, loss[loss=0.1733, simple_loss=0.2724, pruned_loss=0.03711, over 17036.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2568, pruned_loss=0.03867, over 3323954.46 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:56:30,063 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256532.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:56:33,743 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1914, 2.2960, 2.4090, 4.0285, 2.2450, 2.6173, 2.3702, 2.4555], device='cuda:5'), covar=tensor([0.1594, 0.3648, 0.3130, 0.0651, 0.4059, 0.2747, 0.3976, 0.3284], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0471, 0.0386, 0.0339, 0.0447, 0.0538, 0.0442, 0.0551], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:56:56,721 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 2.155e+02 2.522e+02 3.116e+02 8.730e+02, threshold=5.045e+02, percent-clipped=2.0 2023-05-02 03:56:58,721 INFO [train.py:904] (5/8) Epoch 26, batch 2800, loss[loss=0.1895, simple_loss=0.2705, pruned_loss=0.05424, over 12261.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2566, pruned_loss=0.03832, over 3312395.92 frames. ], batch size: 247, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:57:02,643 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0328, 4.3486, 4.2155, 3.1726, 3.7636, 4.2304, 3.9160, 2.0846], device='cuda:5'), covar=tensor([0.0563, 0.0105, 0.0095, 0.0455, 0.0187, 0.0156, 0.0130, 0.0751], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0136, 0.0101, 0.0113, 0.0098, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 03:58:07,644 INFO [train.py:904] (5/8) Epoch 26, batch 2850, loss[loss=0.173, simple_loss=0.2491, pruned_loss=0.04845, over 16477.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2555, pruned_loss=0.03795, over 3312756.41 frames. ], batch size: 146, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:58:21,232 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256612.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:00,663 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256641.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:09,188 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9777, 2.2072, 2.7682, 2.9872, 2.8517, 3.4790, 2.4255, 3.5035], device='cuda:5'), covar=tensor([0.0327, 0.0573, 0.0379, 0.0396, 0.0416, 0.0240, 0.0517, 0.0226], device='cuda:5'), in_proj_covar=tensor([0.0197, 0.0199, 0.0188, 0.0191, 0.0206, 0.0166, 0.0203, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 03:59:15,040 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.116e+02 2.591e+02 2.981e+02 5.904e+02, threshold=5.182e+02, percent-clipped=1.0 2023-05-02 03:59:16,865 INFO [train.py:904] (5/8) Epoch 26, batch 2900, loss[loss=0.1828, simple_loss=0.2488, pruned_loss=0.05835, over 16272.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2544, pruned_loss=0.03832, over 3317906.59 frames. ], batch size: 165, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:59:27,996 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256660.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:29,244 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5395, 5.9729, 5.7066, 5.7822, 5.3447, 5.3741, 5.3189, 6.0625], device='cuda:5'), covar=tensor([0.1487, 0.0991, 0.1120, 0.0826, 0.0935, 0.0691, 0.1360, 0.0951], device='cuda:5'), in_proj_covar=tensor([0.0716, 0.0870, 0.0713, 0.0671, 0.0553, 0.0551, 0.0730, 0.0677], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:00:13,277 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256694.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:00:25,265 INFO [train.py:904] (5/8) Epoch 26, batch 2950, loss[loss=0.1712, simple_loss=0.2584, pruned_loss=0.04201, over 17210.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2545, pruned_loss=0.03851, over 3316031.11 frames. ], batch size: 45, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:00:33,193 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9033, 2.6900, 2.5607, 4.3770, 3.5009, 4.1332, 1.6367, 3.0096], device='cuda:5'), covar=tensor([0.1340, 0.0749, 0.1213, 0.0167, 0.0148, 0.0360, 0.1630, 0.0789], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0180, 0.0199, 0.0200, 0.0206, 0.0219, 0.0209, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:01:01,190 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256728.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:14,678 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:18,105 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256740.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:20,854 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256742.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:35,082 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.243e+02 2.622e+02 3.035e+02 4.998e+02, threshold=5.244e+02, percent-clipped=0.0 2023-05-02 04:01:35,103 INFO [train.py:904] (5/8) Epoch 26, batch 3000, loss[loss=0.184, simple_loss=0.2789, pruned_loss=0.04456, over 17079.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2551, pruned_loss=0.03946, over 3317219.91 frames. ], batch size: 55, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:01:35,103 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 04:01:44,002 INFO [train.py:938] (5/8) Epoch 26, validation: loss=0.1339, simple_loss=0.2392, pruned_loss=0.01435, over 944034.00 frames. 2023-05-02 04:01:44,003 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 04:02:03,769 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5227, 3.5515, 2.1868, 3.7975, 2.8469, 3.7014, 2.3834, 2.9344], device='cuda:5'), covar=tensor([0.0278, 0.0450, 0.1584, 0.0323, 0.0745, 0.0868, 0.1417, 0.0699], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0184, 0.0201, 0.0177, 0.0182, 0.0225, 0.0208, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:02:27,415 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,531 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,648 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:53,045 INFO [train.py:904] (5/8) Epoch 26, batch 3050, loss[loss=0.1462, simple_loss=0.2365, pruned_loss=0.02795, over 17177.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2549, pruned_loss=0.03897, over 3324697.68 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:03:19,810 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9300, 4.9075, 4.8342, 4.4608, 4.5272, 4.9118, 4.6520, 4.6225], device='cuda:5'), covar=tensor([0.0724, 0.0824, 0.0341, 0.0368, 0.0927, 0.0528, 0.0522, 0.0719], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0473, 0.0370, 0.0375, 0.0373, 0.0428, 0.0253, 0.0445], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 04:03:27,117 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256827.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:03:27,250 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256827.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:03:38,905 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:04:02,824 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.111e+02 2.504e+02 3.018e+02 1.623e+03, threshold=5.009e+02, percent-clipped=3.0 2023-05-02 04:04:02,848 INFO [train.py:904] (5/8) Epoch 26, batch 3100, loss[loss=0.1475, simple_loss=0.2365, pruned_loss=0.02923, over 16831.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2539, pruned_loss=0.03949, over 3328745.90 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:04:47,264 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256883.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:04:53,664 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256888.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:05:13,364 INFO [train.py:904] (5/8) Epoch 26, batch 3150, loss[loss=0.1529, simple_loss=0.2334, pruned_loss=0.03623, over 16815.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2533, pruned_loss=0.03962, over 3329095.34 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:06:05,139 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256941.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:06:09,384 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256944.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:06:21,405 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.162e+02 2.494e+02 3.116e+02 1.108e+03, threshold=4.988e+02, percent-clipped=4.0 2023-05-02 04:06:21,420 INFO [train.py:904] (5/8) Epoch 26, batch 3200, loss[loss=0.1729, simple_loss=0.2663, pruned_loss=0.03977, over 16410.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2523, pruned_loss=0.03885, over 3325116.29 frames. ], batch size: 68, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:06:36,854 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2193, 4.1933, 4.1488, 3.5589, 4.1882, 1.6301, 3.9680, 3.6340], device='cuda:5'), covar=tensor([0.0130, 0.0115, 0.0192, 0.0270, 0.0096, 0.3040, 0.0130, 0.0276], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0174, 0.0212, 0.0188, 0.0188, 0.0218, 0.0201, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:06:40,287 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 04:07:11,968 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:07:24,593 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 04:07:29,906 INFO [train.py:904] (5/8) Epoch 26, batch 3250, loss[loss=0.1682, simple_loss=0.2581, pruned_loss=0.03918, over 16767.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.253, pruned_loss=0.03933, over 3325356.85 frames. ], batch size: 83, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:07:43,723 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:08:03,439 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:08:38,568 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.275e+02 2.633e+02 3.156e+02 7.653e+02, threshold=5.267e+02, percent-clipped=3.0 2023-05-02 04:08:38,584 INFO [train.py:904] (5/8) Epoch 26, batch 3300, loss[loss=0.1468, simple_loss=0.241, pruned_loss=0.0263, over 17199.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2531, pruned_loss=0.03887, over 3334967.70 frames. ], batch size: 44, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:09:07,471 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257074.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:09,568 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257076.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:48,092 INFO [train.py:904] (5/8) Epoch 26, batch 3350, loss[loss=0.1731, simple_loss=0.2562, pruned_loss=0.04498, over 16688.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2539, pruned_loss=0.0394, over 3325445.81 frames. ], batch size: 89, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:10:20,827 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:10:56,576 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.179e+02 2.494e+02 3.258e+02 7.595e+02, threshold=4.988e+02, percent-clipped=3.0 2023-05-02 04:10:56,592 INFO [train.py:904] (5/8) Epoch 26, batch 3400, loss[loss=0.152, simple_loss=0.24, pruned_loss=0.03198, over 17223.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2546, pruned_loss=0.0394, over 3323023.28 frames. ], batch size: 45, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:11:13,559 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257165.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:27,561 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257175.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:39,285 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257183.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:11:44,309 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2654, 3.4611, 3.7696, 2.2519, 3.0795, 2.4801, 3.7326, 3.7148], device='cuda:5'), covar=tensor([0.0266, 0.0928, 0.0543, 0.2043, 0.0875, 0.1042, 0.0599, 0.0943], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0157, 0.0148, 0.0132, 0.0147, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 04:11:57,049 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3334, 5.2781, 5.0836, 4.5953, 5.1958, 2.0168, 4.9117, 5.0199], device='cuda:5'), covar=tensor([0.0100, 0.0100, 0.0240, 0.0430, 0.0103, 0.2906, 0.0158, 0.0216], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0188, 0.0189, 0.0219, 0.0202, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:12:06,426 INFO [train.py:904] (5/8) Epoch 26, batch 3450, loss[loss=0.1461, simple_loss=0.2266, pruned_loss=0.03286, over 15804.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2532, pruned_loss=0.03892, over 3319402.16 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:12:34,035 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:12:39,898 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:12:57,445 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257239.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:12:59,876 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:13:14,025 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7588, 4.5434, 4.7645, 4.9351, 5.1293, 4.5382, 5.1274, 5.1369], device='cuda:5'), covar=tensor([0.1844, 0.1333, 0.1975, 0.0987, 0.0612, 0.1223, 0.0784, 0.0671], device='cuda:5'), in_proj_covar=tensor([0.0696, 0.0854, 0.0989, 0.0867, 0.0657, 0.0689, 0.0715, 0.0832], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:13:16,457 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.059e+02 2.361e+02 2.793e+02 7.239e+02, threshold=4.722e+02, percent-clipped=1.0 2023-05-02 04:13:16,473 INFO [train.py:904] (5/8) Epoch 26, batch 3500, loss[loss=0.1603, simple_loss=0.2465, pruned_loss=0.03712, over 16779.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2526, pruned_loss=0.03862, over 3300703.66 frames. ], batch size: 102, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:13:58,945 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:25,674 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257302.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:26,380 INFO [train.py:904] (5/8) Epoch 26, batch 3550, loss[loss=0.1452, simple_loss=0.2287, pruned_loss=0.03087, over 17200.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2512, pruned_loss=0.03836, over 3308341.97 frames. ], batch size: 43, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:15:28,534 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2630, 5.2370, 4.9691, 4.4453, 5.0988, 1.8897, 4.7970, 4.8077], device='cuda:5'), covar=tensor([0.0108, 0.0096, 0.0264, 0.0440, 0.0119, 0.3071, 0.0191, 0.0272], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0174, 0.0212, 0.0188, 0.0189, 0.0218, 0.0202, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:15:34,859 INFO [train.py:904] (5/8) Epoch 26, batch 3600, loss[loss=0.1711, simple_loss=0.2519, pruned_loss=0.04509, over 16490.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2502, pruned_loss=0.03828, over 3305481.38 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:15:35,978 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 2.009e+02 2.326e+02 2.677e+02 5.900e+02, threshold=4.651e+02, percent-clipped=2.0 2023-05-02 04:15:47,152 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3204, 3.3745, 3.6060, 2.3561, 3.0854, 2.5546, 3.8015, 3.6997], device='cuda:5'), covar=tensor([0.0243, 0.0923, 0.0593, 0.1947, 0.0827, 0.0997, 0.0487, 0.0910], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 04:15:58,543 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257369.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:16:47,825 INFO [train.py:904] (5/8) Epoch 26, batch 3650, loss[loss=0.1568, simple_loss=0.2337, pruned_loss=0.03995, over 16801.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2489, pruned_loss=0.03878, over 3314405.71 frames. ], batch size: 96, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:58,063 INFO [train.py:904] (5/8) Epoch 26, batch 3700, loss[loss=0.1501, simple_loss=0.2291, pruned_loss=0.03552, over 16675.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2481, pruned_loss=0.04012, over 3305394.81 frames. ], batch size: 89, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:59,893 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.445e+02 2.076e+02 2.563e+02 3.144e+02 6.860e+02, threshold=5.126e+02, percent-clipped=4.0 2023-05-02 04:18:18,589 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:18:41,341 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257483.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:18:44,595 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8854, 1.9700, 2.4768, 2.7258, 2.7544, 2.7654, 2.0087, 3.0457], device='cuda:5'), covar=tensor([0.0187, 0.0533, 0.0369, 0.0313, 0.0340, 0.0313, 0.0617, 0.0183], device='cuda:5'), in_proj_covar=tensor([0.0198, 0.0199, 0.0188, 0.0192, 0.0206, 0.0166, 0.0203, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:19:09,137 INFO [train.py:904] (5/8) Epoch 26, batch 3750, loss[loss=0.1843, simple_loss=0.2709, pruned_loss=0.04887, over 16648.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2496, pruned_loss=0.04166, over 3291887.45 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:19:33,999 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:19:44,928 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257528.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:19:49,562 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257531.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:20:01,886 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257539.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:20:20,038 INFO [train.py:904] (5/8) Epoch 26, batch 3800, loss[loss=0.1863, simple_loss=0.258, pruned_loss=0.05727, over 16889.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2511, pruned_loss=0.04322, over 3284084.29 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:20:22,164 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.842e+02 2.272e+02 2.465e+02 2.848e+02 4.858e+02, threshold=4.931e+02, percent-clipped=0.0 2023-05-02 04:20:55,733 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:20:55,821 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257578.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:21:09,655 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257587.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:21:22,707 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:21:30,548 INFO [train.py:904] (5/8) Epoch 26, batch 3850, loss[loss=0.1473, simple_loss=0.2253, pruned_loss=0.03465, over 16773.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2511, pruned_loss=0.04371, over 3285161.41 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:22:20,395 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257639.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:22:41,230 INFO [train.py:904] (5/8) Epoch 26, batch 3900, loss[loss=0.1749, simple_loss=0.2599, pruned_loss=0.04492, over 16518.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2504, pruned_loss=0.04408, over 3286989.38 frames. ], batch size: 35, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:22:42,471 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.694e+02 2.236e+02 2.511e+02 3.000e+02 4.952e+02, threshold=5.021e+02, percent-clipped=1.0 2023-05-02 04:22:54,578 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-02 04:23:04,654 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257669.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:23:04,756 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7783, 2.9899, 3.2470, 2.0880, 2.8772, 2.2078, 3.3747, 3.3326], device='cuda:5'), covar=tensor([0.0254, 0.0874, 0.0607, 0.1987, 0.0854, 0.1068, 0.0555, 0.0874], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0157, 0.0148, 0.0132, 0.0147, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 04:23:51,422 INFO [train.py:904] (5/8) Epoch 26, batch 3950, loss[loss=0.1518, simple_loss=0.2343, pruned_loss=0.03463, over 16742.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2496, pruned_loss=0.04451, over 3295494.15 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:24:12,553 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:24:36,744 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4490, 4.4726, 4.7612, 4.7277, 4.7875, 4.5343, 4.5197, 4.3846], device='cuda:5'), covar=tensor([0.0360, 0.0713, 0.0427, 0.0454, 0.0553, 0.0439, 0.0787, 0.0719], device='cuda:5'), in_proj_covar=tensor([0.0442, 0.0498, 0.0482, 0.0442, 0.0531, 0.0508, 0.0588, 0.0407], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 04:25:02,904 INFO [train.py:904] (5/8) Epoch 26, batch 4000, loss[loss=0.1701, simple_loss=0.2578, pruned_loss=0.04123, over 16482.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2497, pruned_loss=0.04482, over 3284226.05 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:25:03,991 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.101e+02 2.446e+02 2.886e+02 5.630e+02, threshold=4.893e+02, percent-clipped=0.0 2023-05-02 04:25:41,160 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3400, 3.3911, 2.3887, 2.1095, 2.1572, 2.1335, 3.4051, 2.8906], device='cuda:5'), covar=tensor([0.3313, 0.0725, 0.2362, 0.3006, 0.3050, 0.2500, 0.0701, 0.1442], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0274, 0.0309, 0.0321, 0.0305, 0.0271, 0.0301, 0.0349], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 04:25:44,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3973, 5.7779, 5.3082, 5.6381, 5.1885, 4.9208, 5.2894, 5.8460], device='cuda:5'), covar=tensor([0.2253, 0.1155, 0.2025, 0.1272, 0.1616, 0.1252, 0.2174, 0.1400], device='cuda:5'), in_proj_covar=tensor([0.0721, 0.0875, 0.0715, 0.0677, 0.0556, 0.0555, 0.0736, 0.0684], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:26:13,229 INFO [train.py:904] (5/8) Epoch 26, batch 4050, loss[loss=0.1525, simple_loss=0.2393, pruned_loss=0.03288, over 16657.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2506, pruned_loss=0.04392, over 3284526.26 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:26:39,996 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:26:43,303 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257823.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:27:25,381 INFO [train.py:904] (5/8) Epoch 26, batch 4100, loss[loss=0.1925, simple_loss=0.2804, pruned_loss=0.05228, over 16392.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2523, pruned_loss=0.04337, over 3287467.86 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:27:26,543 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.750e+02 2.045e+02 2.393e+02 4.321e+02, threshold=4.090e+02, percent-clipped=1.0 2023-05-02 04:27:35,025 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257859.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:27:49,184 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257869.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:27:50,988 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4143, 4.6844, 4.5011, 4.4897, 4.1938, 4.1904, 4.1966, 4.7127], device='cuda:5'), covar=tensor([0.1163, 0.0830, 0.0978, 0.0804, 0.0825, 0.1382, 0.1067, 0.0803], device='cuda:5'), in_proj_covar=tensor([0.0720, 0.0875, 0.0714, 0.0676, 0.0556, 0.0555, 0.0736, 0.0683], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:28:02,788 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:10,492 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257883.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:30,716 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:39,737 INFO [train.py:904] (5/8) Epoch 26, batch 4150, loss[loss=0.1955, simple_loss=0.2771, pruned_loss=0.05694, over 16303.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2592, pruned_loss=0.04576, over 3253067.26 frames. ], batch size: 35, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:07,211 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:15,911 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:28,860 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257934.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:29:44,285 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257944.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:29:45,355 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:53,454 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:56,805 INFO [train.py:904] (5/8) Epoch 26, batch 4200, loss[loss=0.1837, simple_loss=0.2805, pruned_loss=0.04348, over 16285.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.266, pruned_loss=0.04753, over 3205112.03 frames. ], batch size: 165, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:58,474 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.083e+02 2.573e+02 2.968e+02 5.903e+02, threshold=5.147e+02, percent-clipped=5.0 2023-05-02 04:31:00,270 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6403, 3.9386, 4.2336, 2.6935, 3.5080, 2.7790, 4.0633, 4.1023], device='cuda:5'), covar=tensor([0.0240, 0.0692, 0.0480, 0.1687, 0.0740, 0.0882, 0.0547, 0.0788], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0156, 0.0148, 0.0132, 0.0146, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 04:31:03,090 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9693, 3.0813, 2.6060, 4.6858, 3.4792, 4.0297, 1.7055, 3.0762], device='cuda:5'), covar=tensor([0.1283, 0.0756, 0.1301, 0.0157, 0.0282, 0.0474, 0.1673, 0.0871], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0181, 0.0199, 0.0200, 0.0207, 0.0219, 0.0208, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:31:15,318 INFO [train.py:904] (5/8) Epoch 26, batch 4250, loss[loss=0.1729, simple_loss=0.2708, pruned_loss=0.03755, over 16616.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2688, pruned_loss=0.04682, over 3196179.11 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:31:29,176 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:31:46,774 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4684, 5.7683, 5.5574, 5.5953, 5.2995, 5.1129, 5.1990, 5.9331], device='cuda:5'), covar=tensor([0.1241, 0.0852, 0.0921, 0.0798, 0.0780, 0.0710, 0.1131, 0.0793], device='cuda:5'), in_proj_covar=tensor([0.0709, 0.0863, 0.0703, 0.0665, 0.0547, 0.0547, 0.0725, 0.0674], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:32:29,625 INFO [train.py:904] (5/8) Epoch 26, batch 4300, loss[loss=0.1748, simple_loss=0.2566, pruned_loss=0.04649, over 16047.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2705, pruned_loss=0.04611, over 3189397.72 frames. ], batch size: 35, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:32:31,424 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.134e+02 2.518e+02 2.964e+02 4.474e+02, threshold=5.035e+02, percent-clipped=0.0 2023-05-02 04:33:45,794 INFO [train.py:904] (5/8) Epoch 26, batch 4350, loss[loss=0.1915, simple_loss=0.2825, pruned_loss=0.05023, over 17050.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2743, pruned_loss=0.04779, over 3182440.07 frames. ], batch size: 55, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:34:15,471 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:34:47,993 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-02 04:34:58,991 INFO [train.py:904] (5/8) Epoch 26, batch 4400, loss[loss=0.202, simple_loss=0.294, pruned_loss=0.05496, over 16641.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2764, pruned_loss=0.04894, over 3193456.41 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:35:00,104 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.313e+02 2.752e+02 3.228e+02 5.813e+02, threshold=5.503e+02, percent-clipped=1.0 2023-05-02 04:35:13,027 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0821, 5.0712, 4.8736, 4.2155, 5.0312, 1.8989, 4.7568, 4.4186], device='cuda:5'), covar=tensor([0.0053, 0.0050, 0.0150, 0.0320, 0.0057, 0.3054, 0.0087, 0.0262], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0172, 0.0210, 0.0185, 0.0187, 0.0216, 0.0200, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:35:25,263 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258171.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:35:35,302 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 04:36:11,041 INFO [train.py:904] (5/8) Epoch 26, batch 4450, loss[loss=0.1868, simple_loss=0.2797, pruned_loss=0.04701, over 16752.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2803, pruned_loss=0.05043, over 3195766.50 frames. ], batch size: 76, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:36:28,176 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:36:55,421 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258234.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:37:03,968 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258239.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:37:22,669 INFO [train.py:904] (5/8) Epoch 26, batch 4500, loss[loss=0.1804, simple_loss=0.2697, pruned_loss=0.04552, over 16802.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2803, pruned_loss=0.05086, over 3190843.75 frames. ], batch size: 102, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:37:23,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 1.900e+02 2.188e+02 2.602e+02 4.588e+02, threshold=4.376e+02, percent-clipped=0.0 2023-05-02 04:38:05,642 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258282.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:38:35,273 INFO [train.py:904] (5/8) Epoch 26, batch 4550, loss[loss=0.2063, simple_loss=0.2764, pruned_loss=0.06807, over 11660.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2806, pruned_loss=0.05174, over 3190273.75 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:38:39,770 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:39:24,639 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-05-02 04:39:48,627 INFO [train.py:904] (5/8) Epoch 26, batch 4600, loss[loss=0.1878, simple_loss=0.2721, pruned_loss=0.05171, over 11713.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2818, pruned_loss=0.05225, over 3193561.34 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:39:50,257 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.394e+02 1.782e+02 2.088e+02 2.418e+02 3.840e+02, threshold=4.176e+02, percent-clipped=0.0 2023-05-02 04:39:53,665 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3900, 3.4076, 2.0587, 3.9315, 2.6544, 3.8820, 2.2466, 2.7630], device='cuda:5'), covar=tensor([0.0340, 0.0428, 0.1806, 0.0170, 0.0885, 0.0536, 0.1659, 0.0835], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0180, 0.0196, 0.0171, 0.0178, 0.0219, 0.0203, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:40:14,726 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7542, 3.7701, 2.4165, 4.6731, 3.0340, 4.5570, 2.5855, 3.0964], device='cuda:5'), covar=tensor([0.0337, 0.0466, 0.1759, 0.0171, 0.0813, 0.0455, 0.1570, 0.0842], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0180, 0.0196, 0.0171, 0.0178, 0.0219, 0.0203, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:40:24,135 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-02 04:41:03,106 INFO [train.py:904] (5/8) Epoch 26, batch 4650, loss[loss=0.1931, simple_loss=0.2856, pruned_loss=0.0503, over 16635.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.281, pruned_loss=0.05217, over 3200932.98 frames. ], batch size: 134, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:41:07,909 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1425, 3.1080, 2.6469, 3.1038, 3.4598, 3.1721, 3.6130, 3.5881], device='cuda:5'), covar=tensor([0.0066, 0.0355, 0.0464, 0.0331, 0.0224, 0.0310, 0.0186, 0.0223], device='cuda:5'), in_proj_covar=tensor([0.0229, 0.0243, 0.0232, 0.0234, 0.0243, 0.0243, 0.0243, 0.0243], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:41:19,162 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8478, 3.6405, 3.9978, 2.0399, 4.2484, 4.2603, 3.0666, 3.1960], device='cuda:5'), covar=tensor([0.0785, 0.0276, 0.0229, 0.1260, 0.0074, 0.0131, 0.0475, 0.0470], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0085, 0.0130, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:42:14,211 INFO [train.py:904] (5/8) Epoch 26, batch 4700, loss[loss=0.1695, simple_loss=0.2618, pruned_loss=0.0386, over 16803.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.278, pruned_loss=0.05111, over 3189648.35 frames. ], batch size: 96, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:42:16,007 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 1.737e+02 2.039e+02 2.431e+02 7.214e+02, threshold=4.078e+02, percent-clipped=1.0 2023-05-02 04:42:16,500 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:42:32,765 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5976, 4.5009, 4.6506, 4.8099, 4.9643, 4.5326, 4.9922, 5.0141], device='cuda:5'), covar=tensor([0.1907, 0.1206, 0.1622, 0.0756, 0.0553, 0.0829, 0.0527, 0.0584], device='cuda:5'), in_proj_covar=tensor([0.0666, 0.0815, 0.0940, 0.0827, 0.0626, 0.0654, 0.0679, 0.0790], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:42:57,775 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6146, 3.6148, 4.0376, 1.9583, 4.2208, 4.2369, 3.0522, 3.0968], device='cuda:5'), covar=tensor([0.0852, 0.0268, 0.0187, 0.1311, 0.0073, 0.0156, 0.0457, 0.0496], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0112, 0.0101, 0.0139, 0.0086, 0.0130, 0.0130, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:43:21,945 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0472, 4.1349, 3.9320, 3.6513, 3.6587, 4.0370, 3.7684, 3.8078], device='cuda:5'), covar=tensor([0.0563, 0.0604, 0.0306, 0.0298, 0.0768, 0.0532, 0.0947, 0.0532], device='cuda:5'), in_proj_covar=tensor([0.0307, 0.0457, 0.0358, 0.0362, 0.0360, 0.0414, 0.0244, 0.0431], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:43:26,806 INFO [train.py:904] (5/8) Epoch 26, batch 4750, loss[loss=0.1519, simple_loss=0.2437, pruned_loss=0.03002, over 16938.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2739, pruned_loss=0.04898, over 3203889.09 frames. ], batch size: 96, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:43:39,679 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:45,048 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:45,161 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:21,275 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258539.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:41,665 INFO [train.py:904] (5/8) Epoch 26, batch 4800, loss[loss=0.1848, simple_loss=0.2822, pruned_loss=0.04367, over 16426.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2701, pruned_loss=0.04691, over 3204452.81 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:44:43,333 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 1.748e+02 2.158e+02 2.549e+02 7.061e+02, threshold=4.316e+02, percent-clipped=3.0 2023-05-02 04:44:56,482 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258563.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:08,236 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7346, 2.7389, 2.5223, 4.4525, 3.0662, 3.9998, 1.5509, 2.9760], device='cuda:5'), covar=tensor([0.1394, 0.0817, 0.1251, 0.0152, 0.0194, 0.0400, 0.1717, 0.0800], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0179, 0.0197, 0.0198, 0.0206, 0.0216, 0.0207, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:45:11,481 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258572.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:34,407 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258587.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:45,121 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6853, 1.8531, 2.2671, 2.6244, 2.6211, 3.0011, 1.9460, 3.0138], device='cuda:5'), covar=tensor([0.0235, 0.0593, 0.0399, 0.0392, 0.0372, 0.0226, 0.0667, 0.0178], device='cuda:5'), in_proj_covar=tensor([0.0198, 0.0200, 0.0188, 0.0193, 0.0208, 0.0166, 0.0205, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:45:57,808 INFO [train.py:904] (5/8) Epoch 26, batch 4850, loss[loss=0.1707, simple_loss=0.2728, pruned_loss=0.03427, over 16688.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2709, pruned_loss=0.04607, over 3190143.83 frames. ], batch size: 134, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:46:03,276 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258606.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:47:14,827 INFO [train.py:904] (5/8) Epoch 26, batch 4900, loss[loss=0.1591, simple_loss=0.2526, pruned_loss=0.03278, over 16887.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2703, pruned_loss=0.04486, over 3175333.96 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:47:16,710 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.370e+02 1.994e+02 2.195e+02 2.536e+02 4.973e+02, threshold=4.389e+02, percent-clipped=2.0 2023-05-02 04:47:17,122 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:48:06,612 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 04:48:18,233 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2141, 5.4986, 5.2493, 5.3045, 5.0290, 4.9472, 4.8520, 5.6040], device='cuda:5'), covar=tensor([0.1244, 0.0785, 0.0971, 0.0878, 0.0740, 0.0847, 0.1250, 0.0847], device='cuda:5'), in_proj_covar=tensor([0.0699, 0.0847, 0.0692, 0.0652, 0.0538, 0.0538, 0.0713, 0.0664], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:48:28,161 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6314, 3.8542, 2.6845, 2.2504, 2.6525, 2.4813, 4.0211, 3.3391], device='cuda:5'), covar=tensor([0.3232, 0.0697, 0.2291, 0.2917, 0.2805, 0.2155, 0.0607, 0.1373], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0273, 0.0310, 0.0320, 0.0304, 0.0270, 0.0301, 0.0347], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 04:48:29,893 INFO [train.py:904] (5/8) Epoch 26, batch 4950, loss[loss=0.1818, simple_loss=0.2662, pruned_loss=0.04872, over 12132.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.27, pruned_loss=0.04428, over 3190140.73 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:48:30,491 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0062, 2.1794, 2.1798, 3.6068, 2.0874, 2.5167, 2.2504, 2.3081], device='cuda:5'), covar=tensor([0.1649, 0.3755, 0.3238, 0.0641, 0.4259, 0.2679, 0.3746, 0.3204], device='cuda:5'), in_proj_covar=tensor([0.0416, 0.0467, 0.0380, 0.0334, 0.0443, 0.0533, 0.0437, 0.0545], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:49:07,622 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7284, 3.6665, 4.2296, 2.0408, 4.4111, 4.4048, 3.1346, 3.1834], device='cuda:5'), covar=tensor([0.0790, 0.0278, 0.0162, 0.1230, 0.0053, 0.0103, 0.0418, 0.0486], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0085, 0.0129, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 04:49:41,073 INFO [train.py:904] (5/8) Epoch 26, batch 5000, loss[loss=0.1673, simple_loss=0.2599, pruned_loss=0.03734, over 17132.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2716, pruned_loss=0.04427, over 3199103.72 frames. ], batch size: 47, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:49:42,196 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.044e+02 2.396e+02 2.809e+02 5.347e+02, threshold=4.792e+02, percent-clipped=3.0 2023-05-02 04:49:54,239 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2077, 5.4718, 5.2144, 5.3055, 5.0729, 4.9323, 4.8392, 5.5988], device='cuda:5'), covar=tensor([0.1388, 0.0856, 0.1120, 0.0843, 0.0782, 0.0775, 0.1335, 0.0860], device='cuda:5'), in_proj_covar=tensor([0.0701, 0.0850, 0.0695, 0.0654, 0.0540, 0.0539, 0.0717, 0.0665], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 04:50:31,792 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 04:50:49,467 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 04:50:54,502 INFO [train.py:904] (5/8) Epoch 26, batch 5050, loss[loss=0.1741, simple_loss=0.2673, pruned_loss=0.04049, over 17225.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2721, pruned_loss=0.04432, over 3220618.08 frames. ], batch size: 45, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:51:05,562 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:52:07,199 INFO [train.py:904] (5/8) Epoch 26, batch 5100, loss[loss=0.192, simple_loss=0.2802, pruned_loss=0.05197, over 12051.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2707, pruned_loss=0.04372, over 3214580.02 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:52:08,932 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.999e+02 2.336e+02 2.714e+02 4.372e+02, threshold=4.672e+02, percent-clipped=0.0 2023-05-02 04:52:27,820 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:53:21,035 INFO [train.py:904] (5/8) Epoch 26, batch 5150, loss[loss=0.1738, simple_loss=0.2693, pruned_loss=0.03915, over 16878.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2703, pruned_loss=0.04306, over 3199881.31 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:54:27,563 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 04:54:35,100 INFO [train.py:904] (5/8) Epoch 26, batch 5200, loss[loss=0.1801, simple_loss=0.2708, pruned_loss=0.04472, over 16392.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2682, pruned_loss=0.04235, over 3217594.55 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:54:36,754 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 1.846e+02 2.126e+02 2.536e+02 3.966e+02, threshold=4.251e+02, percent-clipped=0.0 2023-05-02 04:54:53,534 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:55:16,382 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 04:55:48,690 INFO [train.py:904] (5/8) Epoch 26, batch 5250, loss[loss=0.1666, simple_loss=0.2593, pruned_loss=0.03692, over 16681.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2655, pruned_loss=0.04147, over 3228791.62 frames. ], batch size: 89, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:56:07,477 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8758, 4.8905, 5.2087, 5.1716, 5.2187, 4.9172, 4.8201, 4.7247], device='cuda:5'), covar=tensor([0.0296, 0.0553, 0.0332, 0.0367, 0.0413, 0.0341, 0.0988, 0.0455], device='cuda:5'), in_proj_covar=tensor([0.0420, 0.0471, 0.0458, 0.0421, 0.0507, 0.0483, 0.0561, 0.0389], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 04:56:23,045 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259026.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:57:03,178 INFO [train.py:904] (5/8) Epoch 26, batch 5300, loss[loss=0.2112, simple_loss=0.2956, pruned_loss=0.06338, over 12059.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2622, pruned_loss=0.04063, over 3231544.50 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:57:04,412 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 1.887e+02 2.180e+02 2.580e+02 4.921e+02, threshold=4.360e+02, percent-clipped=4.0 2023-05-02 04:57:06,022 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-02 04:57:56,259 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 04:58:18,030 INFO [train.py:904] (5/8) Epoch 26, batch 5350, loss[loss=0.191, simple_loss=0.2986, pruned_loss=0.0417, over 16339.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2613, pruned_loss=0.04017, over 3231271.28 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:58:27,838 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:31,426 INFO [train.py:904] (5/8) Epoch 26, batch 5400, loss[loss=0.2067, simple_loss=0.2891, pruned_loss=0.06211, over 12293.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.264, pruned_loss=0.04082, over 3217618.30 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:59:32,583 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 1.944e+02 2.298e+02 2.661e+02 4.475e+02, threshold=4.595e+02, percent-clipped=1.0 2023-05-02 04:59:38,450 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:52,965 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:00:17,422 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 05:00:48,654 INFO [train.py:904] (5/8) Epoch 26, batch 5450, loss[loss=0.1988, simple_loss=0.2916, pruned_loss=0.05304, over 16657.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2668, pruned_loss=0.04202, over 3212362.42 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:01:08,447 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:01:25,155 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:05,225 INFO [train.py:904] (5/8) Epoch 26, batch 5500, loss[loss=0.2272, simple_loss=0.31, pruned_loss=0.07217, over 16469.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2738, pruned_loss=0.04638, over 3174407.64 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:02:07,105 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.367e+02 2.972e+02 3.909e+02 6.600e+02, threshold=5.944e+02, percent-clipped=13.0 2023-05-02 05:02:23,224 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259264.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:58,265 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:22,302 INFO [train.py:904] (5/8) Epoch 26, batch 5550, loss[loss=0.1856, simple_loss=0.2782, pruned_loss=0.04648, over 16755.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2807, pruned_loss=0.05094, over 3154351.13 frames. ], batch size: 102, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:03:50,602 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:57,855 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259325.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:04:35,726 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 05:04:40,434 INFO [train.py:904] (5/8) Epoch 26, batch 5600, loss[loss=0.2629, simple_loss=0.3239, pruned_loss=0.1009, over 11628.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2854, pruned_loss=0.05489, over 3121473.76 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 16.0 2023-05-02 05:04:41,810 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 3.012e+02 3.705e+02 4.647e+02 1.087e+03, threshold=7.410e+02, percent-clipped=8.0 2023-05-02 05:05:12,999 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8407, 4.6859, 4.4328, 3.1141, 3.8787, 4.5666, 4.0092, 2.7110], device='cuda:5'), covar=tensor([0.0548, 0.0038, 0.0058, 0.0395, 0.0116, 0.0110, 0.0092, 0.0444], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0087, 0.0089, 0.0134, 0.0101, 0.0113, 0.0097, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 05:05:29,330 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.21 vs. limit=5.0 2023-05-02 05:06:04,647 INFO [train.py:904] (5/8) Epoch 26, batch 5650, loss[loss=0.1871, simple_loss=0.2748, pruned_loss=0.04966, over 16570.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2896, pruned_loss=0.05828, over 3103904.09 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:06:22,316 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259414.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:07:22,192 INFO [train.py:904] (5/8) Epoch 26, batch 5700, loss[loss=0.2433, simple_loss=0.316, pruned_loss=0.08532, over 11787.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2912, pruned_loss=0.06026, over 3074754.09 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:07:25,095 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 3.137e+02 3.722e+02 4.714e+02 8.155e+02, threshold=7.444e+02, percent-clipped=2.0 2023-05-02 05:07:56,966 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259475.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:08:39,982 INFO [train.py:904] (5/8) Epoch 26, batch 5750, loss[loss=0.2082, simple_loss=0.2997, pruned_loss=0.05832, over 16567.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2938, pruned_loss=0.06195, over 3048251.99 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:02,530 INFO [train.py:904] (5/8) Epoch 26, batch 5800, loss[loss=0.1935, simple_loss=0.2849, pruned_loss=0.05103, over 17060.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2928, pruned_loss=0.06026, over 3057015.56 frames. ], batch size: 55, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:05,675 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.780e+02 3.397e+02 4.110e+02 5.922e+02, threshold=6.793e+02, percent-clipped=0.0 2023-05-02 05:10:47,072 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:11:19,040 INFO [train.py:904] (5/8) Epoch 26, batch 5850, loss[loss=0.2147, simple_loss=0.2984, pruned_loss=0.06551, over 16320.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2912, pruned_loss=0.0589, over 3061635.23 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:11:44,966 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259620.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:11:46,986 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:12:39,964 INFO [train.py:904] (5/8) Epoch 26, batch 5900, loss[loss=0.1957, simple_loss=0.2839, pruned_loss=0.05374, over 15288.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2912, pruned_loss=0.05942, over 3060676.87 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:12:43,697 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.720e+02 3.451e+02 4.224e+02 7.867e+02, threshold=6.903e+02, percent-clipped=3.0 2023-05-02 05:13:00,401 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-05-02 05:13:06,338 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2705, 4.3710, 4.4852, 4.3151, 4.3774, 4.8587, 4.4439, 4.2231], device='cuda:5'), covar=tensor([0.1665, 0.1844, 0.2470, 0.2058, 0.2360, 0.1074, 0.1596, 0.2388], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0623, 0.0684, 0.0510, 0.0673, 0.0712, 0.0534, 0.0680], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 05:13:09,229 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259669.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:13:21,941 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5715, 5.6129, 5.3768, 4.9565, 4.9336, 5.4606, 5.3637, 5.1405], device='cuda:5'), covar=tensor([0.0828, 0.0977, 0.0368, 0.0466, 0.1265, 0.0622, 0.0527, 0.0992], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0455, 0.0355, 0.0359, 0.0357, 0.0413, 0.0242, 0.0429], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:13:26,973 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4221, 3.3967, 3.4508, 3.5297, 3.5704, 3.3334, 3.5384, 3.6204], device='cuda:5'), covar=tensor([0.1331, 0.0943, 0.1138, 0.0655, 0.0690, 0.2397, 0.1146, 0.0878], device='cuda:5'), in_proj_covar=tensor([0.0666, 0.0816, 0.0941, 0.0826, 0.0629, 0.0655, 0.0682, 0.0795], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:14:01,057 INFO [train.py:904] (5/8) Epoch 26, batch 5950, loss[loss=0.2261, simple_loss=0.3186, pruned_loss=0.06685, over 16563.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2925, pruned_loss=0.05862, over 3064198.21 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:14:03,342 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6554, 4.6549, 4.9866, 4.9568, 4.9740, 4.6989, 4.6693, 4.5422], device='cuda:5'), covar=tensor([0.0321, 0.0611, 0.0410, 0.0400, 0.0436, 0.0429, 0.0908, 0.0536], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0474, 0.0461, 0.0423, 0.0509, 0.0485, 0.0564, 0.0391], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 05:14:24,148 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3707, 3.4748, 3.6268, 3.6048, 3.6234, 3.4528, 3.4903, 3.5078], device='cuda:5'), covar=tensor([0.0425, 0.0759, 0.0501, 0.0488, 0.0492, 0.0579, 0.0761, 0.0590], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0473, 0.0460, 0.0422, 0.0508, 0.0484, 0.0562, 0.0391], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 05:14:29,214 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 05:15:12,639 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259749.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:15:18,044 INFO [train.py:904] (5/8) Epoch 26, batch 6000, loss[loss=0.187, simple_loss=0.278, pruned_loss=0.04801, over 16485.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.292, pruned_loss=0.05875, over 3056354.22 frames. ], batch size: 35, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:15:18,044 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 05:15:28,184 INFO [train.py:938] (5/8) Epoch 26, validation: loss=0.1485, simple_loss=0.2607, pruned_loss=0.01818, over 944034.00 frames. 2023-05-02 05:15:28,185 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 05:15:30,552 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 2.681e+02 3.578e+02 4.333e+02 8.656e+02, threshold=7.155e+02, percent-clipped=3.0 2023-05-02 05:15:55,091 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:13,721 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:46,441 INFO [train.py:904] (5/8) Epoch 26, batch 6050, loss[loss=0.1912, simple_loss=0.2863, pruned_loss=0.04807, over 16573.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2902, pruned_loss=0.05778, over 3071383.34 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:16:59,647 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259810.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:17:05,657 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3863, 4.4230, 4.7320, 4.6789, 4.7156, 4.4441, 4.3925, 4.3750], device='cuda:5'), covar=tensor([0.0363, 0.0709, 0.0439, 0.0462, 0.0453, 0.0459, 0.1018, 0.0559], device='cuda:5'), in_proj_covar=tensor([0.0420, 0.0473, 0.0460, 0.0421, 0.0508, 0.0484, 0.0561, 0.0389], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 05:17:51,096 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259843.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:18:02,923 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-05-02 05:18:05,782 INFO [train.py:904] (5/8) Epoch 26, batch 6100, loss[loss=0.1669, simple_loss=0.2561, pruned_loss=0.0388, over 16675.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2897, pruned_loss=0.05657, over 3084303.15 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:18:09,302 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.757e+02 3.100e+02 3.601e+02 8.920e+02, threshold=6.201e+02, percent-clipped=2.0 2023-05-02 05:18:27,897 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:18:52,990 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259882.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:19:23,809 INFO [train.py:904] (5/8) Epoch 26, batch 6150, loss[loss=0.2084, simple_loss=0.2833, pruned_loss=0.06675, over 11519.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2876, pruned_loss=0.05609, over 3078101.85 frames. ], batch size: 246, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:19:49,507 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:00,832 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:05,850 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:37,290 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 05:20:42,384 INFO [train.py:904] (5/8) Epoch 26, batch 6200, loss[loss=0.1843, simple_loss=0.2821, pruned_loss=0.04323, over 16548.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2867, pruned_loss=0.05628, over 3072293.83 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:20:44,634 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 2.750e+02 3.295e+02 3.920e+02 8.789e+02, threshold=6.590e+02, percent-clipped=2.0 2023-05-02 05:21:05,650 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259968.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:22:00,557 INFO [train.py:904] (5/8) Epoch 26, batch 6250, loss[loss=0.2156, simple_loss=0.3158, pruned_loss=0.05767, over 16599.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2866, pruned_loss=0.05607, over 3079009.34 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:22:41,231 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:23:15,632 INFO [train.py:904] (5/8) Epoch 26, batch 6300, loss[loss=0.1873, simple_loss=0.2826, pruned_loss=0.04603, over 16683.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2864, pruned_loss=0.05536, over 3090285.17 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:23:19,602 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.817e+02 3.510e+02 4.108e+02 7.742e+02, threshold=7.020e+02, percent-clipped=4.0 2023-05-02 05:23:32,190 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5948, 5.9340, 5.5978, 5.7001, 5.3108, 5.3403, 5.2910, 6.0210], device='cuda:5'), covar=tensor([0.1314, 0.0808, 0.0982, 0.0902, 0.0915, 0.0658, 0.1260, 0.0906], device='cuda:5'), in_proj_covar=tensor([0.0707, 0.0852, 0.0702, 0.0658, 0.0541, 0.0540, 0.0719, 0.0667], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:23:44,904 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260070.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:17,942 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:35,603 INFO [train.py:904] (5/8) Epoch 26, batch 6350, loss[loss=0.1916, simple_loss=0.2835, pruned_loss=0.04987, over 16824.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2873, pruned_loss=0.0567, over 3070865.24 frames. ], batch size: 96, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:24:38,164 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-05-02 05:24:39,481 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260105.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:24:42,737 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9701, 2.0926, 2.2240, 3.4275, 2.0624, 2.3673, 2.2164, 2.2168], device='cuda:5'), covar=tensor([0.1428, 0.3640, 0.2912, 0.0676, 0.4372, 0.2477, 0.3580, 0.3471], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0466, 0.0380, 0.0333, 0.0443, 0.0533, 0.0437, 0.0544], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:24:59,118 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260118.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:25:03,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5590, 3.7412, 2.7714, 2.2923, 2.5117, 2.4311, 3.9184, 3.3839], device='cuda:5'), covar=tensor([0.2934, 0.0540, 0.1759, 0.2661, 0.2521, 0.2074, 0.0513, 0.1216], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0273, 0.0308, 0.0322, 0.0303, 0.0271, 0.0300, 0.0346], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 05:25:22,983 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8075, 1.3539, 1.7138, 1.6843, 1.7654, 1.8244, 1.6321, 1.7621], device='cuda:5'), covar=tensor([0.0238, 0.0396, 0.0223, 0.0272, 0.0282, 0.0179, 0.0460, 0.0158], device='cuda:5'), in_proj_covar=tensor([0.0195, 0.0196, 0.0185, 0.0189, 0.0205, 0.0162, 0.0202, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:25:30,305 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260138.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:25:52,115 INFO [train.py:904] (5/8) Epoch 26, batch 6400, loss[loss=0.1823, simple_loss=0.2745, pruned_loss=0.04507, over 16866.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2877, pruned_loss=0.05792, over 3058981.44 frames. ], batch size: 96, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:25:54,619 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.921e+02 3.424e+02 4.125e+02 9.297e+02, threshold=6.848e+02, percent-clipped=3.0 2023-05-02 05:26:57,133 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 05:27:08,968 INFO [train.py:904] (5/8) Epoch 26, batch 6450, loss[loss=0.1873, simple_loss=0.2733, pruned_loss=0.05061, over 16837.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2877, pruned_loss=0.05726, over 3071164.73 frames. ], batch size: 116, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:27:39,943 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:27:43,835 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260224.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:27:50,745 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3928, 2.6563, 2.1326, 2.4286, 2.9532, 2.5440, 3.0422, 3.1552], device='cuda:5'), covar=tensor([0.0141, 0.0422, 0.0541, 0.0448, 0.0256, 0.0408, 0.0214, 0.0256], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0239, 0.0229, 0.0230, 0.0239, 0.0238, 0.0238, 0.0237], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:28:28,525 INFO [train.py:904] (5/8) Epoch 26, batch 6500, loss[loss=0.1881, simple_loss=0.2749, pruned_loss=0.05069, over 16809.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2858, pruned_loss=0.05658, over 3076870.64 frames. ], batch size: 124, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:28:31,537 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.830e+02 3.427e+02 3.864e+02 5.518e+02, threshold=6.855e+02, percent-clipped=0.0 2023-05-02 05:29:20,114 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260285.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:29:49,662 INFO [train.py:904] (5/8) Epoch 26, batch 6550, loss[loss=0.208, simple_loss=0.2946, pruned_loss=0.06076, over 16459.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2881, pruned_loss=0.05685, over 3096476.84 frames. ], batch size: 35, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:31:07,561 INFO [train.py:904] (5/8) Epoch 26, batch 6600, loss[loss=0.1933, simple_loss=0.2763, pruned_loss=0.05515, over 16725.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2899, pruned_loss=0.05721, over 3091568.45 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:31:09,947 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.646e+02 3.224e+02 4.276e+02 9.076e+02, threshold=6.447e+02, percent-clipped=2.0 2023-05-02 05:31:12,789 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-02 05:31:23,539 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-02 05:31:42,318 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5493, 3.7595, 2.8324, 2.2620, 2.4435, 2.3912, 4.0629, 3.3351], device='cuda:5'), covar=tensor([0.3150, 0.0678, 0.1870, 0.2811, 0.2843, 0.2183, 0.0440, 0.1347], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0273, 0.0308, 0.0321, 0.0302, 0.0270, 0.0300, 0.0345], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 05:31:59,523 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:32:26,386 INFO [train.py:904] (5/8) Epoch 26, batch 6650, loss[loss=0.1882, simple_loss=0.2708, pruned_loss=0.05279, over 16785.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2908, pruned_loss=0.0588, over 3076880.68 frames. ], batch size: 39, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:32:30,313 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260405.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:32:53,471 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1723, 3.2997, 3.0167, 5.2737, 4.1650, 4.3649, 2.0245, 3.2496], device='cuda:5'), covar=tensor([0.1197, 0.0712, 0.1131, 0.0181, 0.0367, 0.0435, 0.1500, 0.0787], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0181, 0.0200, 0.0200, 0.0209, 0.0219, 0.0209, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 05:33:21,083 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:33:43,223 INFO [train.py:904] (5/8) Epoch 26, batch 6700, loss[loss=0.1991, simple_loss=0.2878, pruned_loss=0.05524, over 15230.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2886, pruned_loss=0.05795, over 3085223.44 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:33:43,550 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260453.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:33:45,910 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.097e+02 2.665e+02 3.203e+02 3.728e+02 7.997e+02, threshold=6.406e+02, percent-clipped=3.0 2023-05-02 05:34:35,492 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260486.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:35:01,141 INFO [train.py:904] (5/8) Epoch 26, batch 6750, loss[loss=0.2136, simple_loss=0.2991, pruned_loss=0.06409, over 15343.00 frames. ], tot_loss[loss=0.202, simple_loss=0.288, pruned_loss=0.05804, over 3097670.54 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:35:31,759 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260522.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:36:19,269 INFO [train.py:904] (5/8) Epoch 26, batch 6800, loss[loss=0.1824, simple_loss=0.2747, pruned_loss=0.045, over 17101.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2881, pruned_loss=0.05757, over 3113788.23 frames. ], batch size: 49, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:36:21,473 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.811e+02 2.923e+02 3.478e+02 4.093e+02 6.667e+02, threshold=6.957e+02, percent-clipped=2.0 2023-05-02 05:36:45,309 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:37:01,666 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260580.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:37:26,342 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4501, 1.6249, 2.1116, 2.2999, 2.3833, 2.6350, 1.8688, 2.5932], device='cuda:5'), covar=tensor([0.0239, 0.0580, 0.0353, 0.0376, 0.0369, 0.0259, 0.0617, 0.0179], device='cuda:5'), in_proj_covar=tensor([0.0192, 0.0195, 0.0183, 0.0188, 0.0203, 0.0161, 0.0200, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:37:35,990 INFO [train.py:904] (5/8) Epoch 26, batch 6850, loss[loss=0.1994, simple_loss=0.3079, pruned_loss=0.04545, over 16680.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2895, pruned_loss=0.05783, over 3105688.71 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:50,110 INFO [train.py:904] (5/8) Epoch 26, batch 6900, loss[loss=0.242, simple_loss=0.3194, pruned_loss=0.08233, over 15463.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.292, pruned_loss=0.05813, over 3109253.61 frames. ], batch size: 191, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:53,864 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.543e+02 3.098e+02 3.712e+02 7.299e+02, threshold=6.197e+02, percent-clipped=1.0 2023-05-02 05:39:40,184 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260686.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:40:05,855 INFO [train.py:904] (5/8) Epoch 26, batch 6950, loss[loss=0.1973, simple_loss=0.2871, pruned_loss=0.05374, over 16687.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.293, pruned_loss=0.05896, over 3117556.90 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:40:48,892 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260730.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:40:54,451 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260734.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:21,302 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:23,890 INFO [train.py:904] (5/8) Epoch 26, batch 7000, loss[loss=0.2025, simple_loss=0.2993, pruned_loss=0.05289, over 17040.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2926, pruned_loss=0.05822, over 3109777.68 frames. ], batch size: 55, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:41:29,418 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.843e+02 3.539e+02 4.501e+02 8.180e+02, threshold=7.079e+02, percent-clipped=4.0 2023-05-02 05:42:24,294 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260791.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:42:42,011 INFO [train.py:904] (5/8) Epoch 26, batch 7050, loss[loss=0.2253, simple_loss=0.2987, pruned_loss=0.07594, over 11227.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2931, pruned_loss=0.05756, over 3114597.07 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:42:56,351 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:43:44,101 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5066, 4.6412, 4.8304, 4.5562, 4.6658, 5.1866, 4.7104, 4.4364], device='cuda:5'), covar=tensor([0.1513, 0.2127, 0.2465, 0.2096, 0.2490, 0.1086, 0.1723, 0.2489], device='cuda:5'), in_proj_covar=tensor([0.0427, 0.0634, 0.0695, 0.0515, 0.0683, 0.0723, 0.0544, 0.0691], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 05:43:59,445 INFO [train.py:904] (5/8) Epoch 26, batch 7100, loss[loss=0.1769, simple_loss=0.2693, pruned_loss=0.04223, over 16855.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2913, pruned_loss=0.05749, over 3101027.94 frames. ], batch size: 42, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:44:05,370 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.003e+02 2.616e+02 2.972e+02 3.634e+02 7.546e+02, threshold=5.943e+02, percent-clipped=1.0 2023-05-02 05:44:11,563 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1126, 3.3248, 3.5969, 2.1446, 3.0612, 2.3108, 3.5517, 3.6559], device='cuda:5'), covar=tensor([0.0287, 0.0936, 0.0603, 0.2183, 0.0918, 0.1025, 0.0670, 0.1041], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0171, 0.0172, 0.0158, 0.0149, 0.0133, 0.0148, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 05:44:24,723 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 05:44:42,735 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260880.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:44:44,388 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0679, 3.8713, 4.0311, 4.2582, 4.3276, 4.0531, 4.3507, 4.3789], device='cuda:5'), covar=tensor([0.2020, 0.1672, 0.1927, 0.0992, 0.0987, 0.1430, 0.1114, 0.1051], device='cuda:5'), in_proj_covar=tensor([0.0653, 0.0806, 0.0929, 0.0816, 0.0627, 0.0646, 0.0678, 0.0791], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:44:44,432 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:45:17,433 INFO [train.py:904] (5/8) Epoch 26, batch 7150, loss[loss=0.1793, simple_loss=0.2774, pruned_loss=0.04067, over 16679.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2893, pruned_loss=0.05723, over 3099220.91 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:45:31,013 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7984, 3.7631, 3.9472, 3.6587, 3.8489, 4.2454, 3.9415, 3.6634], device='cuda:5'), covar=tensor([0.2134, 0.2233, 0.2465, 0.2358, 0.2435, 0.1578, 0.1698, 0.2432], device='cuda:5'), in_proj_covar=tensor([0.0425, 0.0631, 0.0693, 0.0515, 0.0682, 0.0721, 0.0542, 0.0688], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 05:45:53,519 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260928.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:46:01,457 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 05:46:12,611 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2612, 2.1042, 1.8002, 1.9154, 2.3451, 2.0578, 1.9622, 2.4853], device='cuda:5'), covar=tensor([0.0232, 0.0485, 0.0637, 0.0550, 0.0338, 0.0456, 0.0236, 0.0316], device='cuda:5'), in_proj_covar=tensor([0.0222, 0.0238, 0.0229, 0.0230, 0.0240, 0.0237, 0.0237, 0.0237], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:46:13,598 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260942.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:46:29,367 INFO [train.py:904] (5/8) Epoch 26, batch 7200, loss[loss=0.1778, simple_loss=0.2719, pruned_loss=0.04183, over 16660.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2874, pruned_loss=0.05617, over 3076079.63 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:46:35,555 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.782e+02 3.336e+02 4.472e+02 6.403e+02, threshold=6.673e+02, percent-clipped=2.0 2023-05-02 05:47:07,472 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-05-02 05:47:48,430 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3356, 4.3198, 4.2098, 3.0010, 4.2933, 1.6529, 3.9521, 3.7845], device='cuda:5'), covar=tensor([0.0176, 0.0165, 0.0273, 0.0659, 0.0156, 0.3739, 0.0217, 0.0429], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0170, 0.0209, 0.0183, 0.0184, 0.0215, 0.0197, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:47:53,036 INFO [train.py:904] (5/8) Epoch 26, batch 7250, loss[loss=0.1819, simple_loss=0.2654, pruned_loss=0.04925, over 16808.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2846, pruned_loss=0.05461, over 3083939.75 frames. ], batch size: 116, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:48:49,249 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7940, 2.4487, 2.2600, 3.3438, 2.3008, 3.5375, 1.4983, 2.6812], device='cuda:5'), covar=tensor([0.1402, 0.0838, 0.1352, 0.0205, 0.0199, 0.0429, 0.1778, 0.0869], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0179, 0.0198, 0.0198, 0.0207, 0.0217, 0.0207, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 05:48:55,110 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5245, 3.5177, 3.4768, 2.7559, 3.4007, 2.0814, 3.2169, 2.7976], device='cuda:5'), covar=tensor([0.0175, 0.0155, 0.0208, 0.0259, 0.0114, 0.2512, 0.0141, 0.0262], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0170, 0.0209, 0.0184, 0.0184, 0.0215, 0.0197, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:49:09,261 INFO [train.py:904] (5/8) Epoch 26, batch 7300, loss[loss=0.2424, simple_loss=0.3157, pruned_loss=0.08453, over 11181.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2839, pruned_loss=0.05473, over 3074802.21 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:49:15,975 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.610e+02 3.123e+02 3.831e+02 8.496e+02, threshold=6.245e+02, percent-clipped=1.0 2023-05-02 05:49:23,384 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 05:49:28,197 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-02 05:49:48,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2940, 5.5723, 5.2831, 5.3195, 5.0566, 4.9449, 4.9994, 5.6426], device='cuda:5'), covar=tensor([0.1233, 0.0731, 0.1337, 0.0897, 0.0778, 0.0845, 0.1236, 0.0830], device='cuda:5'), in_proj_covar=tensor([0.0695, 0.0832, 0.0687, 0.0643, 0.0531, 0.0532, 0.0704, 0.0652], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:50:00,913 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:50:03,893 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3434, 3.2422, 3.3201, 3.4444, 3.4499, 3.2555, 3.4053, 3.5045], device='cuda:5'), covar=tensor([0.1176, 0.1089, 0.1219, 0.0699, 0.0801, 0.2723, 0.1373, 0.0921], device='cuda:5'), in_proj_covar=tensor([0.0654, 0.0806, 0.0931, 0.0816, 0.0626, 0.0648, 0.0678, 0.0789], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 05:50:26,308 INFO [train.py:904] (5/8) Epoch 26, batch 7350, loss[loss=0.1861, simple_loss=0.2795, pruned_loss=0.04635, over 16820.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2853, pruned_loss=0.05585, over 3058184.21 frames. ], batch size: 83, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:50:33,385 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261107.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:51:18,042 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261136.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:51:28,077 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-05-02 05:51:34,813 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4217, 3.2976, 3.7604, 1.7257, 3.8792, 3.9198, 2.9793, 2.8124], device='cuda:5'), covar=tensor([0.0883, 0.0311, 0.0199, 0.1407, 0.0091, 0.0191, 0.0454, 0.0537], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0140, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 05:51:44,998 INFO [train.py:904] (5/8) Epoch 26, batch 7400, loss[loss=0.2048, simple_loss=0.2948, pruned_loss=0.0574, over 16589.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2864, pruned_loss=0.05625, over 3058322.68 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:51:50,755 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.938e+02 2.670e+02 3.232e+02 3.696e+02 7.635e+02, threshold=6.463e+02, percent-clipped=2.0 2023-05-02 05:52:11,670 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7558, 3.8650, 2.5134, 4.5399, 3.0964, 4.4520, 2.5934, 3.1525], device='cuda:5'), covar=tensor([0.0298, 0.0388, 0.1616, 0.0187, 0.0764, 0.0516, 0.1457, 0.0761], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0180, 0.0197, 0.0170, 0.0179, 0.0219, 0.0206, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 05:52:34,613 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9810, 5.4361, 5.6402, 5.3246, 5.4016, 5.9242, 5.4521, 5.1915], device='cuda:5'), covar=tensor([0.0973, 0.1781, 0.2506, 0.1813, 0.2165, 0.0896, 0.1561, 0.2302], device='cuda:5'), in_proj_covar=tensor([0.0422, 0.0627, 0.0687, 0.0510, 0.0677, 0.0717, 0.0537, 0.0685], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 05:52:55,006 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261197.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:53:04,382 INFO [train.py:904] (5/8) Epoch 26, batch 7450, loss[loss=0.1921, simple_loss=0.2731, pruned_loss=0.05553, over 17022.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2871, pruned_loss=0.05675, over 3081056.21 frames. ], batch size: 55, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:54:01,693 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:54:27,107 INFO [train.py:904] (5/8) Epoch 26, batch 7500, loss[loss=0.1815, simple_loss=0.2625, pruned_loss=0.05023, over 16639.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2871, pruned_loss=0.05616, over 3082846.73 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:54:33,503 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.815e+02 3.421e+02 3.855e+02 8.023e+02, threshold=6.841e+02, percent-clipped=2.0 2023-05-02 05:55:32,727 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-02 05:55:45,652 INFO [train.py:904] (5/8) Epoch 26, batch 7550, loss[loss=0.1872, simple_loss=0.2762, pruned_loss=0.04913, over 16431.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.286, pruned_loss=0.05622, over 3093842.39 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:55:55,749 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5082, 4.2070, 4.1316, 2.5288, 3.7085, 4.2442, 3.6491, 2.2848], device='cuda:5'), covar=tensor([0.0566, 0.0052, 0.0063, 0.0479, 0.0111, 0.0118, 0.0116, 0.0510], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0135, 0.0101, 0.0113, 0.0098, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 05:56:43,741 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-05-02 05:56:48,166 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-05-02 05:57:01,750 INFO [train.py:904] (5/8) Epoch 26, batch 7600, loss[loss=0.1741, simple_loss=0.2733, pruned_loss=0.03741, over 16883.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2849, pruned_loss=0.0562, over 3096690.74 frames. ], batch size: 90, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 05:57:07,527 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.730e+02 2.718e+02 3.314e+02 3.844e+02 7.163e+02, threshold=6.628e+02, percent-clipped=1.0 2023-05-02 05:57:10,696 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 05:57:54,569 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:58:20,655 INFO [train.py:904] (5/8) Epoch 26, batch 7650, loss[loss=0.2178, simple_loss=0.3017, pruned_loss=0.06697, over 16738.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2862, pruned_loss=0.05716, over 3081505.44 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 05:58:26,599 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8239, 5.0813, 5.2691, 4.9679, 5.0882, 5.6513, 5.0867, 4.8134], device='cuda:5'), covar=tensor([0.1060, 0.1932, 0.2449, 0.1900, 0.2339, 0.0975, 0.1804, 0.2540], device='cuda:5'), in_proj_covar=tensor([0.0424, 0.0631, 0.0694, 0.0513, 0.0681, 0.0719, 0.0542, 0.0689], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 05:58:26,684 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:58:49,571 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8190, 2.7961, 2.8676, 2.1399, 2.7404, 2.1989, 2.6863, 2.9746], device='cuda:5'), covar=tensor([0.0260, 0.0800, 0.0466, 0.1753, 0.0750, 0.0899, 0.0567, 0.0723], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 05:59:08,103 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:32,231 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3429, 2.9180, 3.0033, 1.9892, 2.7266, 2.1239, 3.0217, 3.1510], device='cuda:5'), covar=tensor([0.0330, 0.0982, 0.0692, 0.2257, 0.0968, 0.1118, 0.0719, 0.0947], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 05:59:36,568 INFO [train.py:904] (5/8) Epoch 26, batch 7700, loss[loss=0.1995, simple_loss=0.2842, pruned_loss=0.05736, over 16652.00 frames. ], tot_loss[loss=0.2, simple_loss=0.286, pruned_loss=0.05695, over 3094331.25 frames. ], batch size: 57, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 05:59:40,110 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261455.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:42,190 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261456.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:46,568 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.911e+02 3.600e+02 4.177e+02 7.928e+02, threshold=7.200e+02, percent-clipped=2.0 2023-05-02 05:59:52,904 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 06:00:36,513 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261492.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 06:00:52,218 INFO [train.py:904] (5/8) Epoch 26, batch 7750, loss[loss=0.1924, simple_loss=0.2828, pruned_loss=0.051, over 16485.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2865, pruned_loss=0.05731, over 3083227.85 frames. ], batch size: 68, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:01:13,784 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:01:29,905 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7145, 6.0541, 5.7759, 5.8694, 5.4373, 5.3338, 5.4359, 6.1992], device='cuda:5'), covar=tensor([0.1250, 0.0831, 0.0964, 0.0841, 0.0870, 0.0685, 0.1356, 0.0828], device='cuda:5'), in_proj_covar=tensor([0.0698, 0.0837, 0.0690, 0.0647, 0.0534, 0.0536, 0.0708, 0.0657], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:01:45,024 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:02:09,790 INFO [train.py:904] (5/8) Epoch 26, batch 7800, loss[loss=0.2345, simple_loss=0.3197, pruned_loss=0.07462, over 16192.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2872, pruned_loss=0.05801, over 3070930.36 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:02:19,355 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.114e+02 2.737e+02 3.262e+02 3.975e+02 6.664e+02, threshold=6.524e+02, percent-clipped=0.0 2023-05-02 06:02:59,537 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4389, 3.5044, 2.7631, 2.1611, 2.2861, 2.3726, 3.6509, 3.1219], device='cuda:5'), covar=tensor([0.3148, 0.0646, 0.1910, 0.3023, 0.2892, 0.2267, 0.0489, 0.1502], device='cuda:5'), in_proj_covar=tensor([0.0334, 0.0274, 0.0311, 0.0324, 0.0305, 0.0273, 0.0302, 0.0350], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 06:03:00,429 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261585.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:03:28,326 INFO [train.py:904] (5/8) Epoch 26, batch 7850, loss[loss=0.1787, simple_loss=0.2775, pruned_loss=0.03997, over 16857.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2883, pruned_loss=0.05803, over 3072505.02 frames. ], batch size: 96, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:46,523 INFO [train.py:904] (5/8) Epoch 26, batch 7900, loss[loss=0.1961, simple_loss=0.2895, pruned_loss=0.05131, over 16777.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2872, pruned_loss=0.05692, over 3082890.49 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:55,567 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.825e+02 2.587e+02 3.177e+02 3.708e+02 5.885e+02, threshold=6.353e+02, percent-clipped=0.0 2023-05-02 06:05:12,998 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0965, 2.4027, 2.5312, 1.9242, 2.7069, 2.7910, 2.4009, 2.4231], device='cuda:5'), covar=tensor([0.0676, 0.0276, 0.0248, 0.0950, 0.0144, 0.0327, 0.0462, 0.0410], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0138, 0.0086, 0.0129, 0.0129, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 06:05:20,373 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9586, 2.7479, 2.8313, 2.1804, 2.7383, 2.1806, 2.7965, 2.9775], device='cuda:5'), covar=tensor([0.0289, 0.0812, 0.0535, 0.1769, 0.0775, 0.0932, 0.0580, 0.0730], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0168, 0.0169, 0.0155, 0.0146, 0.0131, 0.0145, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 06:06:05,695 INFO [train.py:904] (5/8) Epoch 26, batch 7950, loss[loss=0.1738, simple_loss=0.2621, pruned_loss=0.04271, over 16792.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.288, pruned_loss=0.05763, over 3073581.57 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:07:23,918 INFO [train.py:904] (5/8) Epoch 26, batch 8000, loss[loss=0.1971, simple_loss=0.2868, pruned_loss=0.05364, over 16844.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.288, pruned_loss=0.05734, over 3089572.49 frames. ], batch size: 116, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:07:32,672 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.545e+02 3.130e+02 3.720e+02 6.505e+02, threshold=6.260e+02, percent-clipped=1.0 2023-05-02 06:08:24,304 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261792.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 06:08:40,304 INFO [train.py:904] (5/8) Epoch 26, batch 8050, loss[loss=0.1783, simple_loss=0.2634, pruned_loss=0.04658, over 16711.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2876, pruned_loss=0.0568, over 3095040.61 frames. ], batch size: 57, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:08:54,195 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:13,071 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261824.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:27,652 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:37,859 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261840.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 06:09:57,233 INFO [train.py:904] (5/8) Epoch 26, batch 8100, loss[loss=0.1819, simple_loss=0.2684, pruned_loss=0.04769, over 16633.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2862, pruned_loss=0.05574, over 3105156.61 frames. ], batch size: 62, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:10:06,904 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.493e+02 2.989e+02 3.536e+02 5.601e+02, threshold=5.979e+02, percent-clipped=0.0 2023-05-02 06:10:46,810 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:11:00,412 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:11:14,763 INFO [train.py:904] (5/8) Epoch 26, batch 8150, loss[loss=0.1678, simple_loss=0.2636, pruned_loss=0.03606, over 16813.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.284, pruned_loss=0.05482, over 3120409.03 frames. ], batch size: 89, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:33,077 INFO [train.py:904] (5/8) Epoch 26, batch 8200, loss[loss=0.1794, simple_loss=0.2733, pruned_loss=0.04272, over 15442.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2822, pruned_loss=0.05457, over 3104472.63 frames. ], batch size: 190, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:43,207 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.684e+02 3.208e+02 4.137e+02 6.714e+02, threshold=6.416e+02, percent-clipped=2.0 2023-05-02 06:12:49,080 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 06:13:12,970 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261977.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:13:40,316 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4385, 3.2900, 3.5420, 1.9788, 3.6549, 3.7219, 3.0161, 2.9492], device='cuda:5'), covar=tensor([0.0820, 0.0266, 0.0224, 0.1152, 0.0110, 0.0252, 0.0403, 0.0430], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0110, 0.0100, 0.0139, 0.0086, 0.0130, 0.0130, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 06:13:58,699 INFO [train.py:904] (5/8) Epoch 26, batch 8250, loss[loss=0.1905, simple_loss=0.2854, pruned_loss=0.04785, over 16857.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.281, pruned_loss=0.05223, over 3094464.91 frames. ], batch size: 116, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:14:36,823 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-05-02 06:14:57,739 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262038.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:14:58,051 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 06:15:22,691 INFO [train.py:904] (5/8) Epoch 26, batch 8300, loss[loss=0.1795, simple_loss=0.2601, pruned_loss=0.04939, over 11769.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2788, pruned_loss=0.04967, over 3089606.48 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:15:32,831 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.264e+02 2.806e+02 3.280e+02 5.175e+02, threshold=5.611e+02, percent-clipped=0.0 2023-05-02 06:15:44,846 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7949, 5.0134, 5.1986, 4.9072, 4.9679, 5.5641, 5.0673, 4.7424], device='cuda:5'), covar=tensor([0.1086, 0.1977, 0.2299, 0.2016, 0.2669, 0.0937, 0.1513, 0.2409], device='cuda:5'), in_proj_covar=tensor([0.0419, 0.0618, 0.0682, 0.0507, 0.0670, 0.0709, 0.0534, 0.0679], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 06:16:41,866 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7611, 2.5943, 2.4255, 3.6451, 2.1315, 3.7916, 1.4554, 2.8538], device='cuda:5'), covar=tensor([0.1466, 0.0745, 0.1223, 0.0215, 0.0120, 0.0406, 0.1819, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0179, 0.0199, 0.0197, 0.0207, 0.0218, 0.0208, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 06:16:44,235 INFO [train.py:904] (5/8) Epoch 26, batch 8350, loss[loss=0.1658, simple_loss=0.2659, pruned_loss=0.03283, over 16914.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.278, pruned_loss=0.0478, over 3094130.33 frames. ], batch size: 96, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:16:59,091 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:17:14,887 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7907, 5.0682, 4.8807, 4.8733, 4.6278, 4.6220, 4.4448, 5.1620], device='cuda:5'), covar=tensor([0.1185, 0.0876, 0.0931, 0.0864, 0.0764, 0.0990, 0.1320, 0.0757], device='cuda:5'), in_proj_covar=tensor([0.0691, 0.0830, 0.0682, 0.0638, 0.0528, 0.0529, 0.0698, 0.0648], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:18:04,038 INFO [train.py:904] (5/8) Epoch 26, batch 8400, loss[loss=0.1828, simple_loss=0.2764, pruned_loss=0.04456, over 16521.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2754, pruned_loss=0.04613, over 3073037.78 frames. ], batch size: 68, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:18:13,166 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.393e+02 2.671e+02 3.015e+02 4.313e+02, threshold=5.342e+02, percent-clipped=0.0 2023-05-02 06:18:14,934 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:18:22,590 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-02 06:18:48,024 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262180.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:02,343 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:18,962 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8461, 1.4314, 1.7615, 1.7750, 1.9184, 1.9637, 1.8045, 1.8778], device='cuda:5'), covar=tensor([0.0282, 0.0437, 0.0261, 0.0359, 0.0309, 0.0222, 0.0436, 0.0172], device='cuda:5'), in_proj_covar=tensor([0.0191, 0.0194, 0.0181, 0.0186, 0.0202, 0.0160, 0.0198, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:19:24,268 INFO [train.py:904] (5/8) Epoch 26, batch 8450, loss[loss=0.16, simple_loss=0.2588, pruned_loss=0.03059, over 16783.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2737, pruned_loss=0.04468, over 3054308.94 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:48,813 INFO [train.py:904] (5/8) Epoch 26, batch 8500, loss[loss=0.1534, simple_loss=0.2497, pruned_loss=0.02856, over 16715.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2706, pruned_loss=0.0425, over 3065587.80 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:50,965 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2787, 3.4120, 3.3762, 2.3939, 3.1275, 3.4114, 3.2015, 2.1137], device='cuda:5'), covar=tensor([0.0523, 0.0069, 0.0068, 0.0395, 0.0127, 0.0112, 0.0101, 0.0512], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0086, 0.0087, 0.0132, 0.0099, 0.0111, 0.0096, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 06:20:57,922 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.087e+02 2.550e+02 3.186e+02 5.088e+02, threshold=5.100e+02, percent-clipped=0.0 2023-05-02 06:21:34,902 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9943, 2.7644, 2.6240, 1.9937, 2.5967, 2.7729, 2.6371, 1.9566], device='cuda:5'), covar=tensor([0.0439, 0.0092, 0.0081, 0.0368, 0.0143, 0.0117, 0.0111, 0.0447], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0086, 0.0087, 0.0132, 0.0099, 0.0111, 0.0096, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 06:22:09,822 INFO [train.py:904] (5/8) Epoch 26, batch 8550, loss[loss=0.1894, simple_loss=0.2901, pruned_loss=0.04436, over 15369.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2685, pruned_loss=0.04152, over 3053676.83 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:07,972 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:23:17,808 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0574, 4.0265, 3.9313, 3.0948, 3.9858, 1.7963, 3.7846, 3.5127], device='cuda:5'), covar=tensor([0.0109, 0.0111, 0.0188, 0.0257, 0.0098, 0.2923, 0.0133, 0.0286], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0166, 0.0204, 0.0179, 0.0179, 0.0210, 0.0192, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:23:47,974 INFO [train.py:904] (5/8) Epoch 26, batch 8600, loss[loss=0.1693, simple_loss=0.2642, pruned_loss=0.03714, over 16407.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.269, pruned_loss=0.04064, over 3056697.87 frames. ], batch size: 68, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:59,713 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.293e+02 2.800e+02 3.467e+02 6.859e+02, threshold=5.600e+02, percent-clipped=2.0 2023-05-02 06:24:46,068 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0816, 1.4998, 1.9592, 2.0502, 2.1552, 2.3649, 1.7933, 2.3043], device='cuda:5'), covar=tensor([0.0347, 0.0620, 0.0372, 0.0410, 0.0409, 0.0291, 0.0614, 0.0189], device='cuda:5'), in_proj_covar=tensor([0.0192, 0.0194, 0.0182, 0.0186, 0.0201, 0.0160, 0.0199, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:25:02,040 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-05-02 06:25:27,335 INFO [train.py:904] (5/8) Epoch 26, batch 8650, loss[loss=0.1702, simple_loss=0.2707, pruned_loss=0.03488, over 16387.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2668, pruned_loss=0.03931, over 3046153.45 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:26:42,680 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7332, 2.8543, 2.4337, 4.2112, 2.6754, 4.0444, 1.6528, 3.0568], device='cuda:5'), covar=tensor([0.1404, 0.0725, 0.1255, 0.0164, 0.0109, 0.0348, 0.1671, 0.0659], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0177, 0.0196, 0.0194, 0.0203, 0.0214, 0.0205, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 06:27:03,672 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 06:27:13,557 INFO [train.py:904] (5/8) Epoch 26, batch 8700, loss[loss=0.1657, simple_loss=0.2567, pruned_loss=0.03736, over 16961.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2645, pruned_loss=0.03848, over 3057321.15 frames. ], batch size: 109, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:27:25,318 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 2.080e+02 2.583e+02 3.221e+02 6.009e+02, threshold=5.165e+02, percent-clipped=1.0 2023-05-02 06:28:02,823 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:28:20,421 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:28:48,936 INFO [train.py:904] (5/8) Epoch 26, batch 8750, loss[loss=0.1625, simple_loss=0.2542, pruned_loss=0.03537, over 12467.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2642, pruned_loss=0.03788, over 3055175.86 frames. ], batch size: 250, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:29:48,637 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:29:59,857 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8478, 2.2178, 1.8197, 2.0652, 2.5461, 2.2421, 2.1773, 2.6675], device='cuda:5'), covar=tensor([0.0184, 0.0488, 0.0651, 0.0540, 0.0335, 0.0429, 0.0223, 0.0309], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0233, 0.0225, 0.0224, 0.0234, 0.0232, 0.0231, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:30:05,695 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 06:30:09,757 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:40,893 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5093, 4.5132, 4.3375, 3.6870, 4.4180, 1.6947, 4.1260, 4.1191], device='cuda:5'), covar=tensor([0.0134, 0.0116, 0.0212, 0.0342, 0.0120, 0.2855, 0.0166, 0.0248], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0165, 0.0202, 0.0177, 0.0179, 0.0209, 0.0191, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:30:41,540 INFO [train.py:904] (5/8) Epoch 26, batch 8800, loss[loss=0.1754, simple_loss=0.27, pruned_loss=0.04033, over 16777.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.263, pruned_loss=0.03663, over 3071748.68 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:30:52,416 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.093e+02 2.466e+02 2.812e+02 4.660e+02, threshold=4.932e+02, percent-clipped=0.0 2023-05-02 06:32:25,988 INFO [train.py:904] (5/8) Epoch 26, batch 8850, loss[loss=0.1522, simple_loss=0.2489, pruned_loss=0.02768, over 12289.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2647, pruned_loss=0.03646, over 3024349.44 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:33:16,485 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 06:33:32,237 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262633.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:34:13,867 INFO [train.py:904] (5/8) Epoch 26, batch 8900, loss[loss=0.1644, simple_loss=0.2621, pruned_loss=0.03335, over 15413.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2651, pruned_loss=0.03606, over 3023470.81 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:34:26,822 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 2.190e+02 2.649e+02 3.221e+02 8.286e+02, threshold=5.298e+02, percent-clipped=4.0 2023-05-02 06:35:20,127 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262681.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:36:18,595 INFO [train.py:904] (5/8) Epoch 26, batch 8950, loss[loss=0.1454, simple_loss=0.2398, pruned_loss=0.02551, over 15545.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2646, pruned_loss=0.03633, over 3028816.92 frames. ], batch size: 192, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:37:53,181 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8199, 4.4965, 4.6780, 4.9594, 5.1249, 4.6511, 5.1795, 5.1533], device='cuda:5'), covar=tensor([0.2029, 0.1577, 0.2287, 0.1003, 0.0889, 0.1033, 0.0801, 0.0985], device='cuda:5'), in_proj_covar=tensor([0.0626, 0.0776, 0.0890, 0.0785, 0.0598, 0.0622, 0.0649, 0.0761], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:38:00,869 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=262749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:38:08,101 INFO [train.py:904] (5/8) Epoch 26, batch 9000, loss[loss=0.1524, simple_loss=0.2456, pruned_loss=0.02957, over 16695.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2613, pruned_loss=0.03494, over 3037001.43 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:38:08,102 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 06:38:15,006 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8943, 5.1853, 5.0277, 5.1007, 4.8440, 4.9467, 4.5363, 5.2271], device='cuda:5'), covar=tensor([0.1288, 0.0765, 0.0758, 0.0614, 0.0661, 0.0304, 0.1171, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0690, 0.0828, 0.0681, 0.0635, 0.0528, 0.0528, 0.0697, 0.0649], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:38:18,519 INFO [train.py:938] (5/8) Epoch 26, validation: loss=0.1435, simple_loss=0.2475, pruned_loss=0.01976, over 944034.00 frames. 2023-05-02 06:38:18,520 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 06:38:30,744 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.045e+02 2.519e+02 3.077e+02 5.140e+02, threshold=5.037e+02, percent-clipped=0.0 2023-05-02 06:39:15,620 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 06:40:03,328 INFO [train.py:904] (5/8) Epoch 26, batch 9050, loss[loss=0.1551, simple_loss=0.2483, pruned_loss=0.03097, over 16162.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2618, pruned_loss=0.03516, over 3052509.63 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:40:20,492 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:41:48,719 INFO [train.py:904] (5/8) Epoch 26, batch 9100, loss[loss=0.1696, simple_loss=0.2717, pruned_loss=0.03375, over 15522.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2619, pruned_loss=0.0357, over 3054408.79 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:42:01,225 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.262e+02 2.654e+02 3.227e+02 6.240e+02, threshold=5.309e+02, percent-clipped=4.0 2023-05-02 06:43:45,870 INFO [train.py:904] (5/8) Epoch 26, batch 9150, loss[loss=0.1488, simple_loss=0.2418, pruned_loss=0.02791, over 16533.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2623, pruned_loss=0.03514, over 3053931.44 frames. ], batch size: 68, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:44:04,194 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1451, 4.1698, 4.4468, 4.4429, 4.4532, 4.2469, 4.2234, 4.2008], device='cuda:5'), covar=tensor([0.0321, 0.0716, 0.0427, 0.0369, 0.0407, 0.0402, 0.0764, 0.0477], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0464, 0.0453, 0.0415, 0.0500, 0.0475, 0.0546, 0.0380], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 06:44:14,032 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 06:44:25,007 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-05-02 06:45:27,094 INFO [train.py:904] (5/8) Epoch 26, batch 9200, loss[loss=0.1804, simple_loss=0.2704, pruned_loss=0.0452, over 16850.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.258, pruned_loss=0.0341, over 3060870.74 frames. ], batch size: 116, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:45:36,601 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.229e+02 2.531e+02 3.050e+02 5.061e+02, threshold=5.062e+02, percent-clipped=0.0 2023-05-02 06:47:01,952 INFO [train.py:904] (5/8) Epoch 26, batch 9250, loss[loss=0.1484, simple_loss=0.2356, pruned_loss=0.03064, over 12654.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2576, pruned_loss=0.03433, over 3040104.06 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:47:22,760 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9554, 3.7504, 3.9603, 2.0341, 4.1522, 4.3400, 3.3572, 3.2976], device='cuda:5'), covar=tensor([0.0663, 0.0246, 0.0221, 0.1276, 0.0085, 0.0119, 0.0348, 0.0428], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0107, 0.0096, 0.0135, 0.0082, 0.0124, 0.0125, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 06:48:34,967 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6288, 3.6982, 3.4861, 3.1145, 3.2727, 3.5855, 3.3742, 3.4248], device='cuda:5'), covar=tensor([0.0592, 0.0552, 0.0314, 0.0263, 0.0434, 0.0467, 0.1378, 0.0482], device='cuda:5'), in_proj_covar=tensor([0.0296, 0.0439, 0.0343, 0.0345, 0.0341, 0.0397, 0.0236, 0.0413], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:48:49,723 INFO [train.py:904] (5/8) Epoch 26, batch 9300, loss[loss=0.1434, simple_loss=0.2292, pruned_loss=0.02883, over 12141.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2559, pruned_loss=0.03378, over 3037438.64 frames. ], batch size: 250, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:48:53,909 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-05-02 06:49:02,023 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 2.028e+02 2.420e+02 3.075e+02 6.199e+02, threshold=4.840e+02, percent-clipped=1.0 2023-05-02 06:50:20,370 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:33,222 INFO [train.py:904] (5/8) Epoch 26, batch 9350, loss[loss=0.1784, simple_loss=0.2708, pruned_loss=0.04298, over 16648.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2562, pruned_loss=0.03393, over 3037068.22 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:50:38,640 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263105.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:51,186 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:51:48,320 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:51:54,533 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 06:52:14,167 INFO [train.py:904] (5/8) Epoch 26, batch 9400, loss[loss=0.1698, simple_loss=0.2739, pruned_loss=0.03287, over 16714.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2564, pruned_loss=0.03378, over 3045739.56 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:52:19,903 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:52:25,032 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 2.115e+02 2.425e+02 3.007e+02 4.996e+02, threshold=4.851e+02, percent-clipped=1.0 2023-05-02 06:52:46,333 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1625, 1.6001, 1.9316, 2.0427, 2.1551, 2.3769, 1.8329, 2.3138], device='cuda:5'), covar=tensor([0.0365, 0.0582, 0.0373, 0.0438, 0.0427, 0.0256, 0.0564, 0.0210], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0192, 0.0179, 0.0183, 0.0199, 0.0157, 0.0196, 0.0157], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:52:48,222 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3782, 4.6705, 4.5158, 4.4539, 4.1674, 4.1670, 4.1326, 4.7265], device='cuda:5'), covar=tensor([0.1259, 0.0959, 0.0876, 0.0804, 0.0870, 0.1632, 0.1169, 0.0911], device='cuda:5'), in_proj_covar=tensor([0.0683, 0.0821, 0.0673, 0.0631, 0.0524, 0.0525, 0.0691, 0.0644], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 06:52:52,579 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:50,985 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:53,501 INFO [train.py:904] (5/8) Epoch 26, batch 9450, loss[loss=0.1803, simple_loss=0.2654, pruned_loss=0.04758, over 12518.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2581, pruned_loss=0.03427, over 3019155.25 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:54:12,275 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-02 06:54:14,356 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:54:24,681 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 06:55:19,940 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263246.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:55:33,114 INFO [train.py:904] (5/8) Epoch 26, batch 9500, loss[loss=0.1486, simple_loss=0.2425, pruned_loss=0.0274, over 16750.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2576, pruned_loss=0.03411, over 3011965.66 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:55:47,502 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 2.048e+02 2.318e+02 3.052e+02 5.561e+02, threshold=4.636e+02, percent-clipped=2.0 2023-05-02 06:56:14,940 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7007, 4.0453, 4.0958, 2.7974, 3.5367, 4.0125, 3.7545, 2.4122], device='cuda:5'), covar=tensor([0.0478, 0.0050, 0.0047, 0.0410, 0.0125, 0.0128, 0.0075, 0.0486], device='cuda:5'), in_proj_covar=tensor([0.0134, 0.0086, 0.0087, 0.0132, 0.0099, 0.0110, 0.0095, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 06:56:19,465 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:57:17,240 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 06:57:17,619 INFO [train.py:904] (5/8) Epoch 26, batch 9550, loss[loss=0.1946, simple_loss=0.2905, pruned_loss=0.04932, over 12384.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2568, pruned_loss=0.03377, over 3035718.05 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:57:27,279 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:58:58,572 INFO [train.py:904] (5/8) Epoch 26, batch 9600, loss[loss=0.1687, simple_loss=0.2539, pruned_loss=0.04174, over 12356.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2574, pruned_loss=0.03419, over 3009335.75 frames. ], batch size: 249, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:59:09,909 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.249e+02 2.699e+02 3.019e+02 5.664e+02, threshold=5.399e+02, percent-clipped=4.0 2023-05-02 07:00:45,648 INFO [train.py:904] (5/8) Epoch 26, batch 9650, loss[loss=0.1749, simple_loss=0.2729, pruned_loss=0.03844, over 15443.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.26, pruned_loss=0.03471, over 3018660.75 frames. ], batch size: 192, lr: 2.56e-03, grad_scale: 8.0 2023-05-02 07:00:52,308 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:29,135 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263451.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:31,835 INFO [train.py:904] (5/8) Epoch 26, batch 9700, loss[loss=0.1784, simple_loss=0.2676, pruned_loss=0.04463, over 16856.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2601, pruned_loss=0.03473, over 3052947.08 frames. ], batch size: 124, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:02:33,070 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:42,018 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.181e+02 2.600e+02 3.178e+02 8.821e+02, threshold=5.200e+02, percent-clipped=1.0 2023-05-02 07:02:58,827 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:04:02,105 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:04:13,675 INFO [train.py:904] (5/8) Epoch 26, batch 9750, loss[loss=0.164, simple_loss=0.2609, pruned_loss=0.03361, over 16340.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2595, pruned_loss=0.03507, over 3049691.56 frames. ], batch size: 165, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:04:23,940 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1455, 5.4152, 5.2410, 5.1939, 4.8961, 4.8508, 4.7237, 5.5283], device='cuda:5'), covar=tensor([0.1247, 0.0939, 0.0875, 0.0853, 0.0798, 0.0910, 0.1392, 0.0985], device='cuda:5'), in_proj_covar=tensor([0.0682, 0.0821, 0.0671, 0.0631, 0.0525, 0.0524, 0.0692, 0.0645], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:05:19,230 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8746, 3.8687, 4.1520, 4.1359, 4.1508, 3.9447, 3.9624, 3.9851], device='cuda:5'), covar=tensor([0.0344, 0.0716, 0.0436, 0.0417, 0.0455, 0.0487, 0.0807, 0.0410], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0462, 0.0448, 0.0412, 0.0496, 0.0472, 0.0542, 0.0377], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 07:05:51,111 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 07:05:51,432 INFO [train.py:904] (5/8) Epoch 26, batch 9800, loss[loss=0.1552, simple_loss=0.2459, pruned_loss=0.03223, over 12460.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2593, pruned_loss=0.03449, over 3047616.57 frames. ], batch size: 248, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:06:03,263 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 2.162e+02 2.515e+02 2.948e+02 4.356e+02, threshold=5.031e+02, percent-clipped=0.0 2023-05-02 07:06:23,918 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:06:29,543 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 07:07:33,457 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263602.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:07:34,250 INFO [train.py:904] (5/8) Epoch 26, batch 9850, loss[loss=0.1687, simple_loss=0.2631, pruned_loss=0.03722, over 16422.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2602, pruned_loss=0.034, over 3049124.67 frames. ], batch size: 147, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:00,059 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-02 07:09:24,116 INFO [train.py:904] (5/8) Epoch 26, batch 9900, loss[loss=0.159, simple_loss=0.2438, pruned_loss=0.03711, over 11982.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2611, pruned_loss=0.03405, over 3066672.80 frames. ], batch size: 246, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:36,803 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 1.934e+02 2.301e+02 2.936e+02 6.933e+02, threshold=4.602e+02, percent-clipped=3.0 2023-05-02 07:10:00,734 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9240, 4.2406, 4.1169, 4.1132, 3.7405, 3.8320, 3.8685, 4.2474], device='cuda:5'), covar=tensor([0.1224, 0.0922, 0.0866, 0.0765, 0.0862, 0.1646, 0.0976, 0.0938], device='cuda:5'), in_proj_covar=tensor([0.0679, 0.0817, 0.0668, 0.0628, 0.0522, 0.0522, 0.0688, 0.0642], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:11:01,223 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2203, 2.9040, 3.1373, 1.7510, 3.2689, 3.3742, 2.7182, 2.5764], device='cuda:5'), covar=tensor([0.0799, 0.0329, 0.0191, 0.1356, 0.0111, 0.0181, 0.0446, 0.0550], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0106, 0.0094, 0.0134, 0.0081, 0.0122, 0.0125, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 07:11:20,749 INFO [train.py:904] (5/8) Epoch 26, batch 9950, loss[loss=0.1901, simple_loss=0.2727, pruned_loss=0.05373, over 12266.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2639, pruned_loss=0.03441, over 3085701.29 frames. ], batch size: 250, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:11:27,955 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0152, 3.1674, 3.2382, 2.1598, 2.9212, 3.2382, 3.1104, 1.9652], device='cuda:5'), covar=tensor([0.0563, 0.0067, 0.0065, 0.0451, 0.0137, 0.0099, 0.0089, 0.0508], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0085, 0.0086, 0.0131, 0.0098, 0.0109, 0.0094, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 07:11:51,079 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4281, 2.0427, 1.8689, 1.7318, 2.3056, 1.9534, 1.8831, 2.3551], device='cuda:5'), covar=tensor([0.0212, 0.0484, 0.0519, 0.0544, 0.0281, 0.0431, 0.0240, 0.0302], device='cuda:5'), in_proj_covar=tensor([0.0215, 0.0235, 0.0226, 0.0226, 0.0236, 0.0234, 0.0231, 0.0231], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:12:24,278 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0332, 3.7734, 4.3230, 2.1340, 4.4261, 4.5554, 3.3060, 3.3304], device='cuda:5'), covar=tensor([0.0650, 0.0241, 0.0118, 0.1258, 0.0056, 0.0090, 0.0350, 0.0440], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0106, 0.0094, 0.0134, 0.0081, 0.0122, 0.0125, 0.0124], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 07:13:19,286 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:13:21,147 INFO [train.py:904] (5/8) Epoch 26, batch 10000, loss[loss=0.1557, simple_loss=0.2587, pruned_loss=0.0263, over 15428.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2625, pruned_loss=0.03394, over 3102370.00 frames. ], batch size: 191, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:34,913 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.135e+02 2.379e+02 2.761e+02 5.378e+02, threshold=4.758e+02, percent-clipped=3.0 2023-05-02 07:13:51,464 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:14:48,911 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263796.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:14:53,816 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:00,043 INFO [train.py:904] (5/8) Epoch 26, batch 10050, loss[loss=0.1791, simple_loss=0.2813, pruned_loss=0.03844, over 16174.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2626, pruned_loss=0.03409, over 3094732.69 frames. ], batch size: 165, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:15:23,692 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263815.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:32,634 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1245, 2.5131, 2.5983, 1.8852, 2.7676, 2.8971, 2.4361, 2.4452], device='cuda:5'), covar=tensor([0.0649, 0.0277, 0.0218, 0.1056, 0.0139, 0.0212, 0.0492, 0.0443], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0106, 0.0094, 0.0134, 0.0081, 0.0122, 0.0125, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 07:15:57,715 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263833.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:16:17,295 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:16:30,282 INFO [train.py:904] (5/8) Epoch 26, batch 10100, loss[loss=0.156, simple_loss=0.2398, pruned_loss=0.03616, over 12572.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2624, pruned_loss=0.03404, over 3082945.83 frames. ], batch size: 248, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:16:39,585 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.128e+02 2.539e+02 3.188e+02 8.403e+02, threshold=5.079e+02, percent-clipped=3.0 2023-05-02 07:17:02,805 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:17:10,643 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0450, 2.2898, 2.2825, 3.1086, 1.8918, 3.2718, 1.7553, 2.8531], device='cuda:5'), covar=tensor([0.1167, 0.0639, 0.1009, 0.0158, 0.0080, 0.0344, 0.1460, 0.0612], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0175, 0.0194, 0.0190, 0.0197, 0.0211, 0.0204, 0.0192], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:17:20,832 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 07:17:36,344 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263894.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 07:18:08,988 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263902.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:18:09,862 INFO [train.py:904] (5/8) Epoch 27, batch 0, loss[loss=0.1956, simple_loss=0.2788, pruned_loss=0.05625, over 16560.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2788, pruned_loss=0.05625, over 16560.00 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 16.0 2023-05-02 07:18:09,862 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 07:18:17,115 INFO [train.py:938] (5/8) Epoch 27, validation: loss=0.1434, simple_loss=0.2467, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 07:18:17,116 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 07:18:38,687 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:22,226 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:26,174 INFO [train.py:904] (5/8) Epoch 27, batch 50, loss[loss=0.1879, simple_loss=0.2741, pruned_loss=0.05085, over 16735.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2695, pruned_loss=0.04808, over 746851.27 frames. ], batch size: 134, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:19:36,261 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263960.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:39,892 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.464e+02 3.043e+02 3.721e+02 7.050e+02, threshold=6.086e+02, percent-clipped=6.0 2023-05-02 07:19:50,834 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9507, 4.5376, 3.1252, 2.3973, 2.7245, 2.5995, 4.8236, 3.6034], device='cuda:5'), covar=tensor([0.2914, 0.0536, 0.1833, 0.3062, 0.3017, 0.2295, 0.0346, 0.1491], device='cuda:5'), in_proj_covar=tensor([0.0326, 0.0268, 0.0306, 0.0317, 0.0293, 0.0268, 0.0297, 0.0340], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 07:20:36,420 INFO [train.py:904] (5/8) Epoch 27, batch 100, loss[loss=0.1872, simple_loss=0.2713, pruned_loss=0.0516, over 16295.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2622, pruned_loss=0.04464, over 1320200.40 frames. ], batch size: 165, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:01,041 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:21:10,219 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-02 07:21:36,092 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8278, 3.8049, 3.8987, 3.7157, 3.8938, 4.3171, 3.9327, 3.5595], device='cuda:5'), covar=tensor([0.2174, 0.2427, 0.2980, 0.2384, 0.2610, 0.1796, 0.1709, 0.2827], device='cuda:5'), in_proj_covar=tensor([0.0407, 0.0602, 0.0669, 0.0493, 0.0652, 0.0693, 0.0522, 0.0657], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 07:21:44,763 INFO [train.py:904] (5/8) Epoch 27, batch 150, loss[loss=0.1705, simple_loss=0.2604, pruned_loss=0.04032, over 16874.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2614, pruned_loss=0.04421, over 1759785.02 frames. ], batch size: 102, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:57,550 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.228e+02 2.541e+02 3.123e+02 6.802e+02, threshold=5.083e+02, percent-clipped=1.0 2023-05-02 07:22:06,391 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5503, 4.3303, 4.6297, 4.7641, 4.8855, 4.3948, 4.7894, 4.8699], device='cuda:5'), covar=tensor([0.1899, 0.1528, 0.1531, 0.0825, 0.0686, 0.1149, 0.1769, 0.0833], device='cuda:5'), in_proj_covar=tensor([0.0645, 0.0797, 0.0913, 0.0804, 0.0613, 0.0637, 0.0670, 0.0777], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:22:36,115 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2183, 2.1746, 2.8418, 3.2451, 3.0649, 3.6759, 2.2150, 3.7075], device='cuda:5'), covar=tensor([0.0267, 0.0678, 0.0354, 0.0333, 0.0339, 0.0215, 0.0756, 0.0172], device='cuda:5'), in_proj_covar=tensor([0.0191, 0.0194, 0.0181, 0.0186, 0.0202, 0.0159, 0.0199, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:22:53,639 INFO [train.py:904] (5/8) Epoch 27, batch 200, loss[loss=0.1821, simple_loss=0.2757, pruned_loss=0.04429, over 16740.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2613, pruned_loss=0.04379, over 2109445.78 frames. ], batch size: 57, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:23:13,839 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 07:23:34,803 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4697, 4.5497, 4.7181, 4.4828, 4.5604, 5.1498, 4.6253, 4.2672], device='cuda:5'), covar=tensor([0.1792, 0.2262, 0.2732, 0.2325, 0.2759, 0.1217, 0.1897, 0.2756], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0607, 0.0673, 0.0497, 0.0656, 0.0699, 0.0525, 0.0662], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 07:24:00,844 INFO [train.py:904] (5/8) Epoch 27, batch 250, loss[loss=0.1766, simple_loss=0.2543, pruned_loss=0.04943, over 16734.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2601, pruned_loss=0.04396, over 2381412.50 frames. ], batch size: 124, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:24:14,014 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.137e+02 2.566e+02 3.283e+02 9.180e+02, threshold=5.132e+02, percent-clipped=3.0 2023-05-02 07:24:50,573 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264189.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:24:53,467 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-02 07:25:10,479 INFO [train.py:904] (5/8) Epoch 27, batch 300, loss[loss=0.1542, simple_loss=0.2386, pruned_loss=0.03495, over 16523.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2569, pruned_loss=0.04215, over 2591572.61 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:25:38,561 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3696, 3.3800, 3.9264, 2.1668, 3.2580, 2.3830, 3.7833, 3.6515], device='cuda:5'), covar=tensor([0.0242, 0.1065, 0.0480, 0.2214, 0.0785, 0.1071, 0.0584, 0.1235], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0165, 0.0167, 0.0154, 0.0145, 0.0130, 0.0144, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:25:53,766 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:26:19,432 INFO [train.py:904] (5/8) Epoch 27, batch 350, loss[loss=0.1647, simple_loss=0.2392, pruned_loss=0.04514, over 16815.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2539, pruned_loss=0.04027, over 2764481.60 frames. ], batch size: 102, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:26:34,330 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 2.070e+02 2.475e+02 2.951e+02 5.112e+02, threshold=4.951e+02, percent-clipped=0.0 2023-05-02 07:27:17,871 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:27:27,698 INFO [train.py:904] (5/8) Epoch 27, batch 400, loss[loss=0.1733, simple_loss=0.2494, pruned_loss=0.0486, over 16746.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2535, pruned_loss=0.04043, over 2882813.37 frames. ], batch size: 83, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:27:35,608 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5381, 4.5370, 4.8667, 4.8665, 4.9041, 4.6096, 4.6156, 4.4748], device='cuda:5'), covar=tensor([0.0410, 0.0748, 0.0455, 0.0430, 0.0538, 0.0510, 0.0862, 0.0670], device='cuda:5'), in_proj_covar=tensor([0.0418, 0.0473, 0.0459, 0.0421, 0.0506, 0.0483, 0.0555, 0.0385], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 07:27:41,718 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-05-02 07:27:45,295 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264316.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:28:33,911 INFO [train.py:904] (5/8) Epoch 27, batch 450, loss[loss=0.1613, simple_loss=0.2554, pruned_loss=0.03357, over 17126.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2526, pruned_loss=0.04041, over 2981474.88 frames. ], batch size: 48, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:28:47,823 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 07:28:48,386 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.223e+02 2.597e+02 3.088e+02 6.420e+02, threshold=5.195e+02, percent-clipped=1.0 2023-05-02 07:29:12,317 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7220, 3.9295, 2.1457, 4.5080, 3.1438, 4.3615, 2.2394, 3.1654], device='cuda:5'), covar=tensor([0.0352, 0.0462, 0.2110, 0.0360, 0.0861, 0.0478, 0.2150, 0.0817], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0170, 0.0179, 0.0218, 0.0205, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:29:37,174 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7357, 3.8476, 2.1687, 4.4313, 3.1130, 4.3017, 2.2921, 3.1814], device='cuda:5'), covar=tensor([0.0325, 0.0422, 0.1925, 0.0293, 0.0812, 0.0513, 0.1913, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0179, 0.0196, 0.0170, 0.0179, 0.0218, 0.0205, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:29:44,923 INFO [train.py:904] (5/8) Epoch 27, batch 500, loss[loss=0.1946, simple_loss=0.2802, pruned_loss=0.05451, over 16844.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2516, pruned_loss=0.03959, over 3064736.40 frames. ], batch size: 96, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:30:51,116 INFO [train.py:904] (5/8) Epoch 27, batch 550, loss[loss=0.2003, simple_loss=0.281, pruned_loss=0.05976, over 16752.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2512, pruned_loss=0.03896, over 3123631.12 frames. ], batch size: 124, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:31:04,229 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.059e+02 2.347e+02 2.787e+02 4.742e+02, threshold=4.693e+02, percent-clipped=0.0 2023-05-02 07:31:26,711 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264479.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:31:41,335 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264489.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:31:55,542 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264500.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:31:59,358 INFO [train.py:904] (5/8) Epoch 27, batch 600, loss[loss=0.1553, simple_loss=0.2289, pruned_loss=0.0408, over 16709.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2503, pruned_loss=0.03961, over 3172252.29 frames. ], batch size: 124, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:32:46,463 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264537.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:32:50,461 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:32:59,535 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 07:33:08,807 INFO [train.py:904] (5/8) Epoch 27, batch 650, loss[loss=0.1435, simple_loss=0.2251, pruned_loss=0.03097, over 16777.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2483, pruned_loss=0.03912, over 3197061.29 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:33:18,998 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264561.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 07:33:20,707 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.185e+02 2.561e+02 3.175e+02 5.971e+02, threshold=5.123e+02, percent-clipped=4.0 2023-05-02 07:33:59,038 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:34:03,653 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5752, 6.0291, 5.7593, 5.8165, 5.4157, 5.5333, 5.4013, 6.1544], device='cuda:5'), covar=tensor([0.1657, 0.1032, 0.1108, 0.0965, 0.0994, 0.0648, 0.1326, 0.0940], device='cuda:5'), in_proj_covar=tensor([0.0711, 0.0855, 0.0700, 0.0659, 0.0547, 0.0544, 0.0724, 0.0671], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:34:16,000 INFO [train.py:904] (5/8) Epoch 27, batch 700, loss[loss=0.1469, simple_loss=0.2322, pruned_loss=0.03081, over 16991.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2479, pruned_loss=0.03848, over 3232593.03 frames. ], batch size: 41, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:34:30,518 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9826, 2.8240, 2.6584, 5.0574, 4.0040, 4.3337, 1.8509, 3.0711], device='cuda:5'), covar=tensor([0.1267, 0.0769, 0.1214, 0.0182, 0.0193, 0.0468, 0.1506, 0.0796], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0179, 0.0197, 0.0197, 0.0203, 0.0217, 0.0207, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:34:34,298 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264616.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:35:22,125 INFO [train.py:904] (5/8) Epoch 27, batch 750, loss[loss=0.1515, simple_loss=0.2373, pruned_loss=0.03283, over 16800.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2483, pruned_loss=0.0385, over 3244549.63 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:35:35,425 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.184e+02 2.458e+02 2.836e+02 5.785e+02, threshold=4.916e+02, percent-clipped=2.0 2023-05-02 07:35:37,451 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264664.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:36:02,139 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9664, 2.0972, 2.3372, 3.4314, 2.1709, 2.3231, 2.2628, 2.2350], device='cuda:5'), covar=tensor([0.1701, 0.3937, 0.3119, 0.0831, 0.4259, 0.2898, 0.3871, 0.3693], device='cuda:5'), in_proj_covar=tensor([0.0418, 0.0469, 0.0384, 0.0333, 0.0444, 0.0534, 0.0440, 0.0547], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:36:03,202 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6288, 3.6885, 2.3169, 4.1363, 2.9547, 4.0630, 2.4349, 3.0796], device='cuda:5'), covar=tensor([0.0356, 0.0423, 0.1708, 0.0409, 0.0846, 0.0703, 0.1555, 0.0798], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0173, 0.0181, 0.0220, 0.0206, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:36:28,545 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4650, 3.3977, 4.0275, 2.2691, 3.3001, 2.5097, 3.8722, 3.7522], device='cuda:5'), covar=tensor([0.0234, 0.1026, 0.0468, 0.2065, 0.0769, 0.0988, 0.0573, 0.1032], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0168, 0.0169, 0.0156, 0.0147, 0.0132, 0.0146, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:36:29,180 INFO [train.py:904] (5/8) Epoch 27, batch 800, loss[loss=0.1562, simple_loss=0.2484, pruned_loss=0.03194, over 17039.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.248, pruned_loss=0.0381, over 3262297.76 frames. ], batch size: 55, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:36:53,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9724, 3.6466, 4.0599, 2.2956, 4.2199, 4.2583, 3.2526, 3.3086], device='cuda:5'), covar=tensor([0.0724, 0.0293, 0.0232, 0.1148, 0.0104, 0.0211, 0.0477, 0.0449], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0100, 0.0139, 0.0086, 0.0130, 0.0130, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:37:30,001 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6926, 4.7641, 5.1207, 5.1135, 5.1001, 4.8215, 4.7671, 4.6751], device='cuda:5'), covar=tensor([0.0419, 0.0717, 0.0461, 0.0463, 0.0550, 0.0517, 0.0964, 0.0559], device='cuda:5'), in_proj_covar=tensor([0.0432, 0.0488, 0.0472, 0.0433, 0.0523, 0.0500, 0.0573, 0.0398], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 07:37:36,928 INFO [train.py:904] (5/8) Epoch 27, batch 850, loss[loss=0.1794, simple_loss=0.2549, pruned_loss=0.05198, over 16506.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2479, pruned_loss=0.03832, over 3279011.82 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:37:51,808 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.032e+02 2.277e+02 2.694e+02 3.998e+02, threshold=4.553e+02, percent-clipped=0.0 2023-05-02 07:37:56,327 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1908, 3.0935, 3.0794, 5.2614, 4.3667, 4.5842, 2.0873, 3.4081], device='cuda:5'), covar=tensor([0.1238, 0.0760, 0.1088, 0.0216, 0.0213, 0.0367, 0.1511, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0179, 0.0198, 0.0198, 0.0204, 0.0217, 0.0208, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:38:44,875 INFO [train.py:904] (5/8) Epoch 27, batch 900, loss[loss=0.1498, simple_loss=0.242, pruned_loss=0.02877, over 17166.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2465, pruned_loss=0.03741, over 3295370.39 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:38:47,137 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264804.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:39:20,832 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8701, 4.8017, 4.7763, 4.3929, 4.4598, 4.7780, 4.6383, 4.4854], device='cuda:5'), covar=tensor([0.0618, 0.0711, 0.0324, 0.0362, 0.0963, 0.0499, 0.0441, 0.0720], device='cuda:5'), in_proj_covar=tensor([0.0314, 0.0466, 0.0364, 0.0367, 0.0364, 0.0423, 0.0250, 0.0438], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:39:28,077 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:39:51,350 INFO [train.py:904] (5/8) Epoch 27, batch 950, loss[loss=0.1614, simple_loss=0.2455, pruned_loss=0.03865, over 16424.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2476, pruned_loss=0.03835, over 3304823.36 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:39:55,658 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264856.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:40:04,152 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 2.026e+02 2.339e+02 3.072e+02 8.203e+02, threshold=4.678e+02, percent-clipped=3.0 2023-05-02 07:40:07,792 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264865.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:41,801 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264890.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:57,198 INFO [train.py:904] (5/8) Epoch 27, batch 1000, loss[loss=0.1428, simple_loss=0.2233, pruned_loss=0.03115, over 16714.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.246, pruned_loss=0.03825, over 3305722.77 frames. ], batch size: 76, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:41:46,840 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264938.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:42:07,416 INFO [train.py:904] (5/8) Epoch 27, batch 1050, loss[loss=0.179, simple_loss=0.2512, pruned_loss=0.0534, over 16904.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2462, pruned_loss=0.03811, over 3312839.37 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:42:19,712 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.065e+02 2.389e+02 2.901e+02 6.221e+02, threshold=4.777e+02, percent-clipped=3.0 2023-05-02 07:42:22,443 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6844, 3.7396, 2.8306, 2.2686, 2.3527, 2.3869, 3.8501, 3.2340], device='cuda:5'), covar=tensor([0.2816, 0.0595, 0.1795, 0.3212, 0.3024, 0.2226, 0.0496, 0.1679], device='cuda:5'), in_proj_covar=tensor([0.0332, 0.0273, 0.0311, 0.0324, 0.0301, 0.0274, 0.0303, 0.0349], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 07:43:16,452 INFO [train.py:904] (5/8) Epoch 27, batch 1100, loss[loss=0.1427, simple_loss=0.224, pruned_loss=0.03069, over 15476.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2446, pruned_loss=0.03762, over 3308360.44 frames. ], batch size: 190, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:43:58,013 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 07:44:25,479 INFO [train.py:904] (5/8) Epoch 27, batch 1150, loss[loss=0.1593, simple_loss=0.2432, pruned_loss=0.03776, over 16933.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2436, pruned_loss=0.0368, over 3317936.40 frames. ], batch size: 116, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:44:39,251 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.949e+02 2.374e+02 2.990e+02 6.257e+02, threshold=4.748e+02, percent-clipped=3.0 2023-05-02 07:45:34,346 INFO [train.py:904] (5/8) Epoch 27, batch 1200, loss[loss=0.1718, simple_loss=0.2664, pruned_loss=0.03856, over 16637.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2431, pruned_loss=0.03643, over 3305419.75 frames. ], batch size: 62, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:46:19,432 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265135.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:46:42,962 INFO [train.py:904] (5/8) Epoch 27, batch 1250, loss[loss=0.1588, simple_loss=0.2541, pruned_loss=0.03171, over 17057.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2437, pruned_loss=0.03665, over 3317254.69 frames. ], batch size: 55, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:46:48,423 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265156.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:46:53,538 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:46:57,313 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.048e+02 2.480e+02 3.047e+02 6.939e+02, threshold=4.960e+02, percent-clipped=4.0 2023-05-02 07:47:15,593 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1673, 4.0050, 4.2466, 4.3463, 4.3926, 4.0142, 4.2418, 4.4079], device='cuda:5'), covar=tensor([0.1550, 0.1160, 0.1137, 0.0620, 0.0602, 0.1472, 0.2164, 0.0740], device='cuda:5'), in_proj_covar=tensor([0.0683, 0.0840, 0.0970, 0.0848, 0.0644, 0.0669, 0.0706, 0.0820], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:47:25,275 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265183.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:47:53,013 INFO [train.py:904] (5/8) Epoch 27, batch 1300, loss[loss=0.164, simple_loss=0.2552, pruned_loss=0.03641, over 17056.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2439, pruned_loss=0.03674, over 3319012.38 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:47:54,319 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265204.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:49:00,611 INFO [train.py:904] (5/8) Epoch 27, batch 1350, loss[loss=0.1624, simple_loss=0.2417, pruned_loss=0.04154, over 16429.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2446, pruned_loss=0.03651, over 3320535.71 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:49:14,444 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.174e+02 2.461e+02 3.019e+02 8.065e+02, threshold=4.923e+02, percent-clipped=2.0 2023-05-02 07:49:28,315 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4083, 4.3957, 4.3389, 3.8224, 4.3827, 1.7987, 4.1732, 3.8567], device='cuda:5'), covar=tensor([0.0133, 0.0117, 0.0187, 0.0283, 0.0099, 0.2936, 0.0158, 0.0267], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0173, 0.0209, 0.0184, 0.0187, 0.0218, 0.0200, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:49:35,311 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:50:07,766 INFO [train.py:904] (5/8) Epoch 27, batch 1400, loss[loss=0.1377, simple_loss=0.2307, pruned_loss=0.02231, over 17096.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2444, pruned_loss=0.03634, over 3327303.93 frames. ], batch size: 47, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:50:56,845 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:51:12,357 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4006, 3.3515, 3.4247, 3.4877, 3.5549, 3.3264, 3.5109, 3.6261], device='cuda:5'), covar=tensor([0.1279, 0.1039, 0.1125, 0.0660, 0.0604, 0.2135, 0.1270, 0.0847], device='cuda:5'), in_proj_covar=tensor([0.0688, 0.0847, 0.0978, 0.0856, 0.0649, 0.0674, 0.0711, 0.0827], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:51:15,559 INFO [train.py:904] (5/8) Epoch 27, batch 1450, loss[loss=0.1492, simple_loss=0.247, pruned_loss=0.02567, over 17116.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.244, pruned_loss=0.03619, over 3326856.44 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:51:20,068 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6078, 4.5788, 4.5005, 4.0335, 4.5591, 1.9489, 4.2936, 4.0396], device='cuda:5'), covar=tensor([0.0149, 0.0126, 0.0176, 0.0296, 0.0105, 0.2742, 0.0169, 0.0270], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0173, 0.0209, 0.0184, 0.0187, 0.0218, 0.0200, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:51:29,670 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.111e+02 2.470e+02 3.011e+02 5.399e+02, threshold=4.940e+02, percent-clipped=1.0 2023-05-02 07:52:24,303 INFO [train.py:904] (5/8) Epoch 27, batch 1500, loss[loss=0.1554, simple_loss=0.2584, pruned_loss=0.02618, over 17109.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2444, pruned_loss=0.03644, over 3325205.85 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:52:32,997 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5011, 3.3576, 2.7864, 2.1987, 2.2438, 2.3383, 3.5117, 3.0129], device='cuda:5'), covar=tensor([0.2822, 0.0712, 0.1733, 0.2893, 0.2756, 0.2262, 0.0539, 0.1632], device='cuda:5'), in_proj_covar=tensor([0.0334, 0.0275, 0.0312, 0.0326, 0.0304, 0.0276, 0.0304, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 07:53:10,806 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9149, 4.6985, 4.9451, 5.1531, 5.2958, 4.6971, 5.3230, 5.3140], device='cuda:5'), covar=tensor([0.2154, 0.1668, 0.2239, 0.0992, 0.0763, 0.1155, 0.0682, 0.0834], device='cuda:5'), in_proj_covar=tensor([0.0692, 0.0851, 0.0983, 0.0860, 0.0653, 0.0679, 0.0715, 0.0832], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:53:17,159 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9884, 4.0297, 4.3101, 4.2895, 4.3054, 4.0602, 4.0875, 4.0566], device='cuda:5'), covar=tensor([0.0384, 0.0774, 0.0429, 0.0408, 0.0538, 0.0491, 0.0787, 0.0573], device='cuda:5'), in_proj_covar=tensor([0.0435, 0.0493, 0.0474, 0.0435, 0.0525, 0.0502, 0.0577, 0.0401], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 07:53:28,022 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-02 07:53:36,984 INFO [train.py:904] (5/8) Epoch 27, batch 1550, loss[loss=0.1866, simple_loss=0.2565, pruned_loss=0.05829, over 16786.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2457, pruned_loss=0.03739, over 3324636.86 frames. ], batch size: 124, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:53:46,015 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265460.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:53:49,838 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.243e+02 2.540e+02 3.021e+02 6.885e+02, threshold=5.080e+02, percent-clipped=3.0 2023-05-02 07:54:34,729 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-05-02 07:54:43,799 INFO [train.py:904] (5/8) Epoch 27, batch 1600, loss[loss=0.1438, simple_loss=0.2413, pruned_loss=0.02313, over 17124.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2478, pruned_loss=0.03819, over 3325603.63 frames. ], batch size: 47, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:54:51,468 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:21,348 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:47,292 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7022, 2.6393, 2.7896, 4.5922, 2.6136, 2.9617, 2.6792, 2.7922], device='cuda:5'), covar=tensor([0.1281, 0.3382, 0.2947, 0.0540, 0.3905, 0.2597, 0.3477, 0.3603], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0473, 0.0387, 0.0337, 0.0446, 0.0539, 0.0444, 0.0553], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:55:51,087 INFO [train.py:904] (5/8) Epoch 27, batch 1650, loss[loss=0.1599, simple_loss=0.2707, pruned_loss=0.02461, over 17141.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2496, pruned_loss=0.0392, over 3325091.13 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:55:57,715 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 07:56:04,418 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.682e+02 2.309e+02 2.667e+02 3.273e+02 6.043e+02, threshold=5.334e+02, percent-clipped=4.0 2023-05-02 07:56:42,639 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5380, 5.9341, 5.6982, 5.7946, 5.3843, 5.3674, 5.3403, 6.0901], device='cuda:5'), covar=tensor([0.1540, 0.1123, 0.1066, 0.0964, 0.0969, 0.0751, 0.1468, 0.1003], device='cuda:5'), in_proj_covar=tensor([0.0722, 0.0872, 0.0712, 0.0672, 0.0555, 0.0550, 0.0739, 0.0683], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 07:56:44,710 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:57:00,229 INFO [train.py:904] (5/8) Epoch 27, batch 1700, loss[loss=0.1955, simple_loss=0.2805, pruned_loss=0.05522, over 16619.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2509, pruned_loss=0.03922, over 3329503.46 frames. ], batch size: 62, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:57:15,991 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:57:28,148 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2226, 3.4534, 3.5865, 2.2383, 3.0781, 2.4594, 3.7330, 3.7998], device='cuda:5'), covar=tensor([0.0275, 0.0898, 0.0642, 0.2004, 0.0854, 0.1020, 0.0514, 0.0884], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0169, 0.0169, 0.0156, 0.0147, 0.0132, 0.0145, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 07:57:32,250 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-05-02 07:57:42,662 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265634.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:58:07,540 INFO [train.py:904] (5/8) Epoch 27, batch 1750, loss[loss=0.1669, simple_loss=0.2437, pruned_loss=0.04503, over 16934.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2515, pruned_loss=0.03898, over 3330204.08 frames. ], batch size: 116, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:58:22,127 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.457e+02 2.193e+02 2.494e+02 2.973e+02 6.444e+02, threshold=4.988e+02, percent-clipped=1.0 2023-05-02 07:58:38,494 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265675.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:59:15,655 INFO [train.py:904] (5/8) Epoch 27, batch 1800, loss[loss=0.1806, simple_loss=0.2734, pruned_loss=0.04386, over 16449.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2524, pruned_loss=0.03894, over 3329021.94 frames. ], batch size: 75, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:59:25,519 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9492, 4.9912, 5.3823, 5.3592, 5.3837, 5.0264, 4.9682, 4.8334], device='cuda:5'), covar=tensor([0.0383, 0.0579, 0.0418, 0.0437, 0.0535, 0.0483, 0.1078, 0.0526], device='cuda:5'), in_proj_covar=tensor([0.0435, 0.0493, 0.0475, 0.0437, 0.0527, 0.0504, 0.0578, 0.0401], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 07:59:32,513 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:00:10,204 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4341, 3.3794, 3.4689, 3.5222, 3.5832, 3.3308, 3.5142, 3.6521], device='cuda:5'), covar=tensor([0.1196, 0.0979, 0.1031, 0.0644, 0.0628, 0.2512, 0.1357, 0.0806], device='cuda:5'), in_proj_covar=tensor([0.0690, 0.0849, 0.0981, 0.0861, 0.0652, 0.0679, 0.0714, 0.0831], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:00:23,857 INFO [train.py:904] (5/8) Epoch 27, batch 1850, loss[loss=0.1538, simple_loss=0.2484, pruned_loss=0.02961, over 17184.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2532, pruned_loss=0.03886, over 3324511.70 frames. ], batch size: 46, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:00:37,857 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.122e+02 2.552e+02 2.913e+02 6.429e+02, threshold=5.105e+02, percent-clipped=1.0 2023-05-02 08:00:55,764 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:01:23,073 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0748, 5.6012, 5.7653, 5.3974, 5.5240, 6.1342, 5.5400, 5.2511], device='cuda:5'), covar=tensor([0.0970, 0.2086, 0.2523, 0.2110, 0.2770, 0.1054, 0.1628, 0.2305], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0639, 0.0708, 0.0525, 0.0696, 0.0735, 0.0552, 0.0699], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 08:01:33,295 INFO [train.py:904] (5/8) Epoch 27, batch 1900, loss[loss=0.1471, simple_loss=0.2437, pruned_loss=0.02521, over 17118.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2527, pruned_loss=0.03825, over 3328717.20 frames. ], batch size: 47, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:01:44,945 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-02 08:02:42,101 INFO [train.py:904] (5/8) Epoch 27, batch 1950, loss[loss=0.1557, simple_loss=0.2531, pruned_loss=0.02912, over 17269.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2529, pruned_loss=0.03809, over 3323384.57 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:02:54,909 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.085e+02 2.462e+02 3.209e+02 6.071e+02, threshold=4.924e+02, percent-clipped=6.0 2023-05-02 08:03:08,224 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265872.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:03:23,110 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7344, 2.5565, 2.4601, 4.1204, 3.3500, 4.0909, 1.5943, 2.9147], device='cuda:5'), covar=tensor([0.1502, 0.0806, 0.1267, 0.0183, 0.0139, 0.0360, 0.1660, 0.0874], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0180, 0.0200, 0.0200, 0.0206, 0.0220, 0.0208, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:03:27,059 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:03:50,721 INFO [train.py:904] (5/8) Epoch 27, batch 2000, loss[loss=0.1555, simple_loss=0.2371, pruned_loss=0.03689, over 16901.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2526, pruned_loss=0.03791, over 3317672.66 frames. ], batch size: 96, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:04:10,583 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:33,615 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265933.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:34,799 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:01,223 INFO [train.py:904] (5/8) Epoch 27, batch 2050, loss[loss=0.1485, simple_loss=0.2351, pruned_loss=0.03097, over 17223.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2527, pruned_loss=0.03814, over 3311255.01 frames. ], batch size: 44, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:05:14,205 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.179e+02 2.495e+02 3.022e+02 6.334e+02, threshold=4.990e+02, percent-clipped=2.0 2023-05-02 08:05:25,363 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265970.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:05:36,751 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:41,863 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265982.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:55,149 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2736, 4.2846, 4.4113, 4.1634, 4.2478, 4.8202, 4.3103, 3.9874], device='cuda:5'), covar=tensor([0.1803, 0.2437, 0.2719, 0.2358, 0.2920, 0.1356, 0.2039, 0.2924], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0641, 0.0708, 0.0525, 0.0697, 0.0735, 0.0551, 0.0697], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 08:06:14,069 INFO [train.py:904] (5/8) Epoch 27, batch 2100, loss[loss=0.1824, simple_loss=0.2616, pruned_loss=0.05165, over 16784.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2537, pruned_loss=0.03922, over 3303465.23 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:06:20,220 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 08:07:07,987 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1708, 2.6655, 2.2180, 2.4979, 2.9835, 2.7277, 3.0651, 3.1099], device='cuda:5'), covar=tensor([0.0224, 0.0448, 0.0591, 0.0457, 0.0288, 0.0382, 0.0236, 0.0304], device='cuda:5'), in_proj_covar=tensor([0.0233, 0.0248, 0.0237, 0.0237, 0.0249, 0.0248, 0.0247, 0.0246], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:07:22,091 INFO [train.py:904] (5/8) Epoch 27, batch 2150, loss[loss=0.1542, simple_loss=0.2416, pruned_loss=0.03335, over 16034.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2541, pruned_loss=0.03924, over 3303396.54 frames. ], batch size: 35, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:07:37,033 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.274e+02 2.676e+02 3.152e+02 5.314e+02, threshold=5.352e+02, percent-clipped=2.0 2023-05-02 08:07:43,519 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 08:07:47,124 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:08:32,968 INFO [train.py:904] (5/8) Epoch 27, batch 2200, loss[loss=0.1886, simple_loss=0.2621, pruned_loss=0.05754, over 16907.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2545, pruned_loss=0.03956, over 3304163.57 frames. ], batch size: 109, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:08:51,979 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9994, 3.8536, 4.2637, 2.1427, 4.4713, 4.5891, 3.3373, 3.5557], device='cuda:5'), covar=tensor([0.0789, 0.0288, 0.0254, 0.1318, 0.0111, 0.0236, 0.0454, 0.0432], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0086, 0.0131, 0.0131, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:09:01,234 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0965, 3.0427, 2.0588, 3.2588, 2.4019, 3.2876, 2.1467, 2.6221], device='cuda:5'), covar=tensor([0.0331, 0.0461, 0.1567, 0.0408, 0.0880, 0.0662, 0.1545, 0.0777], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0182, 0.0198, 0.0175, 0.0182, 0.0223, 0.0207, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:09:35,492 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 08:09:41,316 INFO [train.py:904] (5/8) Epoch 27, batch 2250, loss[loss=0.1802, simple_loss=0.2564, pruned_loss=0.05202, over 16706.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2553, pruned_loss=0.04005, over 3303028.97 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:09:56,596 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.173e+02 2.503e+02 2.978e+02 4.988e+02, threshold=5.005e+02, percent-clipped=0.0 2023-05-02 08:10:28,032 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:10:51,415 INFO [train.py:904] (5/8) Epoch 27, batch 2300, loss[loss=0.165, simple_loss=0.252, pruned_loss=0.03898, over 16257.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.254, pruned_loss=0.03909, over 3316128.43 frames. ], batch size: 165, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:10:54,029 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 08:11:06,890 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:23,572 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266226.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:11:25,877 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266228.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:34,365 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:59,651 INFO [train.py:904] (5/8) Epoch 27, batch 2350, loss[loss=0.1852, simple_loss=0.262, pruned_loss=0.05422, over 16706.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2547, pruned_loss=0.03955, over 3314634.42 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:12:01,538 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 08:12:14,159 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.140e+02 2.442e+02 2.999e+02 5.112e+02, threshold=4.884e+02, percent-clipped=1.0 2023-05-02 08:12:22,924 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266270.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 08:12:27,936 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:30,388 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:46,508 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266287.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 08:13:01,901 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 08:13:07,775 INFO [train.py:904] (5/8) Epoch 27, batch 2400, loss[loss=0.1663, simple_loss=0.2514, pruned_loss=0.04058, over 16870.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2561, pruned_loss=0.04006, over 3305962.27 frames. ], batch size: 96, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:13:28,957 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:14:07,970 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266346.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:14:16,222 INFO [train.py:904] (5/8) Epoch 27, batch 2450, loss[loss=0.1691, simple_loss=0.2644, pruned_loss=0.03689, over 16630.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2572, pruned_loss=0.04023, over 3310128.70 frames. ], batch size: 62, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:14:25,894 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 08:14:31,085 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.203e+02 2.440e+02 2.935e+02 4.080e+02, threshold=4.880e+02, percent-clipped=0.0 2023-05-02 08:14:34,590 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0280, 2.6214, 2.1367, 2.4075, 2.9438, 2.7364, 2.9450, 3.0892], device='cuda:5'), covar=tensor([0.0262, 0.0460, 0.0574, 0.0482, 0.0301, 0.0346, 0.0273, 0.0303], device='cuda:5'), in_proj_covar=tensor([0.0234, 0.0249, 0.0238, 0.0238, 0.0250, 0.0250, 0.0248, 0.0247], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:14:42,010 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:15:26,459 INFO [train.py:904] (5/8) Epoch 27, batch 2500, loss[loss=0.2035, simple_loss=0.2794, pruned_loss=0.06374, over 16321.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2565, pruned_loss=0.03979, over 3311824.71 frames. ], batch size: 165, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:15:31,579 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:15:48,874 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:16:28,997 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0384, 4.5123, 4.5133, 3.3311, 3.7567, 4.5139, 3.9998, 2.6766], device='cuda:5'), covar=tensor([0.0481, 0.0081, 0.0049, 0.0353, 0.0143, 0.0090, 0.0092, 0.0473], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0090, 0.0091, 0.0137, 0.0103, 0.0116, 0.0099, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 08:16:35,711 INFO [train.py:904] (5/8) Epoch 27, batch 2550, loss[loss=0.146, simple_loss=0.2407, pruned_loss=0.02564, over 17236.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2565, pruned_loss=0.03971, over 3313549.12 frames. ], batch size: 44, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:16:51,125 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.440e+02 2.045e+02 2.342e+02 2.686e+02 5.745e+02, threshold=4.685e+02, percent-clipped=2.0 2023-05-02 08:16:51,571 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266464.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:17:13,736 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1114, 4.8541, 5.1160, 5.2876, 5.5348, 4.8719, 5.4934, 5.5414], device='cuda:5'), covar=tensor([0.1943, 0.1401, 0.1912, 0.0881, 0.0559, 0.0854, 0.0574, 0.0698], device='cuda:5'), in_proj_covar=tensor([0.0691, 0.0847, 0.0983, 0.0859, 0.0652, 0.0680, 0.0714, 0.0829], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:17:33,923 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8442, 2.5251, 2.4686, 3.9431, 3.1822, 3.9726, 1.6177, 2.9291], device='cuda:5'), covar=tensor([0.1330, 0.0763, 0.1157, 0.0182, 0.0143, 0.0367, 0.1513, 0.0789], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0201, 0.0206, 0.0220, 0.0208, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:17:45,262 INFO [train.py:904] (5/8) Epoch 27, batch 2600, loss[loss=0.177, simple_loss=0.2605, pruned_loss=0.04677, over 16746.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2565, pruned_loss=0.03965, over 3304083.86 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:18:17,146 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266525.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:21,130 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:55,444 INFO [train.py:904] (5/8) Epoch 27, batch 2650, loss[loss=0.1658, simple_loss=0.2462, pruned_loss=0.0427, over 16383.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2563, pruned_loss=0.03914, over 3310628.95 frames. ], batch size: 145, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:19:07,724 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8780, 3.0137, 2.6816, 5.0116, 3.9516, 4.2354, 1.8410, 3.1275], device='cuda:5'), covar=tensor([0.1343, 0.0762, 0.1229, 0.0260, 0.0230, 0.0534, 0.1553, 0.0842], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0200, 0.0205, 0.0219, 0.0208, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:19:11,569 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.008e+02 2.392e+02 2.897e+02 5.210e+02, threshold=4.785e+02, percent-clipped=1.0 2023-05-02 08:19:21,014 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:24,648 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:29,431 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:37,437 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266582.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:20:05,800 INFO [train.py:904] (5/8) Epoch 27, batch 2700, loss[loss=0.1661, simple_loss=0.2535, pruned_loss=0.03931, over 16525.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2563, pruned_loss=0.03835, over 3317084.12 frames. ], batch size: 75, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:20:31,907 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:20:42,935 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6743, 2.6195, 1.9543, 2.7535, 2.1821, 2.8509, 2.1701, 2.4222], device='cuda:5'), covar=tensor([0.0358, 0.0386, 0.1335, 0.0301, 0.0692, 0.0511, 0.1233, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0183, 0.0199, 0.0176, 0.0183, 0.0224, 0.0207, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:21:15,412 INFO [train.py:904] (5/8) Epoch 27, batch 2750, loss[loss=0.1794, simple_loss=0.2616, pruned_loss=0.04857, over 16563.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2574, pruned_loss=0.03885, over 3318287.83 frames. ], batch size: 75, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:21:16,992 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7733, 3.9639, 2.6318, 4.5833, 3.1620, 4.4871, 2.6653, 3.3083], device='cuda:5'), covar=tensor([0.0351, 0.0432, 0.1584, 0.0327, 0.0829, 0.0666, 0.1537, 0.0754], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0184, 0.0199, 0.0177, 0.0184, 0.0225, 0.0208, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:21:29,200 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.109e+02 2.416e+02 2.772e+02 5.063e+02, threshold=4.832e+02, percent-clipped=2.0 2023-05-02 08:21:51,228 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:22,677 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:23,528 INFO [train.py:904] (5/8) Epoch 27, batch 2800, loss[loss=0.1648, simple_loss=0.2637, pruned_loss=0.03298, over 17133.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2569, pruned_loss=0.03873, over 3321662.45 frames. ], batch size: 48, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:22:25,198 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266704.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:52,011 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-02 08:23:15,485 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266740.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:23:31,358 INFO [train.py:904] (5/8) Epoch 27, batch 2850, loss[loss=0.1409, simple_loss=0.2343, pruned_loss=0.02377, over 16840.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2558, pruned_loss=0.03845, over 3317902.09 frames. ], batch size: 42, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:23:48,197 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.101e+02 2.361e+02 2.749e+02 4.923e+02, threshold=4.721e+02, percent-clipped=1.0 2023-05-02 08:23:49,925 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:24:41,247 INFO [train.py:904] (5/8) Epoch 27, batch 2900, loss[loss=0.1974, simple_loss=0.283, pruned_loss=0.05596, over 17066.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2546, pruned_loss=0.03857, over 3308999.88 frames. ], batch size: 55, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:25:04,255 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266820.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:25:49,407 INFO [train.py:904] (5/8) Epoch 27, batch 2950, loss[loss=0.1711, simple_loss=0.2698, pruned_loss=0.03621, over 17032.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2537, pruned_loss=0.03875, over 3319004.76 frames. ], batch size: 50, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:02,642 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7928, 3.4926, 3.9147, 2.1186, 3.9915, 4.0393, 3.2582, 3.1131], device='cuda:5'), covar=tensor([0.0756, 0.0261, 0.0210, 0.1188, 0.0107, 0.0220, 0.0397, 0.0421], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0086, 0.0132, 0.0131, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:26:04,413 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.163e+02 2.599e+02 3.216e+02 6.007e+02, threshold=5.198e+02, percent-clipped=1.0 2023-05-02 08:26:14,752 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:26:16,131 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3296, 3.6864, 3.8744, 2.2378, 3.0569, 2.4970, 3.7930, 3.8923], device='cuda:5'), covar=tensor([0.0328, 0.0983, 0.0564, 0.2159, 0.0957, 0.1060, 0.0641, 0.1053], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0171, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 08:26:29,723 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266882.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:26:58,266 INFO [train.py:904] (5/8) Epoch 27, batch 3000, loss[loss=0.1683, simple_loss=0.262, pruned_loss=0.03728, over 17071.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2543, pruned_loss=0.03935, over 3320586.19 frames. ], batch size: 55, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:58,266 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 08:27:07,046 INFO [train.py:938] (5/8) Epoch 27, validation: loss=0.1336, simple_loss=0.2386, pruned_loss=0.01429, over 944034.00 frames. 2023-05-02 08:27:07,046 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 08:27:27,844 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:30,436 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:45,321 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266930.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:28:16,693 INFO [train.py:904] (5/8) Epoch 27, batch 3050, loss[loss=0.1568, simple_loss=0.2484, pruned_loss=0.03258, over 16843.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2535, pruned_loss=0.03893, over 3324995.43 frames. ], batch size: 42, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:28:30,554 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.121e+02 2.455e+02 2.792e+02 5.832e+02, threshold=4.910e+02, percent-clipped=2.0 2023-05-02 08:28:49,112 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2224, 5.8949, 6.0880, 5.6807, 5.9112, 6.3535, 5.9149, 5.5228], device='cuda:5'), covar=tensor([0.0949, 0.1708, 0.2155, 0.2029, 0.2390, 0.0975, 0.1560, 0.2436], device='cuda:5'), in_proj_covar=tensor([0.0435, 0.0643, 0.0712, 0.0527, 0.0703, 0.0737, 0.0553, 0.0703], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 08:28:55,885 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266981.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:29:24,845 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267002.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:29:25,705 INFO [train.py:904] (5/8) Epoch 27, batch 3100, loss[loss=0.165, simple_loss=0.2394, pruned_loss=0.04527, over 16290.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2533, pruned_loss=0.039, over 3325312.04 frames. ], batch size: 145, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:29:54,264 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2910, 2.3525, 2.2642, 4.1689, 2.2800, 2.6866, 2.3804, 2.4443], device='cuda:5'), covar=tensor([0.1440, 0.3811, 0.3424, 0.0581, 0.4306, 0.2772, 0.3812, 0.3806], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0475, 0.0388, 0.0340, 0.0447, 0.0543, 0.0446, 0.0555], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:30:11,878 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:33,230 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267050.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:36,926 INFO [train.py:904] (5/8) Epoch 27, batch 3150, loss[loss=0.1784, simple_loss=0.2585, pruned_loss=0.04912, over 11943.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2528, pruned_loss=0.0392, over 3317010.61 frames. ], batch size: 246, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:30:45,813 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267060.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:49,205 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 08:30:50,834 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.137e+02 2.489e+02 2.954e+02 4.893e+02, threshold=4.979e+02, percent-clipped=1.0 2023-05-02 08:31:15,489 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:29,854 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2750, 3.3334, 3.4830, 2.4242, 3.1953, 3.5761, 3.2686, 2.1433], device='cuda:5'), covar=tensor([0.0536, 0.0135, 0.0079, 0.0419, 0.0152, 0.0124, 0.0127, 0.0486], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0090, 0.0092, 0.0137, 0.0103, 0.0117, 0.0100, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 08:31:37,665 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267098.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:43,526 INFO [train.py:904] (5/8) Epoch 27, batch 3200, loss[loss=0.1723, simple_loss=0.2511, pruned_loss=0.0467, over 16906.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2518, pruned_loss=0.03881, over 3320149.22 frames. ], batch size: 90, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:32:08,035 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267120.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:39,048 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:50,809 INFO [train.py:904] (5/8) Epoch 27, batch 3250, loss[loss=0.1344, simple_loss=0.2224, pruned_loss=0.02321, over 17011.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2515, pruned_loss=0.03854, over 3328833.22 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:33:01,476 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:07,332 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 2.100e+02 2.558e+02 2.992e+02 5.304e+02, threshold=5.115e+02, percent-clipped=1.0 2023-05-02 08:33:12,827 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267168.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:29,393 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-05-02 08:33:52,574 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7361, 4.7066, 4.6636, 4.3002, 4.3636, 4.6970, 4.5245, 4.4684], device='cuda:5'), covar=tensor([0.0655, 0.0766, 0.0322, 0.0342, 0.0861, 0.0535, 0.0437, 0.0710], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0483, 0.0376, 0.0381, 0.0376, 0.0438, 0.0258, 0.0453], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 08:34:00,064 INFO [train.py:904] (5/8) Epoch 27, batch 3300, loss[loss=0.1675, simple_loss=0.2619, pruned_loss=0.03662, over 16697.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2527, pruned_loss=0.0384, over 3329108.47 frames. ], batch size: 62, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:34:37,850 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7300, 3.9048, 2.6295, 4.5028, 3.0669, 4.4010, 2.6298, 3.2493], device='cuda:5'), covar=tensor([0.0362, 0.0420, 0.1492, 0.0292, 0.0864, 0.0526, 0.1527, 0.0798], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0183, 0.0198, 0.0177, 0.0182, 0.0224, 0.0206, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:35:07,432 INFO [train.py:904] (5/8) Epoch 27, batch 3350, loss[loss=0.1702, simple_loss=0.2534, pruned_loss=0.04347, over 16800.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2537, pruned_loss=0.03842, over 3332779.87 frames. ], batch size: 102, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:35:22,680 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.912e+02 2.198e+02 2.588e+02 3.700e+02, threshold=4.395e+02, percent-clipped=0.0 2023-05-02 08:35:38,228 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:36:14,896 INFO [train.py:904] (5/8) Epoch 27, batch 3400, loss[loss=0.1653, simple_loss=0.2586, pruned_loss=0.03599, over 17129.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.254, pruned_loss=0.03886, over 3331684.33 frames. ], batch size: 48, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:36:23,540 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7805, 4.8554, 5.1922, 5.1873, 5.2078, 4.8903, 4.8375, 4.7255], device='cuda:5'), covar=tensor([0.0382, 0.0516, 0.0431, 0.0443, 0.0472, 0.0400, 0.0910, 0.0512], device='cuda:5'), in_proj_covar=tensor([0.0445, 0.0500, 0.0482, 0.0445, 0.0531, 0.0511, 0.0588, 0.0407], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 08:36:49,913 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 08:36:58,135 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267335.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:22,636 INFO [train.py:904] (5/8) Epoch 27, batch 3450, loss[loss=0.1432, simple_loss=0.2301, pruned_loss=0.02818, over 16844.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2526, pruned_loss=0.03805, over 3335695.88 frames. ], batch size: 42, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:37:31,874 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267360.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:36,677 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 2.041e+02 2.356e+02 2.660e+02 5.077e+02, threshold=4.713e+02, percent-clipped=1.0 2023-05-02 08:38:03,485 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:38:30,924 INFO [train.py:904] (5/8) Epoch 27, batch 3500, loss[loss=0.1766, simple_loss=0.2525, pruned_loss=0.05035, over 16756.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2515, pruned_loss=0.03752, over 3335402.23 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:38:39,648 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267408.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:07,371 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 08:39:11,760 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0107, 3.1551, 3.3130, 2.2231, 2.8890, 2.2832, 3.5338, 3.5219], device='cuda:5'), covar=tensor([0.0262, 0.1037, 0.0616, 0.1962, 0.0903, 0.1084, 0.0524, 0.0887], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0174, 0.0171, 0.0158, 0.0149, 0.0133, 0.0148, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 08:39:18,473 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 08:39:19,827 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:40,534 INFO [train.py:904] (5/8) Epoch 27, batch 3550, loss[loss=0.1638, simple_loss=0.2591, pruned_loss=0.03421, over 16735.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2506, pruned_loss=0.03721, over 3330770.36 frames. ], batch size: 57, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:39:41,967 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:53,963 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.045e+02 2.366e+02 2.928e+02 5.405e+02, threshold=4.732e+02, percent-clipped=1.0 2023-05-02 08:40:08,300 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-02 08:40:36,676 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:40:41,010 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 08:40:48,176 INFO [train.py:904] (5/8) Epoch 27, batch 3600, loss[loss=0.1715, simple_loss=0.2458, pruned_loss=0.04861, over 16747.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2498, pruned_loss=0.03692, over 3332062.61 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:41:11,805 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7235, 3.9964, 2.7097, 4.5094, 3.0302, 4.4347, 2.6886, 3.2264], device='cuda:5'), covar=tensor([0.0343, 0.0353, 0.1431, 0.0367, 0.0878, 0.0541, 0.1465, 0.0779], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0185, 0.0200, 0.0178, 0.0183, 0.0226, 0.0207, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:41:54,012 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7038, 2.5550, 2.5746, 4.5112, 2.5508, 2.9691, 2.6204, 2.7311], device='cuda:5'), covar=tensor([0.1323, 0.3760, 0.3212, 0.0552, 0.4167, 0.2656, 0.3744, 0.3490], device='cuda:5'), in_proj_covar=tensor([0.0425, 0.0477, 0.0389, 0.0341, 0.0449, 0.0545, 0.0448, 0.0557], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:42:00,666 INFO [train.py:904] (5/8) Epoch 27, batch 3650, loss[loss=0.1865, simple_loss=0.2578, pruned_loss=0.05757, over 16711.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2488, pruned_loss=0.03739, over 3315450.04 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:42:03,563 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:42:16,690 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 2.009e+02 2.392e+02 2.845e+02 4.698e+02, threshold=4.785e+02, percent-clipped=1.0 2023-05-02 08:42:36,614 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:43:14,274 INFO [train.py:904] (5/8) Epoch 27, batch 3700, loss[loss=0.1591, simple_loss=0.2356, pruned_loss=0.04135, over 16361.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2477, pruned_loss=0.03896, over 3299844.29 frames. ], batch size: 68, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:43:45,318 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267624.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:43:47,606 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 08:44:19,783 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8068, 2.7937, 2.6401, 4.6067, 3.6002, 4.0822, 1.8691, 3.0724], device='cuda:5'), covar=tensor([0.1304, 0.0816, 0.1246, 0.0173, 0.0345, 0.0408, 0.1503, 0.0846], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0202, 0.0208, 0.0220, 0.0209, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:44:27,026 INFO [train.py:904] (5/8) Epoch 27, batch 3750, loss[loss=0.17, simple_loss=0.2429, pruned_loss=0.04857, over 16729.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2482, pruned_loss=0.04022, over 3261376.85 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:44:42,770 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.122e+02 2.445e+02 2.969e+02 5.138e+02, threshold=4.890e+02, percent-clipped=1.0 2023-05-02 08:45:40,198 INFO [train.py:904] (5/8) Epoch 27, batch 3800, loss[loss=0.1506, simple_loss=0.2264, pruned_loss=0.03739, over 16800.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.249, pruned_loss=0.0416, over 3256669.01 frames. ], batch size: 102, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:45:41,591 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 08:46:21,684 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7952, 2.7556, 2.7268, 4.8110, 3.7479, 4.2437, 1.6627, 3.0875], device='cuda:5'), covar=tensor([0.1362, 0.0847, 0.1252, 0.0157, 0.0347, 0.0367, 0.1669, 0.0899], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0180, 0.0198, 0.0202, 0.0207, 0.0219, 0.0208, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:46:31,910 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:46:48,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5953, 3.3510, 3.6894, 2.0663, 3.7499, 3.7943, 3.2120, 2.9482], device='cuda:5'), covar=tensor([0.0778, 0.0266, 0.0217, 0.1160, 0.0145, 0.0256, 0.0400, 0.0432], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0112, 0.0102, 0.0139, 0.0087, 0.0132, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:46:52,907 INFO [train.py:904] (5/8) Epoch 27, batch 3850, loss[loss=0.1748, simple_loss=0.2508, pruned_loss=0.04946, over 16413.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2492, pruned_loss=0.0425, over 3257374.11 frames. ], batch size: 68, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:46:54,317 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:47:08,879 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.090e+02 2.451e+02 2.947e+02 6.738e+02, threshold=4.901e+02, percent-clipped=4.0 2023-05-02 08:47:13,738 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 08:47:39,701 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:47:44,076 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5026, 4.5750, 4.7231, 4.5133, 4.6023, 5.1585, 4.6960, 4.4005], device='cuda:5'), covar=tensor([0.1634, 0.2142, 0.2255, 0.2274, 0.2474, 0.1094, 0.1618, 0.2569], device='cuda:5'), in_proj_covar=tensor([0.0437, 0.0646, 0.0710, 0.0527, 0.0703, 0.0737, 0.0552, 0.0701], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 08:48:02,763 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267802.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:48:03,556 INFO [train.py:904] (5/8) Epoch 27, batch 3900, loss[loss=0.1678, simple_loss=0.248, pruned_loss=0.04383, over 16296.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2492, pruned_loss=0.04323, over 3260102.94 frames. ], batch size: 165, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:48:10,281 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8300, 3.8077, 4.0403, 2.4663, 3.5048, 2.8844, 4.2758, 4.2444], device='cuda:5'), covar=tensor([0.0174, 0.0773, 0.0521, 0.1959, 0.0758, 0.0853, 0.0405, 0.0755], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 08:49:11,715 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267850.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:49:14,869 INFO [train.py:904] (5/8) Epoch 27, batch 3950, loss[loss=0.1879, simple_loss=0.2635, pruned_loss=0.05613, over 16247.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2492, pruned_loss=0.04383, over 3257509.57 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:49:23,524 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1925, 3.9842, 4.0112, 4.3850, 4.4655, 4.1556, 4.2967, 4.4440], device='cuda:5'), covar=tensor([0.1768, 0.1585, 0.2100, 0.0939, 0.0963, 0.1609, 0.2440, 0.1333], device='cuda:5'), in_proj_covar=tensor([0.0699, 0.0860, 0.0999, 0.0870, 0.0662, 0.0691, 0.0722, 0.0841], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:49:25,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7898, 4.6596, 4.7913, 4.9771, 5.0974, 4.6060, 5.0386, 5.1267], device='cuda:5'), covar=tensor([0.1764, 0.1247, 0.1535, 0.0693, 0.0629, 0.1071, 0.1351, 0.0812], device='cuda:5'), in_proj_covar=tensor([0.0698, 0.0860, 0.0999, 0.0870, 0.0662, 0.0691, 0.0722, 0.0841], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:49:32,329 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.233e+02 2.530e+02 3.152e+02 5.510e+02, threshold=5.059e+02, percent-clipped=1.0 2023-05-02 08:49:55,647 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6656, 3.7134, 2.3741, 3.9352, 2.9743, 3.9067, 2.4989, 3.0780], device='cuda:5'), covar=tensor([0.0252, 0.0430, 0.1578, 0.0326, 0.0800, 0.0742, 0.1450, 0.0676], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0183, 0.0198, 0.0176, 0.0181, 0.0223, 0.0205, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:50:28,772 INFO [train.py:904] (5/8) Epoch 27, batch 4000, loss[loss=0.1618, simple_loss=0.2412, pruned_loss=0.04123, over 16812.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2494, pruned_loss=0.04412, over 3255048.53 frames. ], batch size: 102, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:50:49,000 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 08:51:05,535 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9909, 3.0673, 3.3855, 2.1717, 2.9343, 2.3359, 3.5329, 3.4151], device='cuda:5'), covar=tensor([0.0254, 0.0902, 0.0597, 0.2079, 0.0884, 0.0974, 0.0521, 0.0882], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0157, 0.0149, 0.0133, 0.0148, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 08:51:23,409 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-02 08:51:42,555 INFO [train.py:904] (5/8) Epoch 27, batch 4050, loss[loss=0.1611, simple_loss=0.2566, pruned_loss=0.03279, over 15581.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2499, pruned_loss=0.04338, over 3265236.21 frames. ], batch size: 191, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:51:58,303 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 1.841e+02 2.143e+02 2.539e+02 4.386e+02, threshold=4.286e+02, percent-clipped=0.0 2023-05-02 08:52:31,224 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 08:52:42,147 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6999, 2.6473, 2.0027, 2.7434, 2.1551, 2.8491, 2.1816, 2.4237], device='cuda:5'), covar=tensor([0.0306, 0.0335, 0.1256, 0.0201, 0.0626, 0.0352, 0.1246, 0.0607], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0182, 0.0197, 0.0176, 0.0181, 0.0223, 0.0205, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 08:52:59,894 INFO [train.py:904] (5/8) Epoch 27, batch 4100, loss[loss=0.181, simple_loss=0.2694, pruned_loss=0.04632, over 16902.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2518, pruned_loss=0.04303, over 3244227.64 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:54:16,169 INFO [train.py:904] (5/8) Epoch 27, batch 4150, loss[loss=0.2107, simple_loss=0.3024, pruned_loss=0.05953, over 16323.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2584, pruned_loss=0.04492, over 3226722.10 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:54:33,181 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 2.031e+02 2.466e+02 3.151e+02 6.245e+02, threshold=4.932e+02, percent-clipped=8.0 2023-05-02 08:54:40,519 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-02 08:55:32,120 INFO [train.py:904] (5/8) Epoch 27, batch 4200, loss[loss=0.2023, simple_loss=0.2964, pruned_loss=0.0541, over 16703.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2654, pruned_loss=0.04685, over 3201466.08 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:56:22,450 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 08:56:40,949 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268150.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:56:44,663 INFO [train.py:904] (5/8) Epoch 27, batch 4250, loss[loss=0.1569, simple_loss=0.2563, pruned_loss=0.02877, over 16265.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.269, pruned_loss=0.04624, over 3208354.65 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:57:00,876 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.686e+02 2.148e+02 2.458e+02 3.113e+02 4.874e+02, threshold=4.916e+02, percent-clipped=0.0 2023-05-02 08:57:15,337 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6083, 3.6884, 3.4309, 3.0614, 3.2575, 3.5733, 3.3253, 3.4038], device='cuda:5'), covar=tensor([0.0591, 0.0785, 0.0325, 0.0309, 0.0554, 0.0526, 0.1425, 0.0490], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0477, 0.0370, 0.0376, 0.0371, 0.0431, 0.0253, 0.0445], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 08:57:49,657 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 08:57:50,363 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:57:50,452 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4812, 5.7521, 5.5417, 5.6218, 5.3071, 5.0859, 5.1833, 5.9132], device='cuda:5'), covar=tensor([0.1280, 0.0837, 0.1035, 0.0875, 0.0790, 0.0701, 0.1230, 0.0777], device='cuda:5'), in_proj_covar=tensor([0.0719, 0.0873, 0.0708, 0.0670, 0.0553, 0.0549, 0.0734, 0.0686], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 08:57:57,384 INFO [train.py:904] (5/8) Epoch 27, batch 4300, loss[loss=0.198, simple_loss=0.2887, pruned_loss=0.05366, over 16720.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2702, pruned_loss=0.04552, over 3212323.50 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:58:10,929 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268212.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:58:11,000 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3178, 3.4330, 3.5906, 2.1931, 3.0025, 2.3991, 3.6273, 3.6143], device='cuda:5'), covar=tensor([0.0265, 0.0839, 0.0627, 0.2141, 0.0935, 0.0962, 0.0631, 0.1208], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0171, 0.0169, 0.0155, 0.0147, 0.0131, 0.0146, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 08:59:12,227 INFO [train.py:904] (5/8) Epoch 27, batch 4350, loss[loss=0.2276, simple_loss=0.3043, pruned_loss=0.0755, over 11676.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2735, pruned_loss=0.04687, over 3191147.15 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:59:27,931 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.120e+02 2.432e+02 2.957e+02 3.904e+02, threshold=4.863e+02, percent-clipped=0.0 2023-05-02 08:59:43,163 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268273.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 08:59:45,606 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:59:47,209 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 09:00:24,028 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-02 09:00:27,003 INFO [train.py:904] (5/8) Epoch 27, batch 4400, loss[loss=0.1825, simple_loss=0.2681, pruned_loss=0.04848, over 16630.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2758, pruned_loss=0.04816, over 3183199.45 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:14,946 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268336.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:01:37,973 INFO [train.py:904] (5/8) Epoch 27, batch 4450, loss[loss=0.2093, simple_loss=0.3053, pruned_loss=0.05663, over 16822.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.279, pruned_loss=0.04935, over 3187932.36 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:55,124 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 1.898e+02 2.329e+02 2.751e+02 5.089e+02, threshold=4.658e+02, percent-clipped=1.0 2023-05-02 09:02:50,820 INFO [train.py:904] (5/8) Epoch 27, batch 4500, loss[loss=0.1966, simple_loss=0.2872, pruned_loss=0.053, over 16865.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2791, pruned_loss=0.05001, over 3198608.59 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:02:53,036 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6484, 3.6447, 2.3680, 4.3454, 2.8909, 4.2357, 2.5815, 3.0665], device='cuda:5'), covar=tensor([0.0303, 0.0388, 0.1692, 0.0150, 0.0828, 0.0431, 0.1432, 0.0793], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0181, 0.0197, 0.0173, 0.0180, 0.0222, 0.0204, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 09:04:03,668 INFO [train.py:904] (5/8) Epoch 27, batch 4550, loss[loss=0.199, simple_loss=0.2888, pruned_loss=0.0546, over 16743.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.28, pruned_loss=0.05104, over 3210014.52 frames. ], batch size: 124, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:04:20,731 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.908e+02 2.230e+02 2.489e+02 1.268e+03, threshold=4.461e+02, percent-clipped=3.0 2023-05-02 09:04:38,775 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3471, 5.6029, 5.3933, 5.4399, 5.1163, 4.9601, 5.0274, 5.7315], device='cuda:5'), covar=tensor([0.1422, 0.0793, 0.0975, 0.0827, 0.0858, 0.0867, 0.1211, 0.0786], device='cuda:5'), in_proj_covar=tensor([0.0714, 0.0865, 0.0702, 0.0665, 0.0550, 0.0544, 0.0727, 0.0680], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:04:55,626 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4622, 2.7946, 2.6101, 4.1594, 2.6163, 3.0254, 2.6613, 2.7954], device='cuda:5'), covar=tensor([0.1280, 0.2800, 0.2675, 0.0542, 0.3669, 0.2042, 0.3070, 0.2957], device='cuda:5'), in_proj_covar=tensor([0.0420, 0.0473, 0.0384, 0.0336, 0.0446, 0.0541, 0.0443, 0.0553], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:05:15,506 INFO [train.py:904] (5/8) Epoch 27, batch 4600, loss[loss=0.191, simple_loss=0.2856, pruned_loss=0.0482, over 17166.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2818, pruned_loss=0.05188, over 3203481.01 frames. ], batch size: 46, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:06:23,488 INFO [train.py:904] (5/8) Epoch 27, batch 4650, loss[loss=0.1648, simple_loss=0.2592, pruned_loss=0.03524, over 16889.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2811, pruned_loss=0.05189, over 3189587.51 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:06:40,835 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 1.843e+02 2.064e+02 2.500e+02 6.008e+02, threshold=4.128e+02, percent-clipped=2.0 2023-05-02 09:06:45,263 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268568.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:07:32,896 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268601.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:07:34,896 INFO [train.py:904] (5/8) Epoch 27, batch 4700, loss[loss=0.1758, simple_loss=0.2679, pruned_loss=0.04187, over 16541.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2786, pruned_loss=0.05087, over 3204773.23 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:07:43,072 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0392, 2.0661, 2.6123, 3.1100, 2.9362, 3.5310, 2.1725, 3.4396], device='cuda:5'), covar=tensor([0.0269, 0.0617, 0.0381, 0.0341, 0.0348, 0.0189, 0.0648, 0.0164], device='cuda:5'), in_proj_covar=tensor([0.0199, 0.0199, 0.0187, 0.0194, 0.0209, 0.0167, 0.0205, 0.0166], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-05-02 09:08:06,436 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 09:08:15,780 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268631.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:08:45,893 INFO [train.py:904] (5/8) Epoch 27, batch 4750, loss[loss=0.143, simple_loss=0.2378, pruned_loss=0.02408, over 16835.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2741, pruned_loss=0.04865, over 3211996.14 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:09:00,226 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268662.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:09:04,566 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 1.855e+02 2.157e+02 2.478e+02 7.312e+02, threshold=4.313e+02, percent-clipped=3.0 2023-05-02 09:09:59,303 INFO [train.py:904] (5/8) Epoch 27, batch 4800, loss[loss=0.1803, simple_loss=0.2685, pruned_loss=0.04606, over 11906.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2709, pruned_loss=0.04662, over 3198061.34 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:10:29,595 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-02 09:11:14,175 INFO [train.py:904] (5/8) Epoch 27, batch 4850, loss[loss=0.1679, simple_loss=0.271, pruned_loss=0.03237, over 16312.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2708, pruned_loss=0.04548, over 3197849.30 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:11:31,500 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.812e+02 2.075e+02 2.444e+02 6.308e+02, threshold=4.151e+02, percent-clipped=1.0 2023-05-02 09:12:27,710 INFO [train.py:904] (5/8) Epoch 27, batch 4900, loss[loss=0.1692, simple_loss=0.2681, pruned_loss=0.03516, over 15287.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2703, pruned_loss=0.04456, over 3174073.22 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:20,094 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-02 09:13:37,969 INFO [train.py:904] (5/8) Epoch 27, batch 4950, loss[loss=0.1704, simple_loss=0.2713, pruned_loss=0.03473, over 16743.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2696, pruned_loss=0.04378, over 3184009.85 frames. ], batch size: 89, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:54,421 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 1.970e+02 2.304e+02 2.682e+02 4.795e+02, threshold=4.609e+02, percent-clipped=2.0 2023-05-02 09:13:59,143 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268868.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:14:36,329 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:14:50,671 INFO [train.py:904] (5/8) Epoch 27, batch 5000, loss[loss=0.1693, simple_loss=0.2763, pruned_loss=0.0311, over 16856.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2716, pruned_loss=0.04404, over 3184955.78 frames. ], batch size: 102, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:15:09,315 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268916.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:15:29,369 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:00,163 INFO [train.py:904] (5/8) Epoch 27, batch 5050, loss[loss=0.1794, simple_loss=0.2701, pruned_loss=0.04431, over 16673.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.272, pruned_loss=0.04409, over 3186368.00 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:16:03,765 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268955.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:05,785 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:17,236 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 1.970e+02 2.360e+02 2.844e+02 4.584e+02, threshold=4.719e+02, percent-clipped=0.0 2023-05-02 09:16:37,087 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:54,237 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268991.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:17:11,195 INFO [train.py:904] (5/8) Epoch 27, batch 5100, loss[loss=0.1817, simple_loss=0.2674, pruned_loss=0.04802, over 12122.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2702, pruned_loss=0.04332, over 3195409.29 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:17:26,914 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7961, 4.0745, 3.0918, 2.4752, 2.6949, 2.6875, 4.3785, 3.4717], device='cuda:5'), covar=tensor([0.2751, 0.0561, 0.1755, 0.2563, 0.2534, 0.1914, 0.0388, 0.1245], device='cuda:5'), in_proj_covar=tensor([0.0334, 0.0276, 0.0313, 0.0326, 0.0306, 0.0275, 0.0304, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 09:18:22,939 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:18:23,599 INFO [train.py:904] (5/8) Epoch 27, batch 5150, loss[loss=0.174, simple_loss=0.2734, pruned_loss=0.03729, over 16863.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2705, pruned_loss=0.04288, over 3190065.84 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:18:41,483 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 1.963e+02 2.152e+02 2.508e+02 3.926e+02, threshold=4.304e+02, percent-clipped=0.0 2023-05-02 09:19:07,807 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0858, 2.4859, 2.6361, 1.9146, 2.7666, 2.8288, 2.5389, 2.4048], device='cuda:5'), covar=tensor([0.0720, 0.0255, 0.0205, 0.1024, 0.0130, 0.0206, 0.0429, 0.0449], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0101, 0.0141, 0.0087, 0.0131, 0.0131, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 09:19:36,081 INFO [train.py:904] (5/8) Epoch 27, batch 5200, loss[loss=0.1667, simple_loss=0.2608, pruned_loss=0.03628, over 12318.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2693, pruned_loss=0.04257, over 3177025.40 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:19:40,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4176, 4.6882, 4.4714, 4.5454, 4.2965, 4.2336, 4.1781, 4.7218], device='cuda:5'), covar=tensor([0.1352, 0.0907, 0.0987, 0.0808, 0.0836, 0.1611, 0.1200, 0.0981], device='cuda:5'), in_proj_covar=tensor([0.0710, 0.0862, 0.0703, 0.0662, 0.0547, 0.0543, 0.0726, 0.0677], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:20:05,706 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0293, 2.9499, 2.4060, 4.7423, 3.3890, 4.0899, 1.8421, 2.8335], device='cuda:5'), covar=tensor([0.1246, 0.0763, 0.1378, 0.0147, 0.0280, 0.0468, 0.1573, 0.0963], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0179, 0.0199, 0.0200, 0.0206, 0.0217, 0.0208, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 09:20:11,045 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 09:20:47,206 INFO [train.py:904] (5/8) Epoch 27, batch 5250, loss[loss=0.1705, simple_loss=0.264, pruned_loss=0.03856, over 16860.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2669, pruned_loss=0.04224, over 3182888.24 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:21:04,387 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 1.922e+02 2.287e+02 2.667e+02 4.356e+02, threshold=4.574e+02, percent-clipped=2.0 2023-05-02 09:22:00,572 INFO [train.py:904] (5/8) Epoch 27, batch 5300, loss[loss=0.1622, simple_loss=0.2506, pruned_loss=0.03688, over 12148.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2632, pruned_loss=0.04108, over 3183140.77 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:22:55,408 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:08,889 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269250.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:12,969 INFO [train.py:904] (5/8) Epoch 27, batch 5350, loss[loss=0.1689, simple_loss=0.2578, pruned_loss=0.03997, over 12386.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2612, pruned_loss=0.04036, over 3191821.55 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:23:18,865 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269257.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:29,780 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.050e+02 2.380e+02 2.740e+02 5.067e+02, threshold=4.761e+02, percent-clipped=1.0 2023-05-02 09:24:22,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4665, 2.8976, 3.0626, 2.0572, 2.7373, 2.1531, 2.9988, 3.2348], device='cuda:5'), covar=tensor([0.0393, 0.0838, 0.0631, 0.1991, 0.0942, 0.0989, 0.0767, 0.0871], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 09:24:24,480 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269302.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:24:25,147 INFO [train.py:904] (5/8) Epoch 27, batch 5400, loss[loss=0.182, simple_loss=0.2782, pruned_loss=0.0429, over 16900.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2642, pruned_loss=0.0413, over 3188081.48 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:24:28,456 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269305.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:24:51,994 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-05-02 09:25:14,093 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7662, 3.5959, 4.0895, 2.0358, 4.2591, 4.2392, 3.1011, 3.1705], device='cuda:5'), covar=tensor([0.0799, 0.0299, 0.0183, 0.1304, 0.0072, 0.0141, 0.0452, 0.0471], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0102, 0.0141, 0.0087, 0.0131, 0.0131, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 09:25:31,317 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:41,593 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:42,385 INFO [train.py:904] (5/8) Epoch 27, batch 5450, loss[loss=0.2088, simple_loss=0.2882, pruned_loss=0.06474, over 11987.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2667, pruned_loss=0.04271, over 3194256.09 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:26:01,151 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.011e+02 2.409e+02 3.005e+02 5.920e+02, threshold=4.819e+02, percent-clipped=1.0 2023-05-02 09:27:00,964 INFO [train.py:904] (5/8) Epoch 27, batch 5500, loss[loss=0.1943, simple_loss=0.2886, pruned_loss=0.05003, over 16488.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2738, pruned_loss=0.04758, over 3134997.69 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:27:16,731 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:28:00,757 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269441.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:28:19,429 INFO [train.py:904] (5/8) Epoch 27, batch 5550, loss[loss=0.2199, simple_loss=0.3118, pruned_loss=0.06401, over 16447.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2811, pruned_loss=0.05232, over 3117586.47 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:28:22,045 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2257, 2.3381, 2.2839, 3.8696, 2.2253, 2.7389, 2.3589, 2.5300], device='cuda:5'), covar=tensor([0.1390, 0.3412, 0.2987, 0.0592, 0.4043, 0.2371, 0.3509, 0.3103], device='cuda:5'), in_proj_covar=tensor([0.0419, 0.0470, 0.0382, 0.0335, 0.0444, 0.0539, 0.0441, 0.0551], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:28:26,572 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4083, 3.5482, 3.6965, 2.2462, 3.1998, 2.4179, 3.7415, 3.8359], device='cuda:5'), covar=tensor([0.0249, 0.0823, 0.0598, 0.2123, 0.0835, 0.1026, 0.0614, 0.1047], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 09:28:38,504 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.260e+02 3.053e+02 3.491e+02 4.207e+02 9.161e+02, threshold=6.983e+02, percent-clipped=6.0 2023-05-02 09:29:37,503 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269502.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:29:38,096 INFO [train.py:904] (5/8) Epoch 27, batch 5600, loss[loss=0.2254, simple_loss=0.3156, pruned_loss=0.06765, over 17034.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2857, pruned_loss=0.05627, over 3084947.72 frames. ], batch size: 55, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:30:56,829 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:31:01,866 INFO [train.py:904] (5/8) Epoch 27, batch 5650, loss[loss=0.189, simple_loss=0.2848, pruned_loss=0.04655, over 16916.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2903, pruned_loss=0.05975, over 3054332.13 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:31:20,025 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.426e+02 3.478e+02 4.295e+02 5.160e+02 1.255e+03, threshold=8.591e+02, percent-clipped=5.0 2023-05-02 09:31:31,205 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3900, 2.5600, 2.0707, 2.3061, 2.9508, 2.5160, 2.9480, 3.1097], device='cuda:5'), covar=tensor([0.0137, 0.0455, 0.0624, 0.0514, 0.0277, 0.0458, 0.0264, 0.0280], device='cuda:5'), in_proj_covar=tensor([0.0225, 0.0240, 0.0230, 0.0231, 0.0241, 0.0240, 0.0238, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:32:11,883 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:13,080 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269598.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:20,979 INFO [train.py:904] (5/8) Epoch 27, batch 5700, loss[loss=0.217, simple_loss=0.3144, pruned_loss=0.05984, over 16350.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2919, pruned_loss=0.06089, over 3059796.04 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:32:46,452 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 09:33:31,976 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269647.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:33:41,290 INFO [train.py:904] (5/8) Epoch 27, batch 5750, loss[loss=0.2262, simple_loss=0.2966, pruned_loss=0.07792, over 10809.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2949, pruned_loss=0.06242, over 3049314.46 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:33:56,051 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3641, 3.3677, 1.7989, 3.7300, 2.4871, 3.6310, 1.9338, 2.6949], device='cuda:5'), covar=tensor([0.0298, 0.0404, 0.2161, 0.0214, 0.0910, 0.0614, 0.2156, 0.0880], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0172, 0.0181, 0.0220, 0.0204, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 09:33:59,193 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 2.911e+02 3.444e+02 4.109e+02 8.393e+02, threshold=6.889e+02, percent-clipped=0.0 2023-05-02 09:34:50,304 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:35:02,102 INFO [train.py:904] (5/8) Epoch 27, batch 5800, loss[loss=0.1809, simple_loss=0.2705, pruned_loss=0.04563, over 16779.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2939, pruned_loss=0.06046, over 3062501.62 frames. ], batch size: 76, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:35:11,582 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269708.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:35:22,001 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4195, 4.5082, 4.3381, 3.9989, 4.0218, 4.4277, 4.1814, 4.1667], device='cuda:5'), covar=tensor([0.0662, 0.0679, 0.0293, 0.0339, 0.0847, 0.0491, 0.0647, 0.0669], device='cuda:5'), in_proj_covar=tensor([0.0307, 0.0461, 0.0358, 0.0362, 0.0358, 0.0417, 0.0246, 0.0429], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:36:07,881 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-02 09:36:21,270 INFO [train.py:904] (5/8) Epoch 27, batch 5850, loss[loss=0.2123, simple_loss=0.3003, pruned_loss=0.06213, over 16494.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2915, pruned_loss=0.05873, over 3083225.42 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:36:33,014 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8285, 2.7281, 2.5577, 4.6924, 3.5738, 4.1528, 1.6417, 3.0325], device='cuda:5'), covar=tensor([0.1354, 0.0801, 0.1251, 0.0168, 0.0270, 0.0372, 0.1666, 0.0828], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0179, 0.0199, 0.0200, 0.0207, 0.0217, 0.0208, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 09:36:37,036 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 09:36:40,942 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 3.085e+02 3.568e+02 4.510e+02 7.349e+02, threshold=7.136e+02, percent-clipped=2.0 2023-05-02 09:36:53,555 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:37:10,898 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0385, 2.4077, 2.5820, 1.9275, 2.6913, 2.7737, 2.4693, 2.3944], device='cuda:5'), covar=tensor([0.0740, 0.0323, 0.0250, 0.0993, 0.0146, 0.0306, 0.0471, 0.0471], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0140, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 09:37:23,664 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6913, 1.7929, 1.6302, 1.5183, 1.9656, 1.5945, 1.6213, 1.9562], device='cuda:5'), covar=tensor([0.0215, 0.0315, 0.0405, 0.0339, 0.0216, 0.0260, 0.0154, 0.0199], device='cuda:5'), in_proj_covar=tensor([0.0226, 0.0240, 0.0231, 0.0231, 0.0242, 0.0240, 0.0239, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:37:34,887 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269797.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:37:45,797 INFO [train.py:904] (5/8) Epoch 27, batch 5900, loss[loss=0.1848, simple_loss=0.2833, pruned_loss=0.0432, over 16547.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2911, pruned_loss=0.05867, over 3092544.96 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:38:38,107 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269834.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:38:40,760 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:39:06,826 INFO [train.py:904] (5/8) Epoch 27, batch 5950, loss[loss=0.213, simple_loss=0.2958, pruned_loss=0.06514, over 11574.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2916, pruned_loss=0.05779, over 3089616.03 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:39:21,982 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9246, 2.1816, 2.4641, 3.1121, 2.2385, 2.3722, 2.3504, 2.2675], device='cuda:5'), covar=tensor([0.1418, 0.3254, 0.2547, 0.0739, 0.3943, 0.2461, 0.3226, 0.3371], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0466, 0.0378, 0.0332, 0.0440, 0.0534, 0.0437, 0.0546], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:39:27,606 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.704e+02 3.273e+02 4.077e+02 6.463e+02, threshold=6.547e+02, percent-clipped=0.0 2023-05-02 09:40:17,660 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:17,757 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:24,920 INFO [train.py:904] (5/8) Epoch 27, batch 6000, loss[loss=0.1708, simple_loss=0.2599, pruned_loss=0.04087, over 16878.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2914, pruned_loss=0.05766, over 3096082.71 frames. ], batch size: 102, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:40:24,920 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 09:40:35,107 INFO [train.py:938] (5/8) Epoch 27, validation: loss=0.148, simple_loss=0.2603, pruned_loss=0.01783, over 944034.00 frames. 2023-05-02 09:40:35,108 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 09:41:37,580 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:41:52,127 INFO [train.py:904] (5/8) Epoch 27, batch 6050, loss[loss=0.2012, simple_loss=0.2957, pruned_loss=0.05329, over 17058.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2896, pruned_loss=0.05702, over 3107760.93 frames. ], batch size: 55, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:42:12,242 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.707e+02 3.068e+02 3.739e+02 6.756e+02, threshold=6.136e+02, percent-clipped=2.0 2023-05-02 09:42:52,939 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-05-02 09:43:13,685 INFO [train.py:904] (5/8) Epoch 27, batch 6100, loss[loss=0.1948, simple_loss=0.2831, pruned_loss=0.05327, over 16885.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2891, pruned_loss=0.05647, over 3100311.97 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:43:23,142 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270008.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:33,204 INFO [train.py:904] (5/8) Epoch 27, batch 6150, loss[loss=0.2347, simple_loss=0.3069, pruned_loss=0.08128, over 11380.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2871, pruned_loss=0.05545, over 3110442.98 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:44:37,805 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270056.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:53,896 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.602e+02 3.189e+02 4.066e+02 6.857e+02, threshold=6.378e+02, percent-clipped=4.0 2023-05-02 09:45:03,390 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8587, 5.1455, 4.9237, 4.8904, 4.6930, 4.6773, 4.6166, 5.2145], device='cuda:5'), covar=tensor([0.1263, 0.0848, 0.0960, 0.0916, 0.0790, 0.1071, 0.1207, 0.0867], device='cuda:5'), in_proj_covar=tensor([0.0704, 0.0853, 0.0695, 0.0656, 0.0538, 0.0537, 0.0717, 0.0668], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:45:42,290 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270097.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:45:51,131 INFO [train.py:904] (5/8) Epoch 27, batch 6200, loss[loss=0.2164, simple_loss=0.2967, pruned_loss=0.06799, over 16912.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.286, pruned_loss=0.05571, over 3097186.07 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:46:33,644 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270129.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:46:58,077 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:47:09,725 INFO [train.py:904] (5/8) Epoch 27, batch 6250, loss[loss=0.1747, simple_loss=0.2664, pruned_loss=0.04151, over 16325.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.285, pruned_loss=0.05479, over 3103933.14 frames. ], batch size: 35, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:47:29,203 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.710e+02 3.440e+02 4.218e+02 8.566e+02, threshold=6.880e+02, percent-clipped=4.0 2023-05-02 09:48:08,662 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:48:25,160 INFO [train.py:904] (5/8) Epoch 27, batch 6300, loss[loss=0.1867, simple_loss=0.2739, pruned_loss=0.0497, over 16956.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2846, pruned_loss=0.05397, over 3109113.17 frames. ], batch size: 109, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:48:46,549 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8048, 4.7896, 4.6075, 3.8248, 4.7291, 1.7171, 4.4883, 4.2547], device='cuda:5'), covar=tensor([0.0128, 0.0116, 0.0212, 0.0426, 0.0102, 0.2970, 0.0144, 0.0281], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0173, 0.0210, 0.0185, 0.0185, 0.0215, 0.0199, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:49:17,464 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5198, 4.4820, 4.3888, 3.5755, 4.4573, 1.7168, 4.1759, 3.9931], device='cuda:5'), covar=tensor([0.0132, 0.0115, 0.0189, 0.0369, 0.0104, 0.2954, 0.0142, 0.0277], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0173, 0.0210, 0.0185, 0.0185, 0.0215, 0.0199, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:49:45,991 INFO [train.py:904] (5/8) Epoch 27, batch 6350, loss[loss=0.1963, simple_loss=0.2843, pruned_loss=0.05419, over 16276.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2845, pruned_loss=0.05439, over 3110683.79 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:50:05,611 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.880e+02 2.767e+02 3.274e+02 3.776e+02 6.685e+02, threshold=6.549e+02, percent-clipped=0.0 2023-05-02 09:50:09,958 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 09:50:28,874 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270280.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:51:02,495 INFO [train.py:904] (5/8) Epoch 27, batch 6400, loss[loss=0.1858, simple_loss=0.2749, pruned_loss=0.04832, over 16649.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2846, pruned_loss=0.05527, over 3106239.56 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:51:32,592 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.91 vs. limit=5.0 2023-05-02 09:52:01,134 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270341.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:52:19,089 INFO [train.py:904] (5/8) Epoch 27, batch 6450, loss[loss=0.1878, simple_loss=0.2828, pruned_loss=0.04645, over 16782.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2851, pruned_loss=0.05482, over 3104397.66 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:52:39,053 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 2.688e+02 3.236e+02 4.371e+02 9.664e+02, threshold=6.472e+02, percent-clipped=9.0 2023-05-02 09:53:27,520 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270396.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:53:37,250 INFO [train.py:904] (5/8) Epoch 27, batch 6500, loss[loss=0.2211, simple_loss=0.2853, pruned_loss=0.07849, over 11740.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2835, pruned_loss=0.05458, over 3103723.30 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:54:17,096 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270429.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:54:19,296 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 09:54:33,524 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270439.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:54:58,130 INFO [train.py:904] (5/8) Epoch 27, batch 6550, loss[loss=0.1876, simple_loss=0.2917, pruned_loss=0.04177, over 16761.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2862, pruned_loss=0.05476, over 3118749.11 frames. ], batch size: 102, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:55:04,775 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:17,873 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 2.679e+02 3.213e+02 3.936e+02 7.742e+02, threshold=6.427e+02, percent-clipped=6.0 2023-05-02 09:55:34,672 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:58,298 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:56:10,646 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270500.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:56:14,135 INFO [train.py:904] (5/8) Epoch 27, batch 6600, loss[loss=0.2369, simple_loss=0.3089, pruned_loss=0.0825, over 11408.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2882, pruned_loss=0.05545, over 3111786.04 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:56:22,034 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9268, 2.1329, 2.3022, 3.3724, 2.0886, 2.3982, 2.2589, 2.2615], device='cuda:5'), covar=tensor([0.1613, 0.3843, 0.2909, 0.0760, 0.4412, 0.2637, 0.3665, 0.3522], device='cuda:5'), in_proj_covar=tensor([0.0417, 0.0468, 0.0380, 0.0333, 0.0442, 0.0536, 0.0439, 0.0548], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:57:09,777 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:57:09,914 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6523, 4.6840, 4.5462, 4.2057, 4.1974, 4.6348, 4.4166, 4.3384], device='cuda:5'), covar=tensor([0.0643, 0.0538, 0.0326, 0.0351, 0.1006, 0.0479, 0.0482, 0.0697], device='cuda:5'), in_proj_covar=tensor([0.0305, 0.0461, 0.0357, 0.0361, 0.0356, 0.0414, 0.0245, 0.0429], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:57:30,495 INFO [train.py:904] (5/8) Epoch 27, batch 6650, loss[loss=0.1664, simple_loss=0.2553, pruned_loss=0.03876, over 16502.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2888, pruned_loss=0.05676, over 3101480.05 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:57:50,324 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.739e+02 3.273e+02 3.972e+02 7.700e+02, threshold=6.545e+02, percent-clipped=3.0 2023-05-02 09:57:52,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0476, 4.0217, 3.9590, 3.1343, 3.9962, 1.7966, 3.7744, 3.4161], device='cuda:5'), covar=tensor([0.0133, 0.0114, 0.0191, 0.0337, 0.0101, 0.2992, 0.0143, 0.0340], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0173, 0.0210, 0.0185, 0.0185, 0.0215, 0.0199, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 09:58:46,119 INFO [train.py:904] (5/8) Epoch 27, batch 6700, loss[loss=0.1918, simple_loss=0.2799, pruned_loss=0.05178, over 16449.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.288, pruned_loss=0.05723, over 3094396.05 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:59:01,574 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:59:18,386 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4002, 4.5551, 4.7171, 4.4630, 4.5974, 5.0478, 4.5689, 4.3354], device='cuda:5'), covar=tensor([0.1512, 0.1902, 0.2264, 0.1901, 0.2274, 0.1036, 0.1707, 0.2389], device='cuda:5'), in_proj_covar=tensor([0.0424, 0.0629, 0.0688, 0.0511, 0.0680, 0.0717, 0.0540, 0.0686], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 09:59:27,504 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 09:59:36,018 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270636.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:00:01,102 INFO [train.py:904] (5/8) Epoch 27, batch 6750, loss[loss=0.1714, simple_loss=0.2617, pruned_loss=0.04054, over 16522.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2872, pruned_loss=0.05742, over 3091628.91 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:00:20,167 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 2.756e+02 3.273e+02 4.054e+02 1.247e+03, threshold=6.545e+02, percent-clipped=2.0 2023-05-02 10:00:32,956 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:00:53,483 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4525, 2.9457, 3.0280, 1.8720, 2.6386, 2.0815, 3.1019, 3.1959], device='cuda:5'), covar=tensor([0.0337, 0.0884, 0.0710, 0.2317, 0.1028, 0.1105, 0.0748, 0.0944], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 10:01:15,225 INFO [train.py:904] (5/8) Epoch 27, batch 6800, loss[loss=0.2058, simple_loss=0.3015, pruned_loss=0.05498, over 16456.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2876, pruned_loss=0.05804, over 3073518.44 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:01:53,403 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270726.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:33,337 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:34,176 INFO [train.py:904] (5/8) Epoch 27, batch 6850, loss[loss=0.1833, simple_loss=0.2933, pruned_loss=0.03666, over 16517.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2883, pruned_loss=0.05767, over 3089213.94 frames. ], batch size: 75, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:02:37,573 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1777, 3.8455, 3.8925, 2.4439, 3.6759, 3.9860, 3.7898, 2.0161], device='cuda:5'), covar=tensor([0.0657, 0.0073, 0.0064, 0.0476, 0.0090, 0.0113, 0.0087, 0.0582], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0090, 0.0090, 0.0136, 0.0101, 0.0115, 0.0098, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 10:02:39,153 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1686, 5.5013, 5.2309, 5.2504, 4.9821, 4.9829, 4.7744, 5.5810], device='cuda:5'), covar=tensor([0.1214, 0.0855, 0.1037, 0.0897, 0.0867, 0.0799, 0.1403, 0.0847], device='cuda:5'), in_proj_covar=tensor([0.0705, 0.0853, 0.0698, 0.0657, 0.0538, 0.0537, 0.0717, 0.0666], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:02:53,222 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.704e+02 3.220e+02 3.695e+02 5.610e+02, threshold=6.440e+02, percent-clipped=0.0 2023-05-02 10:03:24,147 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5120, 3.5762, 3.5518, 1.9530, 2.9550, 2.1546, 3.8788, 3.9265], device='cuda:5'), covar=tensor([0.0215, 0.0830, 0.0758, 0.2476, 0.1030, 0.1189, 0.0571, 0.0789], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0171, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 10:03:25,241 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:37,652 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270795.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:49,487 INFO [train.py:904] (5/8) Epoch 27, batch 6900, loss[loss=0.2042, simple_loss=0.2952, pruned_loss=0.0566, over 16959.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2906, pruned_loss=0.05712, over 3101801.17 frames. ], batch size: 109, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:04:39,271 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8727, 2.7200, 2.5557, 1.9192, 2.5449, 2.6602, 2.5417, 1.9339], device='cuda:5'), covar=tensor([0.0429, 0.0097, 0.0102, 0.0376, 0.0153, 0.0137, 0.0126, 0.0429], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0090, 0.0090, 0.0137, 0.0101, 0.0116, 0.0098, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 10:05:07,710 INFO [train.py:904] (5/8) Epoch 27, batch 6950, loss[loss=0.1834, simple_loss=0.2681, pruned_loss=0.04934, over 15420.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2922, pruned_loss=0.05871, over 3082566.63 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:05:28,451 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.915e+02 3.513e+02 4.432e+02 8.204e+02, threshold=7.026e+02, percent-clipped=5.0 2023-05-02 10:05:31,808 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0826, 2.4457, 2.5377, 1.9754, 2.6537, 2.7659, 2.4364, 2.3924], device='cuda:5'), covar=tensor([0.0679, 0.0280, 0.0233, 0.0884, 0.0133, 0.0277, 0.0455, 0.0440], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0110, 0.0101, 0.0139, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:06:06,619 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0736, 5.0176, 4.9095, 4.0428, 4.9380, 1.7737, 4.6518, 4.4574], device='cuda:5'), covar=tensor([0.0176, 0.0139, 0.0230, 0.0442, 0.0142, 0.3091, 0.0289, 0.0314], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0173, 0.0211, 0.0185, 0.0186, 0.0216, 0.0199, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:06:10,991 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0712, 5.3859, 5.1765, 5.1408, 4.9175, 4.8842, 4.6723, 5.4644], device='cuda:5'), covar=tensor([0.1292, 0.0823, 0.1002, 0.0995, 0.0849, 0.0865, 0.1514, 0.0895], device='cuda:5'), in_proj_covar=tensor([0.0700, 0.0847, 0.0694, 0.0652, 0.0534, 0.0534, 0.0713, 0.0663], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:06:23,858 INFO [train.py:904] (5/8) Epoch 27, batch 7000, loss[loss=0.1714, simple_loss=0.268, pruned_loss=0.03738, over 16766.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2917, pruned_loss=0.05733, over 3094148.19 frames. ], batch size: 39, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:06:38,382 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4681, 3.9997, 4.0253, 2.7021, 3.6670, 4.0547, 3.7283, 2.3423], device='cuda:5'), covar=tensor([0.0549, 0.0079, 0.0064, 0.0424, 0.0110, 0.0132, 0.0099, 0.0480], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0090, 0.0090, 0.0137, 0.0101, 0.0116, 0.0098, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 10:06:46,550 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:07:15,571 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270936.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 10:07:40,701 INFO [train.py:904] (5/8) Epoch 27, batch 7050, loss[loss=0.1893, simple_loss=0.2784, pruned_loss=0.0501, over 16733.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2919, pruned_loss=0.05708, over 3092515.27 frames. ], batch size: 57, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:08:01,040 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.738e+02 3.222e+02 3.975e+02 6.307e+02, threshold=6.443e+02, percent-clipped=0.0 2023-05-02 10:08:05,095 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:18,568 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:27,636 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270984.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:08:32,250 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3447, 5.3680, 5.1942, 4.8049, 4.8104, 5.2635, 5.1787, 4.9323], device='cuda:5'), covar=tensor([0.0607, 0.0443, 0.0309, 0.0336, 0.1103, 0.0428, 0.0295, 0.0680], device='cuda:5'), in_proj_covar=tensor([0.0304, 0.0459, 0.0356, 0.0360, 0.0356, 0.0413, 0.0244, 0.0427], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:08:58,245 INFO [train.py:904] (5/8) Epoch 27, batch 7100, loss[loss=0.1866, simple_loss=0.2817, pruned_loss=0.04577, over 16882.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2906, pruned_loss=0.05744, over 3080141.47 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:08:59,794 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 10:10:15,233 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:10:16,505 INFO [train.py:904] (5/8) Epoch 27, batch 7150, loss[loss=0.2108, simple_loss=0.2947, pruned_loss=0.06342, over 16414.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2896, pruned_loss=0.05743, over 3087137.42 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:10:37,971 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 2.616e+02 3.372e+02 3.961e+02 8.125e+02, threshold=6.744e+02, percent-clipped=3.0 2023-05-02 10:10:59,870 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:18,933 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:28,071 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271100.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:31,649 INFO [train.py:904] (5/8) Epoch 27, batch 7200, loss[loss=0.1658, simple_loss=0.2566, pruned_loss=0.03747, over 16685.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2876, pruned_loss=0.05649, over 3060511.98 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:12:37,997 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:12:53,498 INFO [train.py:904] (5/8) Epoch 27, batch 7250, loss[loss=0.1754, simple_loss=0.2643, pruned_loss=0.04325, over 16955.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2846, pruned_loss=0.05475, over 3062677.56 frames. ], batch size: 109, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:13:16,041 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.666e+02 3.085e+02 3.699e+02 8.603e+02, threshold=6.169e+02, percent-clipped=3.0 2023-05-02 10:14:10,781 INFO [train.py:904] (5/8) Epoch 27, batch 7300, loss[loss=0.2182, simple_loss=0.3038, pruned_loss=0.06629, over 16862.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2843, pruned_loss=0.0548, over 3068641.37 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:14:31,264 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-05-02 10:15:29,744 INFO [train.py:904] (5/8) Epoch 27, batch 7350, loss[loss=0.1911, simple_loss=0.2757, pruned_loss=0.05321, over 15217.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2849, pruned_loss=0.05543, over 3048573.18 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:15:42,286 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:15:50,538 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.875e+02 3.303e+02 4.103e+02 7.171e+02, threshold=6.606e+02, percent-clipped=3.0 2023-05-02 10:15:54,740 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:01,224 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:48,732 INFO [train.py:904] (5/8) Epoch 27, batch 7400, loss[loss=0.1934, simple_loss=0.2843, pruned_loss=0.0512, over 17030.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2865, pruned_loss=0.05616, over 3067411.35 frames. ], batch size: 55, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:17:11,160 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:17:18,865 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:17:27,879 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-05-02 10:17:38,744 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9539, 3.3962, 3.3872, 2.1658, 3.1851, 3.3954, 3.1653, 1.9939], device='cuda:5'), covar=tensor([0.0619, 0.0080, 0.0080, 0.0482, 0.0128, 0.0150, 0.0127, 0.0505], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0134, 0.0099, 0.0114, 0.0097, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 10:18:06,480 INFO [train.py:904] (5/8) Epoch 27, batch 7450, loss[loss=0.2139, simple_loss=0.31, pruned_loss=0.05892, over 15460.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2872, pruned_loss=0.0571, over 3058077.68 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:18:11,070 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3760, 3.2092, 3.5983, 1.7998, 3.6776, 3.7851, 2.8716, 2.7636], device='cuda:5'), covar=tensor([0.0828, 0.0307, 0.0227, 0.1276, 0.0117, 0.0207, 0.0493, 0.0498], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0110, 0.0102, 0.0139, 0.0086, 0.0131, 0.0131, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:18:26,105 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0232, 2.3131, 2.3022, 2.8083, 1.9143, 3.2145, 1.8471, 2.7122], device='cuda:5'), covar=tensor([0.1168, 0.0691, 0.1138, 0.0220, 0.0124, 0.0395, 0.1500, 0.0724], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0179, 0.0200, 0.0200, 0.0207, 0.0218, 0.0209, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:18:30,891 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.819e+02 3.560e+02 4.384e+02 9.484e+02, threshold=7.119e+02, percent-clipped=1.0 2023-05-02 10:18:45,273 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:18:55,985 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271382.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:19:30,459 INFO [train.py:904] (5/8) Epoch 27, batch 7500, loss[loss=0.1973, simple_loss=0.2852, pruned_loss=0.05465, over 15361.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2873, pruned_loss=0.05626, over 3059261.34 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:19:52,268 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 10:20:08,142 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:14,064 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:22,939 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271436.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:37,584 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8454, 3.9055, 4.1400, 4.1141, 4.1244, 3.8977, 3.9227, 3.9444], device='cuda:5'), covar=tensor([0.0381, 0.0744, 0.0443, 0.0455, 0.0478, 0.0485, 0.0842, 0.0555], device='cuda:5'), in_proj_covar=tensor([0.0424, 0.0475, 0.0462, 0.0422, 0.0508, 0.0486, 0.0558, 0.0390], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 10:20:49,676 INFO [train.py:904] (5/8) Epoch 27, batch 7550, loss[loss=0.1796, simple_loss=0.2709, pruned_loss=0.0442, over 16797.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2862, pruned_loss=0.05649, over 3060019.35 frames. ], batch size: 83, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:21:01,224 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7921, 3.8309, 3.9449, 3.6795, 3.8137, 4.2215, 3.8655, 3.6011], device='cuda:5'), covar=tensor([0.2020, 0.2396, 0.2571, 0.2434, 0.2675, 0.1728, 0.1752, 0.2594], device='cuda:5'), in_proj_covar=tensor([0.0424, 0.0631, 0.0692, 0.0514, 0.0685, 0.0722, 0.0544, 0.0690], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 10:21:11,194 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.650e+02 3.300e+02 4.210e+02 7.541e+02, threshold=6.599e+02, percent-clipped=2.0 2023-05-02 10:21:34,046 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8192, 1.4231, 1.7446, 1.6942, 1.8015, 1.9114, 1.6662, 1.8109], device='cuda:5'), covar=tensor([0.0295, 0.0420, 0.0238, 0.0325, 0.0302, 0.0202, 0.0471, 0.0157], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0195, 0.0183, 0.0188, 0.0203, 0.0163, 0.0200, 0.0162], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:21:41,020 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:22:05,504 INFO [train.py:904] (5/8) Epoch 27, batch 7600, loss[loss=0.1857, simple_loss=0.2697, pruned_loss=0.05087, over 16377.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2845, pruned_loss=0.05606, over 3051838.04 frames. ], batch size: 35, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:22,937 INFO [train.py:904] (5/8) Epoch 27, batch 7650, loss[loss=0.2394, simple_loss=0.3022, pruned_loss=0.08833, over 11199.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2849, pruned_loss=0.05632, over 3063902.34 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:45,489 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.699e+02 2.765e+02 3.286e+02 4.189e+02 6.927e+02, threshold=6.573e+02, percent-clipped=1.0 2023-05-02 10:23:55,642 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:24:43,701 INFO [train.py:904] (5/8) Epoch 27, batch 7700, loss[loss=0.1962, simple_loss=0.2778, pruned_loss=0.05729, over 16778.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.285, pruned_loss=0.05657, over 3084224.09 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:25:06,155 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271617.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:25:12,495 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:25:32,601 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8408, 2.0188, 2.4337, 3.0466, 2.1745, 2.2367, 2.2298, 2.1536], device='cuda:5'), covar=tensor([0.1703, 0.3900, 0.2592, 0.0921, 0.4732, 0.2799, 0.3602, 0.3912], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0470, 0.0380, 0.0332, 0.0443, 0.0536, 0.0441, 0.0547], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:26:02,424 INFO [train.py:904] (5/8) Epoch 27, batch 7750, loss[loss=0.1974, simple_loss=0.2899, pruned_loss=0.05245, over 16468.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2847, pruned_loss=0.05626, over 3087087.23 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:26:24,501 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.896e+02 3.348e+02 4.032e+02 8.871e+02, threshold=6.696e+02, percent-clipped=2.0 2023-05-02 10:27:20,094 INFO [train.py:904] (5/8) Epoch 27, batch 7800, loss[loss=0.2478, simple_loss=0.3097, pruned_loss=0.09295, over 11181.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2855, pruned_loss=0.05634, over 3096031.99 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:28:04,772 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271731.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:28:36,664 INFO [train.py:904] (5/8) Epoch 27, batch 7850, loss[loss=0.2042, simple_loss=0.2896, pruned_loss=0.05944, over 16716.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2865, pruned_loss=0.05666, over 3077709.21 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:28:41,498 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6988, 3.8063, 2.5079, 4.4859, 3.0259, 4.3590, 2.6123, 3.0791], device='cuda:5'), covar=tensor([0.0327, 0.0407, 0.1677, 0.0190, 0.0841, 0.0553, 0.1506, 0.0879], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0181, 0.0198, 0.0172, 0.0180, 0.0220, 0.0204, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:28:57,996 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.838e+02 3.379e+02 4.179e+02 9.334e+02, threshold=6.757e+02, percent-clipped=5.0 2023-05-02 10:29:11,968 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1078, 3.3431, 3.5091, 2.0449, 3.0510, 2.2661, 3.5615, 3.5820], device='cuda:5'), covar=tensor([0.0248, 0.0807, 0.0592, 0.2177, 0.0840, 0.1069, 0.0579, 0.0901], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0169, 0.0169, 0.0155, 0.0147, 0.0132, 0.0145, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:29:20,884 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:29:27,941 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6318, 4.3667, 4.2385, 2.9869, 3.7899, 4.2816, 3.7288, 2.4318], device='cuda:5'), covar=tensor([0.0543, 0.0049, 0.0056, 0.0397, 0.0106, 0.0120, 0.0108, 0.0472], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0134, 0.0099, 0.0113, 0.0096, 0.0128], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 10:29:47,278 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8143, 4.8643, 4.6966, 4.3448, 4.3603, 4.7669, 4.6419, 4.4410], device='cuda:5'), covar=tensor([0.0682, 0.0601, 0.0345, 0.0355, 0.1018, 0.0586, 0.0415, 0.0739], device='cuda:5'), in_proj_covar=tensor([0.0301, 0.0454, 0.0353, 0.0355, 0.0352, 0.0409, 0.0243, 0.0423], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:29:52,549 INFO [train.py:904] (5/8) Epoch 27, batch 7900, loss[loss=0.2244, simple_loss=0.291, pruned_loss=0.07889, over 11539.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2859, pruned_loss=0.05604, over 3074001.94 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:30:12,254 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 10:30:59,369 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1109, 3.5063, 3.4917, 2.2458, 3.2435, 3.5457, 3.2747, 2.0400], device='cuda:5'), covar=tensor([0.0576, 0.0073, 0.0072, 0.0454, 0.0118, 0.0124, 0.0119, 0.0523], device='cuda:5'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0134, 0.0099, 0.0113, 0.0096, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 10:31:12,171 INFO [train.py:904] (5/8) Epoch 27, batch 7950, loss[loss=0.1972, simple_loss=0.2945, pruned_loss=0.05, over 16465.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2865, pruned_loss=0.05645, over 3072096.92 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:31:15,329 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:31:27,613 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3395, 3.5123, 2.1502, 3.8454, 2.6215, 3.8351, 2.2218, 2.7506], device='cuda:5'), covar=tensor([0.0369, 0.0451, 0.1757, 0.0252, 0.0898, 0.0612, 0.1653, 0.0930], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0172, 0.0180, 0.0220, 0.0204, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:31:36,732 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.764e+02 3.233e+02 3.778e+02 5.723e+02, threshold=6.466e+02, percent-clipped=0.0 2023-05-02 10:32:29,919 INFO [train.py:904] (5/8) Epoch 27, batch 8000, loss[loss=0.1894, simple_loss=0.2776, pruned_loss=0.05058, over 16712.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2872, pruned_loss=0.05736, over 3058534.72 frames. ], batch size: 134, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:32:49,759 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271915.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:32:52,891 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:03,304 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 10:33:19,529 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:47,191 INFO [train.py:904] (5/8) Epoch 27, batch 8050, loss[loss=0.2035, simple_loss=0.2936, pruned_loss=0.05677, over 16699.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2867, pruned_loss=0.05699, over 3075782.34 frames. ], batch size: 76, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:34:05,655 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:34:09,136 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.577e+02 3.144e+02 3.701e+02 6.468e+02, threshold=6.287e+02, percent-clipped=1.0 2023-05-02 10:34:49,579 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:35:04,679 INFO [train.py:904] (5/8) Epoch 27, batch 8100, loss[loss=0.1945, simple_loss=0.2798, pruned_loss=0.05459, over 16659.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2867, pruned_loss=0.05639, over 3094273.16 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:35:45,705 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:36:17,772 INFO [train.py:904] (5/8) Epoch 27, batch 8150, loss[loss=0.2069, simple_loss=0.2862, pruned_loss=0.06379, over 16842.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2842, pruned_loss=0.05544, over 3096815.25 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:36:39,766 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.743e+02 3.279e+02 3.939e+02 6.192e+02, threshold=6.559e+02, percent-clipped=0.0 2023-05-02 10:36:50,654 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 10:36:57,284 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:01,655 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:06,691 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4694, 1.7109, 2.1384, 2.4250, 2.4443, 2.7250, 1.7896, 2.6720], device='cuda:5'), covar=tensor([0.0292, 0.0648, 0.0371, 0.0432, 0.0368, 0.0254, 0.0684, 0.0192], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0190, 0.0206, 0.0165, 0.0202, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:37:32,132 INFO [train.py:904] (5/8) Epoch 27, batch 8200, loss[loss=0.1751, simple_loss=0.2721, pruned_loss=0.03903, over 16341.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2817, pruned_loss=0.05435, over 3107206.32 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:38:11,765 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 10:38:12,515 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:16,450 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:18,485 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5654, 3.6676, 2.4868, 4.1779, 2.8813, 4.1027, 2.5634, 3.0742], device='cuda:5'), covar=tensor([0.0318, 0.0410, 0.1485, 0.0269, 0.0826, 0.0548, 0.1381, 0.0709], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0181, 0.0198, 0.0172, 0.0180, 0.0221, 0.0204, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:38:30,497 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8197, 3.5547, 3.8360, 2.1238, 4.0018, 4.0480, 3.1379, 3.2703], device='cuda:5'), covar=tensor([0.0669, 0.0237, 0.0242, 0.1135, 0.0095, 0.0197, 0.0450, 0.0389], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0110, 0.0102, 0.0139, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:38:53,596 INFO [train.py:904] (5/8) Epoch 27, batch 8250, loss[loss=0.1534, simple_loss=0.2494, pruned_loss=0.0287, over 11997.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2809, pruned_loss=0.05216, over 3079046.37 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:39:19,038 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.668e+02 3.172e+02 3.719e+02 6.913e+02, threshold=6.344e+02, percent-clipped=1.0 2023-05-02 10:39:40,205 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3428, 3.4416, 2.1663, 3.7441, 2.5983, 3.7164, 2.3300, 2.8280], device='cuda:5'), covar=tensor([0.0285, 0.0358, 0.1551, 0.0253, 0.0786, 0.0528, 0.1413, 0.0700], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0171, 0.0179, 0.0220, 0.0204, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 10:39:40,208 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272181.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:39:50,809 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272188.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:14,330 INFO [train.py:904] (5/8) Epoch 27, batch 8300, loss[loss=0.1804, simple_loss=0.2751, pruned_loss=0.04288, over 16682.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2783, pruned_loss=0.04946, over 3064496.61 frames. ], batch size: 134, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:40:26,374 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:46,111 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:54,438 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 10:41:15,176 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272242.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:41:23,025 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272247.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:31,415 INFO [train.py:904] (5/8) Epoch 27, batch 8350, loss[loss=0.1881, simple_loss=0.2873, pruned_loss=0.04446, over 15315.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2776, pruned_loss=0.0473, over 3082806.90 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:41:54,866 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.144e+02 2.517e+02 3.037e+02 5.148e+02, threshold=5.034e+02, percent-clipped=0.0 2023-05-02 10:42:20,865 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272284.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:30,962 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272290.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:50,445 INFO [train.py:904] (5/8) Epoch 27, batch 8400, loss[loss=0.1653, simple_loss=0.263, pruned_loss=0.03377, over 16604.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2748, pruned_loss=0.04581, over 3053367.61 frames. ], batch size: 62, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:42:59,289 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:43:06,004 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 10:44:09,899 INFO [train.py:904] (5/8) Epoch 27, batch 8450, loss[loss=0.1524, simple_loss=0.2433, pruned_loss=0.0307, over 12410.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2725, pruned_loss=0.04376, over 3041307.96 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:44:34,140 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 2.148e+02 2.573e+02 3.182e+02 7.232e+02, threshold=5.146e+02, percent-clipped=1.0 2023-05-02 10:44:42,835 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272373.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:45:27,627 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 10:45:32,400 INFO [train.py:904] (5/8) Epoch 27, batch 8500, loss[loss=0.1514, simple_loss=0.2552, pruned_loss=0.02376, over 16929.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2692, pruned_loss=0.04148, over 3068295.73 frames. ], batch size: 90, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:45:33,988 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 10:45:50,622 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 10:45:58,841 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4518, 2.9795, 2.7878, 2.3402, 2.2717, 2.4200, 2.9943, 2.8539], device='cuda:5'), covar=tensor([0.2505, 0.0665, 0.1563, 0.2915, 0.2621, 0.2286, 0.0444, 0.1433], device='cuda:5'), in_proj_covar=tensor([0.0328, 0.0269, 0.0306, 0.0320, 0.0299, 0.0270, 0.0298, 0.0342], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 10:46:24,126 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:46:57,058 INFO [train.py:904] (5/8) Epoch 27, batch 8550, loss[loss=0.1649, simple_loss=0.2474, pruned_loss=0.04113, over 11926.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2667, pruned_loss=0.04054, over 3041077.11 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:47:25,010 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.247e+02 2.591e+02 3.219e+02 6.393e+02, threshold=5.181e+02, percent-clipped=2.0 2023-05-02 10:47:55,111 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272483.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:48:35,744 INFO [train.py:904] (5/8) Epoch 27, batch 8600, loss[loss=0.1641, simple_loss=0.2681, pruned_loss=0.03011, over 16740.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2675, pruned_loss=0.04025, over 3027992.85 frames. ], batch size: 76, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:48:50,723 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272510.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:49:43,702 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272537.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:50:13,945 INFO [train.py:904] (5/8) Epoch 27, batch 8650, loss[loss=0.1763, simple_loss=0.2762, pruned_loss=0.03822, over 16430.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2654, pruned_loss=0.03898, over 3002115.43 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:50:27,064 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:50:48,992 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.181e+02 2.564e+02 3.046e+02 5.835e+02, threshold=5.129e+02, percent-clipped=3.0 2023-05-02 10:51:15,239 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:51:36,706 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:00,252 INFO [train.py:904] (5/8) Epoch 27, batch 8700, loss[loss=0.1638, simple_loss=0.2519, pruned_loss=0.03784, over 12401.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2632, pruned_loss=0.03787, over 3005963.02 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:52:01,244 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:21,627 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:44,467 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 10:53:06,583 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272638.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:53:35,419 INFO [train.py:904] (5/8) Epoch 27, batch 8750, loss[loss=0.1593, simple_loss=0.2489, pruned_loss=0.03491, over 12133.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2628, pruned_loss=0.03696, over 3033175.05 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 10:54:15,882 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.360e+02 2.765e+02 3.360e+02 6.939e+02, threshold=5.529e+02, percent-clipped=4.0 2023-05-02 10:54:28,961 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:54:52,888 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 10:55:27,078 INFO [train.py:904] (5/8) Epoch 27, batch 8800, loss[loss=0.1826, simple_loss=0.2694, pruned_loss=0.04793, over 12277.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2618, pruned_loss=0.03597, over 3044209.47 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:55:28,538 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8754, 3.7781, 3.9305, 4.0137, 4.1014, 3.7067, 4.0580, 4.1285], device='cuda:5'), covar=tensor([0.1497, 0.1063, 0.1213, 0.0703, 0.0563, 0.1791, 0.0745, 0.0640], device='cuda:5'), in_proj_covar=tensor([0.0638, 0.0782, 0.0908, 0.0793, 0.0608, 0.0635, 0.0664, 0.0775], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:56:23,110 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:57:06,116 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6273, 3.6923, 3.4621, 3.1506, 3.2820, 3.5820, 3.3900, 3.4703], device='cuda:5'), covar=tensor([0.0478, 0.0436, 0.0284, 0.0259, 0.0475, 0.0396, 0.1369, 0.0441], device='cuda:5'), in_proj_covar=tensor([0.0295, 0.0445, 0.0346, 0.0348, 0.0344, 0.0400, 0.0238, 0.0413], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:57:12,470 INFO [train.py:904] (5/8) Epoch 27, batch 8850, loss[loss=0.1468, simple_loss=0.2414, pruned_loss=0.02611, over 12113.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2648, pruned_loss=0.03568, over 3046649.80 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:57:46,549 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.085e+02 2.509e+02 2.907e+02 5.151e+02, threshold=5.018e+02, percent-clipped=0.0 2023-05-02 10:57:51,868 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-02 10:58:17,428 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272783.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:58:29,363 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0482, 5.1224, 4.9005, 4.5073, 4.5336, 4.9923, 4.9476, 4.6341], device='cuda:5'), covar=tensor([0.0674, 0.0547, 0.0376, 0.0379, 0.1133, 0.0542, 0.0327, 0.0815], device='cuda:5'), in_proj_covar=tensor([0.0295, 0.0445, 0.0346, 0.0349, 0.0344, 0.0400, 0.0239, 0.0413], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 10:58:45,387 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-05-02 10:58:57,772 INFO [train.py:904] (5/8) Epoch 27, batch 8900, loss[loss=0.1548, simple_loss=0.2541, pruned_loss=0.02772, over 16679.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2653, pruned_loss=0.03526, over 3065470.80 frames. ], batch size: 89, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:00:01,537 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272831.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:00:20,735 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272837.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 11:01:00,400 INFO [train.py:904] (5/8) Epoch 27, batch 8950, loss[loss=0.1597, simple_loss=0.2575, pruned_loss=0.03097, over 16151.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2648, pruned_loss=0.03533, over 3076491.56 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:01:35,340 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 1.963e+02 2.469e+02 3.081e+02 5.293e+02, threshold=4.938e+02, percent-clipped=1.0 2023-05-02 11:01:56,905 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272879.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:02:12,417 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272885.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 11:02:37,400 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7258, 6.1023, 5.8371, 5.9436, 5.5493, 5.4309, 5.5297, 6.2156], device='cuda:5'), covar=tensor([0.1496, 0.0888, 0.1023, 0.0839, 0.0841, 0.0634, 0.1325, 0.0825], device='cuda:5'), in_proj_covar=tensor([0.0693, 0.0839, 0.0688, 0.0644, 0.0528, 0.0530, 0.0702, 0.0656], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:02:42,327 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8820, 4.8501, 4.6549, 4.0610, 4.7004, 1.9247, 4.4418, 4.4349], device='cuda:5'), covar=tensor([0.0092, 0.0094, 0.0198, 0.0372, 0.0111, 0.2791, 0.0138, 0.0255], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0169, 0.0206, 0.0180, 0.0183, 0.0213, 0.0195, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:02:49,718 INFO [train.py:904] (5/8) Epoch 27, batch 9000, loss[loss=0.1487, simple_loss=0.2386, pruned_loss=0.02942, over 16972.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2611, pruned_loss=0.03398, over 3060633.27 frames. ], batch size: 116, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:02:49,719 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 11:02:59,739 INFO [train.py:938] (5/8) Epoch 27, validation: loss=0.1436, simple_loss=0.2474, pruned_loss=0.01989, over 944034.00 frames. 2023-05-02 11:02:59,741 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 11:03:00,778 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:03:26,507 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-05-02 11:03:51,522 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:01,240 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-02 11:04:42,537 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272951.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:44,610 INFO [train.py:904] (5/8) Epoch 27, batch 9050, loss[loss=0.1744, simple_loss=0.2625, pruned_loss=0.04309, over 12538.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2616, pruned_loss=0.03424, over 3051572.54 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:05:18,776 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 2.208e+02 2.579e+02 3.072e+02 5.940e+02, threshold=5.159e+02, percent-clipped=5.0 2023-05-02 11:05:19,960 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:06:29,755 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1502, 2.2014, 2.1188, 3.7530, 2.0640, 2.4681, 2.2973, 2.3096], device='cuda:5'), covar=tensor([0.1413, 0.3859, 0.3629, 0.0609, 0.4689, 0.2787, 0.3908, 0.3602], device='cuda:5'), in_proj_covar=tensor([0.0408, 0.0460, 0.0376, 0.0324, 0.0435, 0.0525, 0.0433, 0.0536], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:06:32,230 INFO [train.py:904] (5/8) Epoch 27, batch 9100, loss[loss=0.1652, simple_loss=0.2661, pruned_loss=0.03215, over 16681.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2615, pruned_loss=0.03501, over 3043894.88 frames. ], batch size: 62, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:07:36,909 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273029.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:08:32,223 INFO [train.py:904] (5/8) Epoch 27, batch 9150, loss[loss=0.1611, simple_loss=0.2514, pruned_loss=0.03535, over 16956.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2616, pruned_loss=0.03456, over 3035870.12 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:09:07,676 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-05-02 11:09:08,134 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.260e+02 2.634e+02 3.413e+02 5.711e+02, threshold=5.268e+02, percent-clipped=2.0 2023-05-02 11:09:23,280 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8184, 3.0653, 3.4133, 2.0131, 2.9002, 2.1959, 3.2861, 3.1946], device='cuda:5'), covar=tensor([0.0270, 0.0903, 0.0552, 0.2179, 0.0817, 0.1001, 0.0687, 0.1093], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0165, 0.0165, 0.0153, 0.0144, 0.0130, 0.0142, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:09:27,810 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273077.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:10:18,360 INFO [train.py:904] (5/8) Epoch 27, batch 9200, loss[loss=0.166, simple_loss=0.2612, pruned_loss=0.03535, over 16405.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2579, pruned_loss=0.0341, over 3037799.53 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:11:13,767 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 11:11:53,372 INFO [train.py:904] (5/8) Epoch 27, batch 9250, loss[loss=0.142, simple_loss=0.2383, pruned_loss=0.02291, over 17015.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.258, pruned_loss=0.03455, over 3035606.92 frames. ], batch size: 50, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:12:25,535 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.225e+02 2.595e+02 3.229e+02 5.341e+02, threshold=5.191e+02, percent-clipped=1.0 2023-05-02 11:13:43,828 INFO [train.py:904] (5/8) Epoch 27, batch 9300, loss[loss=0.1525, simple_loss=0.2468, pruned_loss=0.02909, over 16662.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2564, pruned_loss=0.03389, over 3059561.77 frames. ], batch size: 134, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:14:29,841 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 11:14:35,011 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5911, 2.5218, 2.2536, 3.8335, 2.2725, 3.7775, 1.4524, 2.8242], device='cuda:5'), covar=tensor([0.1568, 0.0873, 0.1357, 0.0199, 0.0135, 0.0395, 0.1885, 0.0813], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0177, 0.0197, 0.0195, 0.0202, 0.0214, 0.0207, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:15:28,950 INFO [train.py:904] (5/8) Epoch 27, batch 9350, loss[loss=0.1536, simple_loss=0.2433, pruned_loss=0.03189, over 12381.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2561, pruned_loss=0.03372, over 3070035.55 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:16:02,980 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.192e+02 2.614e+02 3.426e+02 6.584e+02, threshold=5.229e+02, percent-clipped=2.0 2023-05-02 11:16:03,915 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:16:53,606 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7839, 2.9744, 3.3667, 1.9707, 2.8926, 2.1539, 3.2873, 3.1294], device='cuda:5'), covar=tensor([0.0274, 0.0919, 0.0545, 0.2224, 0.0803, 0.1030, 0.0646, 0.1077], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0164, 0.0165, 0.0153, 0.0144, 0.0130, 0.0142, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:17:10,314 INFO [train.py:904] (5/8) Epoch 27, batch 9400, loss[loss=0.1807, simple_loss=0.2816, pruned_loss=0.03987, over 16840.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2559, pruned_loss=0.03339, over 3068174.45 frames. ], batch size: 124, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:17:11,951 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7634, 3.0863, 3.4021, 1.9795, 2.8462, 2.1813, 3.1717, 3.2535], device='cuda:5'), covar=tensor([0.0388, 0.0975, 0.0579, 0.2308, 0.0929, 0.1087, 0.0914, 0.1117], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0164, 0.0165, 0.0153, 0.0144, 0.0129, 0.0142, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:17:39,649 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:18:18,029 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5683, 1.6422, 2.1570, 2.5192, 2.4728, 2.8375, 1.6571, 2.7717], device='cuda:5'), covar=tensor([0.0239, 0.0734, 0.0441, 0.0337, 0.0387, 0.0227, 0.0864, 0.0179], device='cuda:5'), in_proj_covar=tensor([0.0192, 0.0194, 0.0182, 0.0185, 0.0202, 0.0160, 0.0198, 0.0160], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:18:51,374 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6333, 4.4193, 4.6665, 4.8115, 4.9929, 4.5028, 4.9870, 4.9972], device='cuda:5'), covar=tensor([0.1933, 0.1513, 0.1766, 0.0790, 0.0589, 0.1145, 0.0648, 0.0696], device='cuda:5'), in_proj_covar=tensor([0.0639, 0.0779, 0.0903, 0.0793, 0.0608, 0.0633, 0.0665, 0.0773], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:18:52,069 INFO [train.py:904] (5/8) Epoch 27, batch 9450, loss[loss=0.1561, simple_loss=0.2577, pruned_loss=0.02724, over 16715.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2587, pruned_loss=0.0335, over 3089186.77 frames. ], batch size: 76, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:19:21,971 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.197e+02 2.626e+02 3.206e+02 9.861e+02, threshold=5.251e+02, percent-clipped=1.0 2023-05-02 11:20:32,370 INFO [train.py:904] (5/8) Epoch 27, batch 9500, loss[loss=0.147, simple_loss=0.2366, pruned_loss=0.02869, over 17032.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2572, pruned_loss=0.03298, over 3094616.57 frames. ], batch size: 55, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:20:49,209 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273409.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:21:25,433 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-05-02 11:21:38,402 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273435.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:22:07,383 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9442, 5.0043, 5.3501, 5.3548, 5.3717, 5.0919, 5.0037, 4.9216], device='cuda:5'), covar=tensor([0.0491, 0.1449, 0.0707, 0.0606, 0.0651, 0.0567, 0.1129, 0.0509], device='cuda:5'), in_proj_covar=tensor([0.0412, 0.0461, 0.0449, 0.0413, 0.0495, 0.0473, 0.0541, 0.0380], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 11:22:07,520 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0231, 2.9273, 2.7770, 4.9923, 3.4898, 4.4453, 1.7385, 3.3812], device='cuda:5'), covar=tensor([0.1375, 0.0868, 0.1174, 0.0161, 0.0215, 0.0382, 0.1756, 0.0686], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0176, 0.0196, 0.0194, 0.0200, 0.0213, 0.0206, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:22:16,785 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9343, 2.1738, 2.3383, 3.1756, 2.1880, 2.3510, 2.2925, 2.2568], device='cuda:5'), covar=tensor([0.1500, 0.4242, 0.3193, 0.0876, 0.4793, 0.2885, 0.4209, 0.4039], device='cuda:5'), in_proj_covar=tensor([0.0409, 0.0461, 0.0378, 0.0326, 0.0437, 0.0526, 0.0435, 0.0537], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:22:18,635 INFO [train.py:904] (5/8) Epoch 27, batch 9550, loss[loss=0.1747, simple_loss=0.2714, pruned_loss=0.03903, over 16745.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2564, pruned_loss=0.03306, over 3091671.90 frames. ], batch size: 124, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:22:19,368 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2038, 2.9367, 3.1592, 1.8501, 3.2435, 3.3430, 2.8307, 2.6343], device='cuda:5'), covar=tensor([0.0775, 0.0266, 0.0215, 0.1165, 0.0108, 0.0201, 0.0447, 0.0482], device='cuda:5'), in_proj_covar=tensor([0.0143, 0.0106, 0.0097, 0.0134, 0.0082, 0.0124, 0.0125, 0.0125], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 11:22:51,235 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 2.156e+02 2.505e+02 2.980e+02 5.260e+02, threshold=5.009e+02, percent-clipped=1.0 2023-05-02 11:22:54,290 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:46,476 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:57,856 INFO [train.py:904] (5/8) Epoch 27, batch 9600, loss[loss=0.1525, simple_loss=0.2419, pruned_loss=0.03159, over 12368.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2575, pruned_loss=0.03361, over 3080215.36 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:25:44,223 INFO [train.py:904] (5/8) Epoch 27, batch 9650, loss[loss=0.1743, simple_loss=0.2664, pruned_loss=0.04109, over 16948.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2597, pruned_loss=0.03411, over 3088514.97 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:26:23,246 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.167e+02 2.602e+02 2.993e+02 6.217e+02, threshold=5.204e+02, percent-clipped=3.0 2023-05-02 11:26:26,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1828, 3.9730, 4.2035, 4.3468, 4.4826, 4.0343, 4.4760, 4.5007], device='cuda:5'), covar=tensor([0.1743, 0.1276, 0.1690, 0.0782, 0.0563, 0.1259, 0.0646, 0.0710], device='cuda:5'), in_proj_covar=tensor([0.0638, 0.0779, 0.0902, 0.0792, 0.0607, 0.0633, 0.0664, 0.0773], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:27:30,465 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-05-02 11:27:30,807 INFO [train.py:904] (5/8) Epoch 27, batch 9700, loss[loss=0.1964, simple_loss=0.279, pruned_loss=0.0569, over 16958.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2591, pruned_loss=0.03427, over 3080945.19 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:28:50,545 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1773, 1.6006, 1.9630, 2.1429, 2.2446, 2.3575, 1.8307, 2.3026], device='cuda:5'), covar=tensor([0.0285, 0.0594, 0.0338, 0.0376, 0.0384, 0.0262, 0.0581, 0.0188], device='cuda:5'), in_proj_covar=tensor([0.0191, 0.0193, 0.0180, 0.0184, 0.0200, 0.0159, 0.0197, 0.0159], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:29:12,478 INFO [train.py:904] (5/8) Epoch 27, batch 9750, loss[loss=0.194, simple_loss=0.2912, pruned_loss=0.04837, over 16886.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2582, pruned_loss=0.03437, over 3081034.09 frames. ], batch size: 116, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:29:42,116 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 2.115e+02 2.683e+02 3.309e+02 9.068e+02, threshold=5.365e+02, percent-clipped=4.0 2023-05-02 11:30:41,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6949, 4.2958, 2.9703, 2.3474, 2.7184, 2.8067, 4.6260, 3.5732], device='cuda:5'), covar=tensor([0.3200, 0.0500, 0.2016, 0.3206, 0.2740, 0.2007, 0.0425, 0.1319], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0269, 0.0306, 0.0319, 0.0295, 0.0269, 0.0297, 0.0341], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 11:30:51,293 INFO [train.py:904] (5/8) Epoch 27, batch 9800, loss[loss=0.1826, simple_loss=0.2885, pruned_loss=0.03837, over 15342.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2587, pruned_loss=0.03372, over 3091502.07 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:32:36,072 INFO [train.py:904] (5/8) Epoch 27, batch 9850, loss[loss=0.1622, simple_loss=0.263, pruned_loss=0.03069, over 16918.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2597, pruned_loss=0.03359, over 3083338.12 frames. ], batch size: 116, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:33:00,294 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:33:11,084 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 2.029e+02 2.351e+02 3.028e+02 6.435e+02, threshold=4.703e+02, percent-clipped=1.0 2023-05-02 11:34:02,564 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273791.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:34:25,790 INFO [train.py:904] (5/8) Epoch 27, batch 9900, loss[loss=0.1482, simple_loss=0.2539, pruned_loss=0.02124, over 16754.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2594, pruned_loss=0.03331, over 3060452.39 frames. ], batch size: 83, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:35:26,984 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2382, 2.4092, 2.0823, 2.2268, 2.7307, 2.4195, 2.6353, 2.9069], device='cuda:5'), covar=tensor([0.0185, 0.0459, 0.0570, 0.0539, 0.0347, 0.0433, 0.0272, 0.0284], device='cuda:5'), in_proj_covar=tensor([0.0214, 0.0235, 0.0224, 0.0225, 0.0236, 0.0233, 0.0229, 0.0230], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:35:39,293 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8724, 2.5040, 2.3078, 3.5861, 2.0170, 3.6051, 1.5777, 2.7255], device='cuda:5'), covar=tensor([0.1306, 0.0817, 0.1268, 0.0180, 0.0111, 0.0478, 0.1669, 0.0828], device='cuda:5'), in_proj_covar=tensor([0.0169, 0.0175, 0.0196, 0.0193, 0.0199, 0.0212, 0.0206, 0.0194], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:36:22,579 INFO [train.py:904] (5/8) Epoch 27, batch 9950, loss[loss=0.1564, simple_loss=0.2553, pruned_loss=0.02873, over 16603.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2613, pruned_loss=0.0334, over 3070674.87 frames. ], batch size: 62, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:37:03,261 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 1.975e+02 2.538e+02 3.032e+02 5.048e+02, threshold=5.077e+02, percent-clipped=2.0 2023-05-02 11:38:25,120 INFO [train.py:904] (5/8) Epoch 27, batch 10000, loss[loss=0.1474, simple_loss=0.241, pruned_loss=0.02688, over 17211.00 frames. ], tot_loss[loss=0.163, simple_loss=0.26, pruned_loss=0.03301, over 3081487.30 frames. ], batch size: 45, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:39:09,154 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9284, 5.2660, 5.3807, 5.1725, 5.2607, 5.7533, 5.2403, 4.9493], device='cuda:5'), covar=tensor([0.0956, 0.1583, 0.1892, 0.1929, 0.2063, 0.0801, 0.1437, 0.2164], device='cuda:5'), in_proj_covar=tensor([0.0403, 0.0602, 0.0660, 0.0489, 0.0652, 0.0693, 0.0521, 0.0650], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 11:40:06,639 INFO [train.py:904] (5/8) Epoch 27, batch 10050, loss[loss=0.1825, simple_loss=0.2752, pruned_loss=0.04488, over 16980.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2607, pruned_loss=0.03352, over 3077929.25 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:40:39,179 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.137e+02 2.431e+02 3.038e+02 5.006e+02, threshold=4.862e+02, percent-clipped=0.0 2023-05-02 11:40:51,985 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-02 11:41:12,389 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2835, 3.3577, 2.0700, 3.6577, 2.5282, 3.6208, 2.2636, 2.7275], device='cuda:5'), covar=tensor([0.0363, 0.0421, 0.1719, 0.0228, 0.0898, 0.0687, 0.1586, 0.0836], device='cuda:5'), in_proj_covar=tensor([0.0168, 0.0172, 0.0189, 0.0162, 0.0172, 0.0208, 0.0198, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:41:30,585 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-05-02 11:41:41,060 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274001.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:41:43,729 INFO [train.py:904] (5/8) Epoch 27, batch 10100, loss[loss=0.1479, simple_loss=0.2452, pruned_loss=0.02533, over 16660.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2608, pruned_loss=0.0338, over 3077316.62 frames. ], batch size: 76, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:43:27,168 INFO [train.py:904] (5/8) Epoch 28, batch 0, loss[loss=0.2142, simple_loss=0.2833, pruned_loss=0.0726, over 16839.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2833, pruned_loss=0.0726, over 16839.00 frames. ], batch size: 116, lr: 2.42e-03, grad_scale: 8.0 2023-05-02 11:43:27,168 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 11:43:34,597 INFO [train.py:938] (5/8) Epoch 28, validation: loss=0.1434, simple_loss=0.2464, pruned_loss=0.02022, over 944034.00 frames. 2023-05-02 11:43:34,598 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 11:43:48,086 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274062.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:43:52,189 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:44:01,205 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.305e+02 2.657e+02 3.293e+02 5.505e+02, threshold=5.314e+02, percent-clipped=2.0 2023-05-02 11:44:13,587 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9872, 2.9788, 2.6719, 4.6554, 3.6840, 4.2477, 1.8004, 3.1584], device='cuda:5'), covar=tensor([0.1410, 0.0718, 0.1292, 0.0176, 0.0221, 0.0448, 0.1651, 0.0837], device='cuda:5'), in_proj_covar=tensor([0.0170, 0.0176, 0.0197, 0.0194, 0.0199, 0.0213, 0.0206, 0.0195], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:44:27,522 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:44:43,866 INFO [train.py:904] (5/8) Epoch 28, batch 50, loss[loss=0.1462, simple_loss=0.2381, pruned_loss=0.02721, over 16858.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2628, pruned_loss=0.04374, over 745642.86 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:44:59,072 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274113.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:24,206 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:34,072 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274139.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:54,005 INFO [train.py:904] (5/8) Epoch 28, batch 100, loss[loss=0.1891, simple_loss=0.2629, pruned_loss=0.05767, over 16748.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2632, pruned_loss=0.04312, over 1318724.49 frames. ], batch size: 124, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:46:20,297 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.302e+02 2.829e+02 3.561e+02 6.259e+02, threshold=5.657e+02, percent-clipped=1.0 2023-05-02 11:46:23,035 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5135, 4.5308, 4.8759, 4.8700, 4.9368, 4.5699, 4.5420, 4.4529], device='cuda:5'), covar=tensor([0.0431, 0.0721, 0.0443, 0.0434, 0.0539, 0.0519, 0.1124, 0.0684], device='cuda:5'), in_proj_covar=tensor([0.0418, 0.0468, 0.0456, 0.0419, 0.0502, 0.0481, 0.0549, 0.0385], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 11:46:27,184 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6166, 3.3647, 3.0354, 5.2802, 4.2921, 4.4828, 2.3965, 3.3926], device='cuda:5'), covar=tensor([0.1057, 0.0706, 0.1117, 0.0187, 0.0245, 0.0514, 0.1306, 0.0773], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0177, 0.0199, 0.0196, 0.0201, 0.0215, 0.0208, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:46:48,731 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:47:00,154 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0809, 4.8212, 5.0753, 5.2709, 5.4708, 4.8010, 5.4302, 5.4641], device='cuda:5'), covar=tensor([0.1966, 0.1386, 0.1836, 0.0832, 0.0568, 0.0895, 0.0561, 0.0624], device='cuda:5'), in_proj_covar=tensor([0.0642, 0.0784, 0.0906, 0.0798, 0.0611, 0.0635, 0.0668, 0.0778], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:47:02,022 INFO [train.py:904] (5/8) Epoch 28, batch 150, loss[loss=0.1531, simple_loss=0.2562, pruned_loss=0.02503, over 17017.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2619, pruned_loss=0.04264, over 1756038.27 frames. ], batch size: 50, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:47:17,249 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 11:47:29,075 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8280, 4.4860, 4.7555, 5.0493, 5.1605, 4.6290, 5.2317, 5.2041], device='cuda:5'), covar=tensor([0.2062, 0.1605, 0.2283, 0.0949, 0.0854, 0.1125, 0.0893, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0642, 0.0783, 0.0906, 0.0797, 0.0610, 0.0635, 0.0667, 0.0777], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:48:08,599 INFO [train.py:904] (5/8) Epoch 28, batch 200, loss[loss=0.1988, simple_loss=0.2662, pruned_loss=0.06574, over 16799.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2626, pruned_loss=0.04313, over 2102236.27 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:48:30,263 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:48:33,860 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.228e+02 2.594e+02 3.067e+02 5.865e+02, threshold=5.188e+02, percent-clipped=1.0 2023-05-02 11:48:51,080 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 11:49:11,539 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3596, 2.3453, 2.3672, 4.2240, 2.2755, 2.6996, 2.4344, 2.4863], device='cuda:5'), covar=tensor([0.1383, 0.3887, 0.3331, 0.0573, 0.4371, 0.2776, 0.3854, 0.3927], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0468, 0.0382, 0.0331, 0.0442, 0.0532, 0.0440, 0.0544], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:49:16,140 INFO [train.py:904] (5/8) Epoch 28, batch 250, loss[loss=0.1633, simple_loss=0.2421, pruned_loss=0.04226, over 16327.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.26, pruned_loss=0.04247, over 2375676.78 frames. ], batch size: 165, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:49:16,595 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274303.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:49:52,001 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274330.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:05,635 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9713, 3.0700, 3.3357, 2.2033, 2.8758, 2.2838, 3.5355, 3.4898], device='cuda:5'), covar=tensor([0.0251, 0.1027, 0.0632, 0.2075, 0.0925, 0.1082, 0.0523, 0.0881], device='cuda:5'), in_proj_covar=tensor([0.0158, 0.0166, 0.0167, 0.0155, 0.0146, 0.0131, 0.0143, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 11:50:19,911 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2969, 5.9208, 6.0037, 5.7347, 5.8183, 6.3528, 5.8764, 5.5614], device='cuda:5'), covar=tensor([0.0766, 0.1661, 0.2230, 0.1988, 0.2521, 0.0931, 0.1463, 0.2162], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0619, 0.0680, 0.0502, 0.0671, 0.0711, 0.0533, 0.0668], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 11:50:23,021 INFO [train.py:904] (5/8) Epoch 28, batch 300, loss[loss=0.1562, simple_loss=0.2351, pruned_loss=0.03864, over 16404.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2565, pruned_loss=0.04113, over 2582005.50 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:50:30,579 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:39,370 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:49,960 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.173e+02 2.525e+02 3.031e+02 5.896e+02, threshold=5.050e+02, percent-clipped=1.0 2023-05-02 11:51:31,009 INFO [train.py:904] (5/8) Epoch 28, batch 350, loss[loss=0.1495, simple_loss=0.2351, pruned_loss=0.03196, over 12264.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2546, pruned_loss=0.04036, over 2746697.39 frames. ], batch size: 246, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:52:37,191 INFO [train.py:904] (5/8) Epoch 28, batch 400, loss[loss=0.1759, simple_loss=0.2702, pruned_loss=0.04081, over 17025.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2531, pruned_loss=0.03956, over 2873507.16 frames. ], batch size: 55, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:52:46,834 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 11:53:03,897 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.265e+02 2.649e+02 3.069e+02 5.270e+02, threshold=5.298e+02, percent-clipped=3.0 2023-05-02 11:53:04,364 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9631, 5.0386, 5.4395, 5.4036, 5.4608, 5.1042, 5.0315, 4.8264], device='cuda:5'), covar=tensor([0.0429, 0.0647, 0.0513, 0.0534, 0.0524, 0.0498, 0.1017, 0.0541], device='cuda:5'), in_proj_covar=tensor([0.0429, 0.0480, 0.0466, 0.0429, 0.0513, 0.0492, 0.0562, 0.0395], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 11:53:24,165 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:53:44,163 INFO [train.py:904] (5/8) Epoch 28, batch 450, loss[loss=0.1534, simple_loss=0.2334, pruned_loss=0.03667, over 15893.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.251, pruned_loss=0.03863, over 2971809.24 frames. ], batch size: 35, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:54:35,125 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:54:53,433 INFO [train.py:904] (5/8) Epoch 28, batch 500, loss[loss=0.1812, simple_loss=0.2731, pruned_loss=0.04465, over 17030.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2497, pruned_loss=0.0376, over 3037956.56 frames. ], batch size: 55, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:55:21,739 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 2.207e+02 2.697e+02 3.239e+02 1.435e+03, threshold=5.394e+02, percent-clipped=3.0 2023-05-02 11:55:58,852 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274601.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:56:00,845 INFO [train.py:904] (5/8) Epoch 28, batch 550, loss[loss=0.1504, simple_loss=0.2388, pruned_loss=0.031, over 16846.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.249, pruned_loss=0.03761, over 3099378.85 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:56:31,530 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274625.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:02,643 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 11:57:09,865 INFO [train.py:904] (5/8) Epoch 28, batch 600, loss[loss=0.1554, simple_loss=0.249, pruned_loss=0.03084, over 17103.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2479, pruned_loss=0.03802, over 3153320.41 frames. ], batch size: 49, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:57:15,287 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:17,521 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:36,637 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.130e+02 2.496e+02 2.954e+02 2.616e+03, threshold=4.992e+02, percent-clipped=3.0 2023-05-02 11:58:16,074 INFO [train.py:904] (5/8) Epoch 28, batch 650, loss[loss=0.1474, simple_loss=0.2299, pruned_loss=0.03245, over 12455.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2467, pruned_loss=0.03755, over 3191925.19 frames. ], batch size: 246, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:58:18,639 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:58:42,095 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5768, 4.5469, 4.4494, 3.9119, 4.5116, 1.7929, 4.2361, 4.0449], device='cuda:5'), covar=tensor([0.0169, 0.0143, 0.0226, 0.0333, 0.0133, 0.3039, 0.0177, 0.0281], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0170, 0.0207, 0.0179, 0.0184, 0.0214, 0.0196, 0.0175], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 11:59:25,788 INFO [train.py:904] (5/8) Epoch 28, batch 700, loss[loss=0.1564, simple_loss=0.2446, pruned_loss=0.03413, over 17228.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2463, pruned_loss=0.0372, over 3223263.36 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:59:54,346 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.355e+02 2.008e+02 2.338e+02 2.859e+02 5.050e+02, threshold=4.676e+02, percent-clipped=1.0 2023-05-02 12:00:13,258 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:00:34,720 INFO [train.py:904] (5/8) Epoch 28, batch 750, loss[loss=0.1972, simple_loss=0.2705, pruned_loss=0.06191, over 16211.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2463, pruned_loss=0.03703, over 3247242.55 frames. ], batch size: 164, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 12:01:00,201 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.44 vs. limit=2.0 2023-05-02 12:01:19,980 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:01:45,410 INFO [train.py:904] (5/8) Epoch 28, batch 800, loss[loss=0.1679, simple_loss=0.2413, pruned_loss=0.04722, over 16889.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2463, pruned_loss=0.03709, over 3253374.19 frames. ], batch size: 109, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:02:06,124 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 12:02:15,003 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.216e+02 2.000e+02 2.423e+02 2.829e+02 5.216e+02, threshold=4.846e+02, percent-clipped=1.0 2023-05-02 12:02:43,474 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:02:46,870 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274896.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:02:57,024 INFO [train.py:904] (5/8) Epoch 28, batch 850, loss[loss=0.1456, simple_loss=0.2301, pruned_loss=0.03055, over 16782.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2464, pruned_loss=0.03705, over 3265525.95 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:03:26,705 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274925.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:05,754 INFO [train.py:904] (5/8) Epoch 28, batch 900, loss[loss=0.1594, simple_loss=0.2582, pruned_loss=0.03027, over 17039.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2456, pruned_loss=0.03638, over 3274426.19 frames. ], batch size: 53, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:04:09,190 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274955.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:14,400 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274959.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:33,655 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:34,540 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.476e+02 2.075e+02 2.378e+02 2.870e+02 6.521e+02, threshold=4.756e+02, percent-clipped=4.0 2023-05-02 12:05:14,547 INFO [train.py:904] (5/8) Epoch 28, batch 950, loss[loss=0.151, simple_loss=0.2281, pruned_loss=0.03697, over 16865.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2456, pruned_loss=0.03644, over 3287776.69 frames. ], batch size: 96, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:05:20,984 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275007.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:06:23,174 INFO [train.py:904] (5/8) Epoch 28, batch 1000, loss[loss=0.1715, simple_loss=0.2456, pruned_loss=0.04867, over 16865.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2448, pruned_loss=0.03637, over 3293834.34 frames. ], batch size: 109, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:06:52,348 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.067e+02 2.369e+02 2.767e+02 9.232e+02, threshold=4.738e+02, percent-clipped=2.0 2023-05-02 12:07:31,788 INFO [train.py:904] (5/8) Epoch 28, batch 1050, loss[loss=0.1648, simple_loss=0.2594, pruned_loss=0.03507, over 17060.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2442, pruned_loss=0.03672, over 3298170.64 frames. ], batch size: 50, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:12,816 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-05-02 12:08:36,641 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0235, 2.2296, 2.4534, 3.7042, 2.2640, 2.5041, 2.3471, 2.4010], device='cuda:5'), covar=tensor([0.1716, 0.3859, 0.3162, 0.0778, 0.3942, 0.2724, 0.4024, 0.3233], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0474, 0.0387, 0.0336, 0.0446, 0.0541, 0.0445, 0.0553], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:08:39,633 INFO [train.py:904] (5/8) Epoch 28, batch 1100, loss[loss=0.1417, simple_loss=0.2215, pruned_loss=0.03098, over 16994.00 frames. ], tot_loss[loss=0.1581, simple_loss=0.2436, pruned_loss=0.03628, over 3309169.95 frames. ], batch size: 41, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:47,420 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:07,969 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 1.984e+02 2.393e+02 2.792e+02 5.361e+02, threshold=4.787e+02, percent-clipped=2.0 2023-05-02 12:09:39,585 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:49,918 INFO [train.py:904] (5/8) Epoch 28, batch 1150, loss[loss=0.1412, simple_loss=0.2178, pruned_loss=0.03233, over 16599.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.2432, pruned_loss=0.03582, over 3317897.40 frames. ], batch size: 68, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:10:12,523 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:44,955 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:53,446 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275250.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:57,378 INFO [train.py:904] (5/8) Epoch 28, batch 1200, loss[loss=0.1486, simple_loss=0.2396, pruned_loss=0.0288, over 17248.00 frames. ], tot_loss[loss=0.1563, simple_loss=0.2421, pruned_loss=0.03521, over 3316373.58 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:10:57,825 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9311, 4.8932, 4.7698, 4.2723, 4.8533, 2.0238, 4.6203, 4.4809], device='cuda:5'), covar=tensor([0.0164, 0.0154, 0.0215, 0.0368, 0.0114, 0.2785, 0.0171, 0.0224], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0174, 0.0211, 0.0183, 0.0188, 0.0218, 0.0200, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:11:26,949 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.036e+02 2.270e+02 2.715e+02 4.480e+02, threshold=4.540e+02, percent-clipped=0.0 2023-05-02 12:12:06,866 INFO [train.py:904] (5/8) Epoch 28, batch 1250, loss[loss=0.1892, simple_loss=0.2605, pruned_loss=0.05893, over 16873.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.243, pruned_loss=0.03592, over 3320925.97 frames. ], batch size: 116, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:12:12,705 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:13:09,216 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275347.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:13:17,200 INFO [train.py:904] (5/8) Epoch 28, batch 1300, loss[loss=0.1637, simple_loss=0.247, pruned_loss=0.0402, over 16422.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.2432, pruned_loss=0.03579, over 3325106.42 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:13:39,120 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275368.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:13:46,874 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 1.989e+02 2.411e+02 2.818e+02 5.022e+02, threshold=4.822e+02, percent-clipped=1.0 2023-05-02 12:14:02,853 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275385.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:14:26,879 INFO [train.py:904] (5/8) Epoch 28, batch 1350, loss[loss=0.1697, simple_loss=0.2522, pruned_loss=0.04364, over 16494.00 frames. ], tot_loss[loss=0.1579, simple_loss=0.2436, pruned_loss=0.03605, over 3319964.84 frames. ], batch size: 68, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:14:35,376 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275408.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:15:28,664 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:15:37,721 INFO [train.py:904] (5/8) Epoch 28, batch 1400, loss[loss=0.1833, simple_loss=0.2522, pruned_loss=0.05723, over 16465.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2432, pruned_loss=0.03589, over 3310679.69 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:16:00,806 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0038, 4.3617, 4.4263, 3.1514, 3.5939, 4.3266, 4.0091, 2.8228], device='cuda:5'), covar=tensor([0.0470, 0.0079, 0.0053, 0.0392, 0.0160, 0.0110, 0.0089, 0.0436], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0090, 0.0091, 0.0136, 0.0102, 0.0115, 0.0099, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 12:16:06,678 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.102e+02 2.484e+02 3.057e+02 6.401e+02, threshold=4.968e+02, percent-clipped=1.0 2023-05-02 12:16:46,822 INFO [train.py:904] (5/8) Epoch 28, batch 1450, loss[loss=0.1547, simple_loss=0.2533, pruned_loss=0.02808, over 17132.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2433, pruned_loss=0.03596, over 3297845.09 frames. ], batch size: 47, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:17:03,288 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:15,498 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-05-02 12:17:32,656 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8514, 4.8186, 4.7749, 4.4199, 4.4247, 4.8218, 4.6539, 4.5486], device='cuda:5'), covar=tensor([0.0605, 0.0798, 0.0331, 0.0340, 0.0972, 0.0535, 0.0423, 0.0717], device='cuda:5'), in_proj_covar=tensor([0.0317, 0.0474, 0.0369, 0.0372, 0.0367, 0.0426, 0.0254, 0.0442], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:17:44,504 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 12:17:53,976 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:57,226 INFO [train.py:904] (5/8) Epoch 28, batch 1500, loss[loss=0.1269, simple_loss=0.2085, pruned_loss=0.02259, over 16768.00 frames. ], tot_loss[loss=0.1577, simple_loss=0.243, pruned_loss=0.03621, over 3295435.62 frames. ], batch size: 39, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:18:26,337 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.124e+02 2.433e+02 3.002e+02 9.704e+02, threshold=4.866e+02, percent-clipped=1.0 2023-05-02 12:18:39,779 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8068, 4.7191, 4.7416, 4.4414, 4.4372, 4.7536, 4.5783, 4.5465], device='cuda:5'), covar=tensor([0.0706, 0.1150, 0.0371, 0.0339, 0.0953, 0.0759, 0.0498, 0.0717], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0477, 0.0371, 0.0373, 0.0369, 0.0429, 0.0256, 0.0445], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:18:50,459 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8699, 4.7547, 4.7744, 4.4276, 4.4487, 4.8022, 4.6419, 4.5742], device='cuda:5'), covar=tensor([0.0704, 0.1087, 0.0385, 0.0362, 0.1063, 0.0741, 0.0508, 0.0801], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0477, 0.0371, 0.0373, 0.0368, 0.0428, 0.0255, 0.0445], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:19:01,050 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275598.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:19:07,374 INFO [train.py:904] (5/8) Epoch 28, batch 1550, loss[loss=0.1711, simple_loss=0.2698, pruned_loss=0.03623, over 16697.00 frames. ], tot_loss[loss=0.1589, simple_loss=0.2441, pruned_loss=0.03685, over 3305722.64 frames. ], batch size: 57, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:19:41,689 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7463, 4.7208, 4.5963, 3.8438, 4.6864, 1.7194, 4.4060, 4.2670], device='cuda:5'), covar=tensor([0.0181, 0.0164, 0.0245, 0.0518, 0.0147, 0.3295, 0.0228, 0.0331], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0176, 0.0212, 0.0185, 0.0189, 0.0219, 0.0202, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:20:18,415 INFO [train.py:904] (5/8) Epoch 28, batch 1600, loss[loss=0.1583, simple_loss=0.2594, pruned_loss=0.0286, over 17258.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2461, pruned_loss=0.03745, over 3299817.88 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:20:30,682 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 12:20:33,177 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275663.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:20:48,277 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.228e+02 2.653e+02 3.158e+02 7.607e+02, threshold=5.306e+02, percent-clipped=1.0 2023-05-02 12:21:09,314 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 12:21:28,987 INFO [train.py:904] (5/8) Epoch 28, batch 1650, loss[loss=0.1801, simple_loss=0.2534, pruned_loss=0.05343, over 16462.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.248, pruned_loss=0.03836, over 3300441.75 frames. ], batch size: 75, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:21:29,333 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275703.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:22:12,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5620, 3.6355, 2.3665, 3.9076, 2.9704, 3.8683, 2.4663, 2.9707], device='cuda:5'), covar=tensor([0.0316, 0.0488, 0.1539, 0.0371, 0.0752, 0.0851, 0.1378, 0.0738], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0183, 0.0198, 0.0175, 0.0181, 0.0222, 0.0206, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:22:21,217 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275741.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:22:37,931 INFO [train.py:904] (5/8) Epoch 28, batch 1700, loss[loss=0.1329, simple_loss=0.2219, pruned_loss=0.02195, over 17019.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2493, pruned_loss=0.03809, over 3315006.45 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:23:08,714 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.133e+02 2.503e+02 2.990e+02 4.977e+02, threshold=5.005e+02, percent-clipped=0.0 2023-05-02 12:23:32,917 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 12:23:49,178 INFO [train.py:904] (5/8) Epoch 28, batch 1750, loss[loss=0.1559, simple_loss=0.2431, pruned_loss=0.03428, over 16515.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2503, pruned_loss=0.03829, over 3314265.13 frames. ], batch size: 75, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:23:50,784 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0015, 4.5848, 4.5609, 3.2511, 3.7763, 4.5398, 4.0683, 2.6858], device='cuda:5'), covar=tensor([0.0537, 0.0072, 0.0052, 0.0402, 0.0168, 0.0099, 0.0101, 0.0523], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0092, 0.0092, 0.0138, 0.0104, 0.0116, 0.0100, 0.0134], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 12:24:05,556 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:24:16,949 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 12:24:59,141 INFO [train.py:904] (5/8) Epoch 28, batch 1800, loss[loss=0.1367, simple_loss=0.2235, pruned_loss=0.02502, over 17261.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2507, pruned_loss=0.03804, over 3321179.95 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:25:09,011 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275860.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:11,868 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275862.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:30,538 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.248e+02 2.316e+02 2.618e+02 3.157e+02 1.143e+03, threshold=5.235e+02, percent-clipped=6.0 2023-05-02 12:25:37,238 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3572, 5.9241, 6.0605, 5.7647, 5.9089, 6.3723, 5.8602, 5.5728], device='cuda:5'), covar=tensor([0.0851, 0.2064, 0.2636, 0.2038, 0.2403, 0.0936, 0.1569, 0.2410], device='cuda:5'), in_proj_covar=tensor([0.0436, 0.0652, 0.0716, 0.0533, 0.0705, 0.0744, 0.0558, 0.0704], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:25:55,245 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1395, 3.7602, 4.3662, 2.2223, 4.4990, 4.6203, 3.4288, 3.5767], device='cuda:5'), covar=tensor([0.0692, 0.0320, 0.0213, 0.1169, 0.0081, 0.0194, 0.0406, 0.0398], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0110, 0.0102, 0.0139, 0.0086, 0.0132, 0.0130, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:26:08,322 INFO [train.py:904] (5/8) Epoch 28, batch 1850, loss[loss=0.1678, simple_loss=0.2661, pruned_loss=0.03473, over 17158.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2513, pruned_loss=0.03795, over 3324607.15 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:26:33,786 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275921.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:16,692 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275952.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:17,459 INFO [train.py:904] (5/8) Epoch 28, batch 1900, loss[loss=0.1697, simple_loss=0.2504, pruned_loss=0.04446, over 16872.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2509, pruned_loss=0.03771, over 3323424.63 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:27:32,235 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275963.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:47,687 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 2.028e+02 2.382e+02 2.719e+02 5.605e+02, threshold=4.765e+02, percent-clipped=1.0 2023-05-02 12:28:30,507 INFO [train.py:904] (5/8) Epoch 28, batch 1950, loss[loss=0.1868, simple_loss=0.2712, pruned_loss=0.0512, over 15590.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2506, pruned_loss=0.03727, over 3326145.84 frames. ], batch size: 191, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:28:30,936 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276003.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:28:40,999 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:28:44,309 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:28:46,530 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7933, 4.5789, 4.8729, 5.0468, 5.2108, 4.6010, 5.2466, 5.2176], device='cuda:5'), covar=tensor([0.2225, 0.1424, 0.1843, 0.0794, 0.0601, 0.1159, 0.0718, 0.0725], device='cuda:5'), in_proj_covar=tensor([0.0695, 0.0845, 0.0983, 0.0861, 0.0654, 0.0687, 0.0720, 0.0834], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:29:02,830 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8563, 2.9370, 3.2559, 2.1302, 2.9009, 2.1111, 3.4479, 3.3738], device='cuda:5'), covar=tensor([0.0267, 0.1055, 0.0696, 0.2089, 0.0922, 0.1169, 0.0573, 0.0918], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0172, 0.0170, 0.0158, 0.0149, 0.0134, 0.0147, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 12:29:15,281 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 12:29:24,727 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:29:37,516 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276051.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:29:40,574 INFO [train.py:904] (5/8) Epoch 28, batch 2000, loss[loss=0.144, simple_loss=0.2272, pruned_loss=0.03041, over 15912.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2502, pruned_loss=0.03746, over 3318618.15 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:30:06,020 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0076, 2.9949, 2.8064, 4.6267, 3.7153, 4.2765, 1.8966, 3.1561], device='cuda:5'), covar=tensor([0.1314, 0.0723, 0.1129, 0.0208, 0.0207, 0.0414, 0.1511, 0.0798], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0180, 0.0200, 0.0203, 0.0206, 0.0219, 0.0210, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:30:11,367 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.104e+02 2.511e+02 2.919e+02 7.567e+02, threshold=5.023e+02, percent-clipped=3.0 2023-05-02 12:30:21,155 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7975, 3.9820, 2.6806, 4.5529, 3.1619, 4.4690, 2.9566, 3.3583], device='cuda:5'), covar=tensor([0.0376, 0.0451, 0.1587, 0.0430, 0.0834, 0.0663, 0.1347, 0.0838], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0184, 0.0198, 0.0176, 0.0181, 0.0223, 0.0207, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:30:31,277 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276089.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:30:50,236 INFO [train.py:904] (5/8) Epoch 28, batch 2050, loss[loss=0.1484, simple_loss=0.2317, pruned_loss=0.03251, over 16816.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2503, pruned_loss=0.03742, over 3323889.17 frames. ], batch size: 96, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:30:53,253 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4465, 4.1566, 4.5908, 2.7929, 4.7374, 4.8572, 3.7164, 3.8787], device='cuda:5'), covar=tensor([0.0662, 0.0270, 0.0177, 0.1061, 0.0084, 0.0188, 0.0373, 0.0389], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0110, 0.0102, 0.0139, 0.0086, 0.0131, 0.0130, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:32:00,358 INFO [train.py:904] (5/8) Epoch 28, batch 2100, loss[loss=0.1786, simple_loss=0.2639, pruned_loss=0.04668, over 16571.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2506, pruned_loss=0.03792, over 3321804.98 frames. ], batch size: 75, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:32:30,700 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.091e+02 2.490e+02 2.967e+02 8.308e+02, threshold=4.979e+02, percent-clipped=2.0 2023-05-02 12:33:09,365 INFO [train.py:904] (5/8) Epoch 28, batch 2150, loss[loss=0.1709, simple_loss=0.2604, pruned_loss=0.04068, over 17030.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.252, pruned_loss=0.03855, over 3321247.42 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:33:27,994 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276216.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:33:32,190 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8639, 3.7862, 3.9265, 4.0499, 4.0902, 3.7003, 3.9399, 4.1211], device='cuda:5'), covar=tensor([0.1651, 0.1092, 0.1200, 0.0674, 0.0632, 0.2034, 0.2163, 0.0796], device='cuda:5'), in_proj_covar=tensor([0.0699, 0.0851, 0.0991, 0.0866, 0.0658, 0.0691, 0.0725, 0.0838], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:33:34,926 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 12:34:18,069 INFO [train.py:904] (5/8) Epoch 28, batch 2200, loss[loss=0.1883, simple_loss=0.2582, pruned_loss=0.05923, over 16765.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2536, pruned_loss=0.03983, over 3317426.75 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:34:50,518 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.258e+02 2.667e+02 3.359e+02 8.724e+02, threshold=5.333e+02, percent-clipped=6.0 2023-05-02 12:35:27,994 INFO [train.py:904] (5/8) Epoch 28, batch 2250, loss[loss=0.1624, simple_loss=0.2426, pruned_loss=0.04114, over 16798.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2542, pruned_loss=0.03994, over 3318185.67 frames. ], batch size: 96, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:35:35,424 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:35:40,719 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5576, 4.6481, 4.8294, 4.6020, 4.6805, 5.2533, 4.7623, 4.4670], device='cuda:5'), covar=tensor([0.1656, 0.2198, 0.2628, 0.2365, 0.2924, 0.1191, 0.1783, 0.2651], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0648, 0.0713, 0.0530, 0.0700, 0.0739, 0.0555, 0.0701], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:35:48,084 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:36:14,992 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8714, 3.0879, 2.7355, 5.0683, 4.1091, 4.3981, 1.9375, 3.1946], device='cuda:5'), covar=tensor([0.1444, 0.0757, 0.1286, 0.0216, 0.0239, 0.0462, 0.1574, 0.0814], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0204, 0.0207, 0.0220, 0.0210, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:36:37,084 INFO [train.py:904] (5/8) Epoch 28, batch 2300, loss[loss=0.143, simple_loss=0.2295, pruned_loss=0.0282, over 16859.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2541, pruned_loss=0.0394, over 3325687.30 frames. ], batch size: 102, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:36:58,939 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 12:37:08,706 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.043e+02 2.409e+02 2.878e+02 4.638e+02, threshold=4.818e+02, percent-clipped=0.0 2023-05-02 12:37:12,924 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276378.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:37:46,521 INFO [train.py:904] (5/8) Epoch 28, batch 2350, loss[loss=0.1594, simple_loss=0.2591, pruned_loss=0.02987, over 17242.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2544, pruned_loss=0.04007, over 3323315.35 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:38:49,601 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8746, 4.4775, 3.0762, 2.4116, 2.7556, 2.6925, 4.8456, 3.6522], device='cuda:5'), covar=tensor([0.2987, 0.0526, 0.1936, 0.3055, 0.3050, 0.2174, 0.0346, 0.1498], device='cuda:5'), in_proj_covar=tensor([0.0336, 0.0278, 0.0315, 0.0329, 0.0307, 0.0279, 0.0307, 0.0355], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:38:54,363 INFO [train.py:904] (5/8) Epoch 28, batch 2400, loss[loss=0.1726, simple_loss=0.2667, pruned_loss=0.03926, over 16842.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2555, pruned_loss=0.03977, over 3333876.20 frames. ], batch size: 96, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:39:11,972 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4170, 3.5033, 3.9904, 2.2937, 3.1400, 2.4994, 3.8459, 3.6927], device='cuda:5'), covar=tensor([0.0277, 0.0981, 0.0478, 0.2053, 0.0867, 0.1019, 0.0602, 0.1081], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0172, 0.0171, 0.0158, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 12:39:26,375 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.076e+02 2.394e+02 2.729e+02 1.080e+03, threshold=4.789e+02, percent-clipped=2.0 2023-05-02 12:39:34,422 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6277, 4.5599, 4.5340, 4.2283, 4.2859, 4.5627, 4.3755, 4.3362], device='cuda:5'), covar=tensor([0.0685, 0.0992, 0.0344, 0.0351, 0.0871, 0.0601, 0.0573, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0322, 0.0482, 0.0375, 0.0378, 0.0371, 0.0432, 0.0257, 0.0450], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:40:04,326 INFO [train.py:904] (5/8) Epoch 28, batch 2450, loss[loss=0.1802, simple_loss=0.2612, pruned_loss=0.04961, over 16763.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2563, pruned_loss=0.03995, over 3333474.15 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:40:23,743 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276516.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:40:24,344 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 12:41:13,994 INFO [train.py:904] (5/8) Epoch 28, batch 2500, loss[loss=0.1551, simple_loss=0.2501, pruned_loss=0.03007, over 17167.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2557, pruned_loss=0.03922, over 3332402.24 frames. ], batch size: 46, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:41:30,331 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276564.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:41:45,572 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.004e+02 2.344e+02 2.885e+02 4.375e+02, threshold=4.688e+02, percent-clipped=0.0 2023-05-02 12:42:22,936 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 12:42:24,148 INFO [train.py:904] (5/8) Epoch 28, batch 2550, loss[loss=0.1465, simple_loss=0.2368, pruned_loss=0.02811, over 17211.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2549, pruned_loss=0.03879, over 3336934.34 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:42:31,813 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:42:36,037 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276611.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:43:02,885 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7168, 2.7596, 2.6544, 4.8343, 3.7858, 4.2664, 1.6213, 3.0902], device='cuda:5'), covar=tensor([0.1490, 0.0888, 0.1258, 0.0250, 0.0255, 0.0411, 0.1738, 0.0801], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0182, 0.0202, 0.0205, 0.0208, 0.0220, 0.0211, 0.0200], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:43:33,181 INFO [train.py:904] (5/8) Epoch 28, batch 2600, loss[loss=0.1403, simple_loss=0.2297, pruned_loss=0.02547, over 16988.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2546, pruned_loss=0.03817, over 3338163.86 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:43:38,042 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276656.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:43:55,783 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:01,119 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276672.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:02,597 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276673.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:05,672 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.267e+02 2.519e+02 3.133e+02 7.416e+02, threshold=5.037e+02, percent-clipped=2.0 2023-05-02 12:44:43,696 INFO [train.py:904] (5/8) Epoch 28, batch 2650, loss[loss=0.1671, simple_loss=0.2567, pruned_loss=0.03878, over 16497.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2561, pruned_loss=0.03868, over 3329593.81 frames. ], batch size: 146, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:45:19,133 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:45:47,926 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9244, 2.7118, 2.8483, 2.1798, 2.6697, 2.1877, 2.7717, 2.8662], device='cuda:5'), covar=tensor([0.0304, 0.0899, 0.0559, 0.1786, 0.0856, 0.0952, 0.0643, 0.0852], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0158, 0.0149, 0.0134, 0.0148, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 12:45:50,740 INFO [train.py:904] (5/8) Epoch 28, batch 2700, loss[loss=0.1541, simple_loss=0.2423, pruned_loss=0.03294, over 17228.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2558, pruned_loss=0.03868, over 3325410.87 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:46:23,528 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 2.127e+02 2.476e+02 3.244e+02 1.015e+03, threshold=4.951e+02, percent-clipped=5.0 2023-05-02 12:46:34,784 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9159, 4.8913, 4.8142, 4.4285, 4.5028, 4.8874, 4.7050, 4.5979], device='cuda:5'), covar=tensor([0.0731, 0.0865, 0.0361, 0.0366, 0.0960, 0.0548, 0.0453, 0.0732], device='cuda:5'), in_proj_covar=tensor([0.0323, 0.0483, 0.0376, 0.0380, 0.0374, 0.0433, 0.0257, 0.0451], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:46:42,564 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:46:47,784 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 12:46:58,831 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2864, 4.5633, 4.8212, 4.7723, 4.8588, 4.5689, 4.2965, 4.4153], device='cuda:5'), covar=tensor([0.0656, 0.0964, 0.0582, 0.0693, 0.0814, 0.0677, 0.1706, 0.0718], device='cuda:5'), in_proj_covar=tensor([0.0446, 0.0504, 0.0484, 0.0449, 0.0534, 0.0513, 0.0590, 0.0410], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 12:47:00,590 INFO [train.py:904] (5/8) Epoch 28, batch 2750, loss[loss=0.1864, simple_loss=0.2747, pruned_loss=0.04907, over 16583.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.256, pruned_loss=0.03823, over 3326980.47 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:07,272 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276851.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:48:09,131 INFO [train.py:904] (5/8) Epoch 28, batch 2800, loss[loss=0.1719, simple_loss=0.2614, pruned_loss=0.04122, over 16502.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2554, pruned_loss=0.03795, over 3336537.79 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:24,970 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276863.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:48:41,264 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.121e+02 2.382e+02 2.775e+02 4.197e+02, threshold=4.764e+02, percent-clipped=0.0 2023-05-02 12:48:56,009 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 12:49:12,328 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5297, 4.5063, 4.6500, 4.4375, 4.4741, 5.1228, 4.5731, 4.2072], device='cuda:5'), covar=tensor([0.1526, 0.2273, 0.2621, 0.2272, 0.2961, 0.1098, 0.1736, 0.2733], device='cuda:5'), in_proj_covar=tensor([0.0434, 0.0653, 0.0716, 0.0531, 0.0705, 0.0742, 0.0557, 0.0706], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:49:20,049 INFO [train.py:904] (5/8) Epoch 28, batch 2850, loss[loss=0.1501, simple_loss=0.2477, pruned_loss=0.02622, over 17049.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2544, pruned_loss=0.03737, over 3338936.81 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:49:25,928 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7001, 6.0235, 5.7592, 5.8842, 5.4897, 5.4693, 5.4041, 6.1368], device='cuda:5'), covar=tensor([0.1456, 0.1048, 0.1140, 0.0908, 0.0892, 0.0712, 0.1264, 0.1040], device='cuda:5'), in_proj_covar=tensor([0.0730, 0.0893, 0.0727, 0.0687, 0.0562, 0.0558, 0.0748, 0.0695], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:49:50,143 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276924.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:49:57,719 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:18,941 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7650, 4.0079, 2.8400, 4.6204, 3.1895, 4.4934, 2.8186, 3.3464], device='cuda:5'), covar=tensor([0.0377, 0.0413, 0.1403, 0.0352, 0.0814, 0.0588, 0.1463, 0.0736], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0185, 0.0199, 0.0177, 0.0183, 0.0226, 0.0208, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:50:27,313 INFO [train.py:904] (5/8) Epoch 28, batch 2900, loss[loss=0.1577, simple_loss=0.2516, pruned_loss=0.03192, over 17026.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.253, pruned_loss=0.03712, over 3342449.63 frames. ], batch size: 55, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:50:46,681 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:54,975 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:58,052 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.124e+02 2.391e+02 2.943e+02 6.733e+02, threshold=4.783e+02, percent-clipped=5.0 2023-05-02 12:51:01,625 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7697, 2.4105, 1.9516, 2.1458, 2.7884, 2.5252, 2.7174, 2.8608], device='cuda:5'), covar=tensor([0.0256, 0.0510, 0.0649, 0.0571, 0.0311, 0.0424, 0.0267, 0.0361], device='cuda:5'), in_proj_covar=tensor([0.0239, 0.0253, 0.0240, 0.0241, 0.0254, 0.0251, 0.0252, 0.0250], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:51:21,969 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276992.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:51:36,314 INFO [train.py:904] (5/8) Epoch 28, batch 2950, loss[loss=0.1941, simple_loss=0.2723, pruned_loss=0.05796, over 16788.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.253, pruned_loss=0.03817, over 3336511.88 frames. ], batch size: 83, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:51:36,893 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2781, 3.4338, 3.1640, 5.1484, 4.3455, 4.4894, 2.2224, 3.6314], device='cuda:5'), covar=tensor([0.1353, 0.0704, 0.1075, 0.0247, 0.0265, 0.0424, 0.1558, 0.0707], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0180, 0.0200, 0.0204, 0.0206, 0.0219, 0.0209, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 12:51:48,615 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:01,178 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:05,344 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:23,117 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4661, 2.6552, 2.2827, 2.3902, 2.9405, 2.6218, 3.0929, 3.1375], device='cuda:5'), covar=tensor([0.0238, 0.0466, 0.0571, 0.0537, 0.0348, 0.0440, 0.0281, 0.0303], device='cuda:5'), in_proj_covar=tensor([0.0239, 0.0253, 0.0240, 0.0241, 0.0254, 0.0251, 0.0252, 0.0250], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:52:45,421 INFO [train.py:904] (5/8) Epoch 28, batch 3000, loss[loss=0.1803, simple_loss=0.2597, pruned_loss=0.05042, over 16868.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2533, pruned_loss=0.03907, over 3324012.05 frames. ], batch size: 116, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:52:45,422 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 12:52:54,789 INFO [train.py:938] (5/8) Epoch 28, validation: loss=0.1335, simple_loss=0.2385, pruned_loss=0.01427, over 944034.00 frames. 2023-05-02 12:52:54,790 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 12:53:10,914 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0629, 4.6210, 4.5560, 3.1558, 3.7474, 4.5275, 4.0373, 2.9056], device='cuda:5'), covar=tensor([0.0497, 0.0070, 0.0055, 0.0433, 0.0154, 0.0108, 0.0101, 0.0474], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0103, 0.0116, 0.0100, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 12:53:21,160 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277072.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:53:25,666 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.148e+02 2.585e+02 3.132e+02 1.139e+03, threshold=5.170e+02, percent-clipped=2.0 2023-05-02 12:54:02,157 INFO [train.py:904] (5/8) Epoch 28, batch 3050, loss[loss=0.1825, simple_loss=0.2684, pruned_loss=0.04824, over 15526.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2532, pruned_loss=0.03908, over 3323391.48 frames. ], batch size: 191, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:54:29,290 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 12:55:01,332 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277146.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:55:11,790 INFO [train.py:904] (5/8) Epoch 28, batch 3100, loss[loss=0.1649, simple_loss=0.241, pruned_loss=0.04442, over 16759.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2526, pruned_loss=0.03942, over 3309232.63 frames. ], batch size: 83, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:55:20,239 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 12:55:34,895 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4272, 2.4361, 2.3475, 4.3328, 2.4083, 2.8399, 2.4596, 2.5958], device='cuda:5'), covar=tensor([0.1445, 0.3846, 0.3322, 0.0549, 0.4212, 0.2561, 0.3736, 0.3843], device='cuda:5'), in_proj_covar=tensor([0.0425, 0.0478, 0.0390, 0.0341, 0.0448, 0.0548, 0.0450, 0.0559], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 12:55:43,286 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 2.064e+02 2.462e+02 2.812e+02 4.738e+02, threshold=4.925e+02, percent-clipped=0.0 2023-05-02 12:55:45,415 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:21,092 INFO [train.py:904] (5/8) Epoch 28, batch 3150, loss[loss=0.1548, simple_loss=0.2329, pruned_loss=0.03837, over 16659.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2523, pruned_loss=0.03896, over 3304977.89 frames. ], batch size: 89, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:56:44,556 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:44,685 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:57:10,193 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277238.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:57:30,730 INFO [train.py:904] (5/8) Epoch 28, batch 3200, loss[loss=0.1597, simple_loss=0.2448, pruned_loss=0.03733, over 16779.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2519, pruned_loss=0.03853, over 3307884.45 frames. ], batch size: 102, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:57:50,789 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277267.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:01,816 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.445e+02 1.963e+02 2.375e+02 2.950e+02 8.294e+02, threshold=4.751e+02, percent-clipped=4.0 2023-05-02 12:58:07,844 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277280.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:17,887 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277287.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:58:19,049 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1922, 5.6481, 5.8187, 5.4577, 5.6218, 6.1680, 5.6446, 5.3706], device='cuda:5'), covar=tensor([0.0907, 0.2090, 0.2390, 0.2172, 0.2554, 0.0970, 0.1502, 0.2342], device='cuda:5'), in_proj_covar=tensor([0.0436, 0.0653, 0.0718, 0.0533, 0.0707, 0.0741, 0.0557, 0.0708], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:58:39,073 INFO [train.py:904] (5/8) Epoch 28, batch 3250, loss[loss=0.1803, simple_loss=0.2873, pruned_loss=0.03668, over 16688.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2523, pruned_loss=0.03834, over 3312273.52 frames. ], batch size: 62, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:58:42,490 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277305.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:55,900 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:03,288 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7860, 4.4702, 3.1137, 2.4276, 2.7267, 2.7048, 4.8454, 3.6202], device='cuda:5'), covar=tensor([0.3263, 0.0538, 0.1935, 0.3122, 0.2883, 0.2158, 0.0357, 0.1536], device='cuda:5'), in_proj_covar=tensor([0.0336, 0.0278, 0.0314, 0.0329, 0.0307, 0.0279, 0.0307, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 12:59:09,120 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:47,716 INFO [train.py:904] (5/8) Epoch 28, batch 3300, loss[loss=0.1967, simple_loss=0.2773, pruned_loss=0.05803, over 16484.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2536, pruned_loss=0.03906, over 3313374.12 frames. ], batch size: 146, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:00:05,296 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:06,308 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277367.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:12,579 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:18,088 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.541e+02 2.125e+02 2.555e+02 3.019e+02 5.327e+02, threshold=5.111e+02, percent-clipped=1.0 2023-05-02 13:00:55,909 INFO [train.py:904] (5/8) Epoch 28, batch 3350, loss[loss=0.1534, simple_loss=0.2493, pruned_loss=0.02878, over 17112.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2544, pruned_loss=0.03925, over 3320425.31 frames. ], batch size: 47, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:01:55,076 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:02:02,118 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7544, 3.8604, 2.5134, 4.1845, 3.0199, 4.1109, 2.6638, 3.1736], device='cuda:5'), covar=tensor([0.0297, 0.0410, 0.1530, 0.0370, 0.0802, 0.0728, 0.1341, 0.0748], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0185, 0.0199, 0.0178, 0.0183, 0.0226, 0.0208, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 13:02:03,952 INFO [train.py:904] (5/8) Epoch 28, batch 3400, loss[loss=0.1332, simple_loss=0.2239, pruned_loss=0.02125, over 17228.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2537, pruned_loss=0.03893, over 3320462.60 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:02:34,601 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.063e+02 2.423e+02 2.740e+02 5.592e+02, threshold=4.846e+02, percent-clipped=2.0 2023-05-02 13:03:01,406 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:13,673 INFO [train.py:904] (5/8) Epoch 28, batch 3450, loss[loss=0.1637, simple_loss=0.2373, pruned_loss=0.0451, over 16738.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2516, pruned_loss=0.03822, over 3329704.68 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:03:36,113 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277519.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:55,472 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:23,515 INFO [train.py:904] (5/8) Epoch 28, batch 3500, loss[loss=0.1551, simple_loss=0.2531, pruned_loss=0.02852, over 17124.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2509, pruned_loss=0.03762, over 3332856.67 frames. ], batch size: 47, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:04:25,796 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:42,186 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:47,752 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:54,562 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277575.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:55,304 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.163e+02 2.511e+02 2.754e+02 5.260e+02, threshold=5.022e+02, percent-clipped=2.0 2023-05-02 13:05:10,792 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277587.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 13:05:32,643 INFO [train.py:904] (5/8) Epoch 28, batch 3550, loss[loss=0.151, simple_loss=0.2439, pruned_loss=0.02907, over 17225.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2506, pruned_loss=0.03773, over 3320165.01 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:05:49,372 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277615.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:05:57,191 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0808, 2.9560, 3.0003, 5.1118, 4.2505, 4.4723, 1.8240, 3.3123], device='cuda:5'), covar=tensor([0.1279, 0.0824, 0.1114, 0.0245, 0.0240, 0.0440, 0.1618, 0.0792], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0182, 0.0201, 0.0206, 0.0208, 0.0221, 0.0210, 0.0200], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 13:06:11,026 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277631.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:11,440 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-05-02 13:06:16,897 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:21,678 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 13:06:42,335 INFO [train.py:904] (5/8) Epoch 28, batch 3600, loss[loss=0.1869, simple_loss=0.2602, pruned_loss=0.05686, over 16335.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2488, pruned_loss=0.03697, over 3323740.77 frames. ], batch size: 145, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:06:53,515 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:02,090 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277667.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:11,700 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2528, 5.2173, 5.1212, 4.6093, 4.7643, 5.1780, 5.0784, 4.7579], device='cuda:5'), covar=tensor([0.0563, 0.0539, 0.0318, 0.0377, 0.1052, 0.0473, 0.0346, 0.0846], device='cuda:5'), in_proj_covar=tensor([0.0327, 0.0492, 0.0382, 0.0385, 0.0380, 0.0441, 0.0260, 0.0458], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 13:07:14,469 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.038e+02 2.278e+02 2.834e+02 5.503e+02, threshold=4.555e+02, percent-clipped=2.0 2023-05-02 13:07:37,008 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:52,073 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 13:07:53,659 INFO [train.py:904] (5/8) Epoch 28, batch 3650, loss[loss=0.1596, simple_loss=0.2369, pruned_loss=0.04114, over 16710.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2475, pruned_loss=0.03772, over 3326916.11 frames. ], batch size: 134, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:08:03,637 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-02 13:08:12,450 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:08:32,600 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 13:09:06,589 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277752.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 13:09:07,240 INFO [train.py:904] (5/8) Epoch 28, batch 3700, loss[loss=0.1978, simple_loss=0.2728, pruned_loss=0.06138, over 11162.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2467, pruned_loss=0.0393, over 3288999.74 frames. ], batch size: 248, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:09:41,446 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.185e+02 2.658e+02 3.107e+02 6.146e+02, threshold=5.316e+02, percent-clipped=3.0 2023-05-02 13:10:03,503 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 13:10:22,510 INFO [train.py:904] (5/8) Epoch 28, batch 3750, loss[loss=0.1804, simple_loss=0.253, pruned_loss=0.05393, over 16798.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2475, pruned_loss=0.04078, over 3274972.63 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:10:30,780 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2414, 3.2920, 3.5747, 2.2943, 3.0417, 2.3840, 3.7712, 3.6889], device='cuda:5'), covar=tensor([0.0218, 0.0919, 0.0638, 0.1995, 0.0861, 0.1028, 0.0455, 0.0812], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0174, 0.0171, 0.0158, 0.0150, 0.0133, 0.0148, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 13:10:50,920 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8094, 3.8956, 4.1312, 4.1094, 4.1368, 3.9053, 3.9367, 3.9124], device='cuda:5'), covar=tensor([0.0436, 0.0702, 0.0453, 0.0455, 0.0555, 0.0535, 0.0819, 0.0593], device='cuda:5'), in_proj_covar=tensor([0.0450, 0.0509, 0.0489, 0.0451, 0.0537, 0.0515, 0.0594, 0.0414], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 13:11:06,984 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:11:33,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6945, 4.7169, 5.0460, 4.9976, 5.0705, 4.7658, 4.7719, 4.6157], device='cuda:5'), covar=tensor([0.0351, 0.0669, 0.0377, 0.0416, 0.0472, 0.0424, 0.0807, 0.0554], device='cuda:5'), in_proj_covar=tensor([0.0449, 0.0509, 0.0488, 0.0450, 0.0536, 0.0515, 0.0593, 0.0413], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 13:11:35,752 INFO [train.py:904] (5/8) Epoch 28, batch 3800, loss[loss=0.1649, simple_loss=0.2418, pruned_loss=0.04405, over 16695.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2483, pruned_loss=0.04145, over 3281635.16 frames. ], batch size: 134, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:08,736 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:11,084 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.049e+02 2.321e+02 2.888e+02 4.993e+02, threshold=4.641e+02, percent-clipped=0.0 2023-05-02 13:12:17,664 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:48,152 INFO [train.py:904] (5/8) Epoch 28, batch 3850, loss[loss=0.1626, simple_loss=0.2458, pruned_loss=0.03971, over 16551.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2484, pruned_loss=0.04204, over 3278102.46 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:59,692 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:18,296 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:23,018 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:00,822 INFO [train.py:904] (5/8) Epoch 28, batch 3900, loss[loss=0.1662, simple_loss=0.2483, pruned_loss=0.04203, over 16446.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2483, pruned_loss=0.0425, over 3280459.33 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:14:08,049 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8643, 2.6816, 2.7092, 5.0600, 3.9560, 4.2273, 1.7555, 3.0736], device='cuda:5'), covar=tensor([0.1289, 0.0896, 0.1298, 0.0124, 0.0367, 0.0390, 0.1622, 0.0926], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0181, 0.0201, 0.0206, 0.0208, 0.0220, 0.0210, 0.0200], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 13:14:13,214 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277961.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:27,646 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:36,147 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.147e+02 2.438e+02 2.765e+02 5.958e+02, threshold=4.876e+02, percent-clipped=2.0 2023-05-02 13:15:15,026 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9359, 3.0976, 3.1995, 2.1566, 2.8317, 2.2071, 3.4953, 3.4771], device='cuda:5'), covar=tensor([0.0285, 0.0928, 0.0665, 0.2002, 0.0930, 0.1096, 0.0520, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0173, 0.0171, 0.0158, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 13:15:16,810 INFO [train.py:904] (5/8) Epoch 28, batch 3950, loss[loss=0.1673, simple_loss=0.2369, pruned_loss=0.04884, over 16867.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2485, pruned_loss=0.0435, over 3276947.79 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:15:25,692 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278009.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:15:45,375 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 13:15:46,344 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 13:15:58,206 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278032.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:16:19,890 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278047.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 13:16:26,877 INFO [train.py:904] (5/8) Epoch 28, batch 4000, loss[loss=0.165, simple_loss=0.2471, pruned_loss=0.04146, over 16816.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2485, pruned_loss=0.04365, over 3287657.24 frames. ], batch size: 90, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:17:01,537 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 1.968e+02 2.308e+02 2.954e+02 7.046e+02, threshold=4.616e+02, percent-clipped=2.0 2023-05-02 13:17:38,533 INFO [train.py:904] (5/8) Epoch 28, batch 4050, loss[loss=0.1522, simple_loss=0.2457, pruned_loss=0.02941, over 16886.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2492, pruned_loss=0.04258, over 3288134.38 frames. ], batch size: 96, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:18:27,806 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7962, 2.5067, 2.6071, 4.6760, 3.4825, 4.0129, 1.6792, 2.9530], device='cuda:5'), covar=tensor([0.1491, 0.1006, 0.1371, 0.0154, 0.0349, 0.0392, 0.1883, 0.0945], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0181, 0.0201, 0.0206, 0.0209, 0.0220, 0.0210, 0.0200], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 13:18:42,067 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 13:18:52,740 INFO [train.py:904] (5/8) Epoch 28, batch 4100, loss[loss=0.1836, simple_loss=0.2637, pruned_loss=0.05169, over 16668.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2512, pruned_loss=0.04258, over 3284408.82 frames. ], batch size: 57, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:19:20,237 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 13:19:29,063 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 1.827e+02 2.078e+02 2.388e+02 3.753e+02, threshold=4.156e+02, percent-clipped=0.0 2023-05-02 13:20:01,070 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1077, 2.4222, 2.5340, 1.9373, 2.7228, 2.7419, 2.3893, 2.4028], device='cuda:5'), covar=tensor([0.0721, 0.0292, 0.0230, 0.0932, 0.0138, 0.0268, 0.0495, 0.0427], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0113, 0.0104, 0.0141, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 13:20:09,088 INFO [train.py:904] (5/8) Epoch 28, batch 4150, loss[loss=0.1842, simple_loss=0.2805, pruned_loss=0.044, over 16904.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2581, pruned_loss=0.04477, over 3262850.11 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:20:19,237 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:20:38,765 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 13:20:42,380 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:20:42,460 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0854, 3.4357, 3.3600, 2.1008, 3.1118, 3.4441, 3.1654, 1.9085], device='cuda:5'), covar=tensor([0.0580, 0.0060, 0.0079, 0.0492, 0.0126, 0.0117, 0.0118, 0.0537], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0134, 0.0102, 0.0114, 0.0098, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 13:21:22,147 INFO [train.py:904] (5/8) Epoch 28, batch 4200, loss[loss=0.2001, simple_loss=0.2949, pruned_loss=0.05262, over 16698.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2644, pruned_loss=0.04648, over 3209996.92 frames. ], batch size: 76, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:21:27,652 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2924, 3.0624, 3.3612, 1.8589, 3.5163, 3.5103, 2.7601, 2.7363], device='cuda:5'), covar=tensor([0.0853, 0.0331, 0.0240, 0.1207, 0.0108, 0.0199, 0.0513, 0.0482], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0112, 0.0104, 0.0140, 0.0088, 0.0133, 0.0131, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 13:21:30,727 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:30,795 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9076, 4.1607, 3.9956, 4.0671, 3.7295, 3.7725, 3.8398, 4.1592], device='cuda:5'), covar=tensor([0.1084, 0.0957, 0.0986, 0.0863, 0.0839, 0.1786, 0.1001, 0.1045], device='cuda:5'), in_proj_covar=tensor([0.0727, 0.0890, 0.0724, 0.0684, 0.0561, 0.0557, 0.0745, 0.0691], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 13:21:54,355 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278274.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:58,174 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.712e+02 2.281e+02 2.720e+02 3.199e+02 7.364e+02, threshold=5.441e+02, percent-clipped=7.0 2023-05-02 13:22:37,399 INFO [train.py:904] (5/8) Epoch 28, batch 4250, loss[loss=0.1765, simple_loss=0.2711, pruned_loss=0.041, over 16963.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2678, pruned_loss=0.04657, over 3177052.92 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:23:07,872 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:08,158 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 13:23:13,494 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:44,049 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278347.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 13:23:52,556 INFO [train.py:904] (5/8) Epoch 28, batch 4300, loss[loss=0.2064, simple_loss=0.292, pruned_loss=0.06038, over 16621.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2691, pruned_loss=0.04577, over 3185730.14 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:24:25,585 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278374.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:24:29,327 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 2.019e+02 2.401e+02 2.918e+02 4.618e+02, threshold=4.803e+02, percent-clipped=0.0 2023-05-02 13:24:40,791 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:24:56,294 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278395.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:25:07,425 INFO [train.py:904] (5/8) Epoch 28, batch 4350, loss[loss=0.1951, simple_loss=0.2871, pruned_loss=0.05155, over 16730.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2728, pruned_loss=0.04698, over 3193881.92 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:25:57,130 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278435.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:26:22,395 INFO [train.py:904] (5/8) Epoch 28, batch 4400, loss[loss=0.1981, simple_loss=0.2929, pruned_loss=0.05164, over 16509.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2749, pruned_loss=0.04835, over 3170593.12 frames. ], batch size: 35, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:26:58,180 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.247e+02 2.621e+02 3.049e+02 4.749e+02, threshold=5.243e+02, percent-clipped=0.0 2023-05-02 13:27:35,967 INFO [train.py:904] (5/8) Epoch 28, batch 4450, loss[loss=0.1813, simple_loss=0.2782, pruned_loss=0.0422, over 16560.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2784, pruned_loss=0.04935, over 3189984.05 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:27:48,032 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:28:49,084 INFO [train.py:904] (5/8) Epoch 28, batch 4500, loss[loss=0.1999, simple_loss=0.2853, pruned_loss=0.05725, over 16432.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2791, pruned_loss=0.05011, over 3200186.50 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:29:17,639 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278572.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:29:23,508 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 1.834e+02 2.148e+02 2.381e+02 3.720e+02, threshold=4.296e+02, percent-clipped=0.0 2023-05-02 13:30:01,298 INFO [train.py:904] (5/8) Epoch 28, batch 4550, loss[loss=0.2018, simple_loss=0.2901, pruned_loss=0.05674, over 17038.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2803, pruned_loss=0.05121, over 3204671.18 frames. ], batch size: 55, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:30:10,389 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-02 13:30:36,960 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278627.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:12,175 INFO [train.py:904] (5/8) Epoch 28, batch 4600, loss[loss=0.208, simple_loss=0.2929, pruned_loss=0.06151, over 16659.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2809, pruned_loss=0.05131, over 3202978.16 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:31:43,899 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278675.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:46,042 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.292e+02 1.855e+02 2.019e+02 2.344e+02 4.106e+02, threshold=4.037e+02, percent-clipped=0.0 2023-05-02 13:31:49,271 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:57,543 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0529, 5.1142, 5.4103, 5.3647, 5.4632, 5.1044, 5.0680, 4.7308], device='cuda:5'), covar=tensor([0.0290, 0.0407, 0.0321, 0.0399, 0.0402, 0.0344, 0.0866, 0.0483], device='cuda:5'), in_proj_covar=tensor([0.0436, 0.0493, 0.0473, 0.0438, 0.0522, 0.0499, 0.0575, 0.0403], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 13:32:09,090 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 13:32:22,026 INFO [train.py:904] (5/8) Epoch 28, batch 4650, loss[loss=0.1792, simple_loss=0.2632, pruned_loss=0.04756, over 16834.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2803, pruned_loss=0.05131, over 3227258.89 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:33:00,844 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278730.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:33:33,716 INFO [train.py:904] (5/8) Epoch 28, batch 4700, loss[loss=0.1945, simple_loss=0.2857, pruned_loss=0.05162, over 16240.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2779, pruned_loss=0.05039, over 3218895.26 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:34:07,040 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 1.813e+02 1.994e+02 2.322e+02 3.636e+02, threshold=3.989e+02, percent-clipped=0.0 2023-05-02 13:34:45,686 INFO [train.py:904] (5/8) Epoch 28, batch 4750, loss[loss=0.1712, simple_loss=0.2611, pruned_loss=0.04062, over 15431.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2738, pruned_loss=0.04815, over 3209181.36 frames. ], batch size: 191, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:35:57,557 INFO [train.py:904] (5/8) Epoch 28, batch 4800, loss[loss=0.1887, simple_loss=0.276, pruned_loss=0.05075, over 12063.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2707, pruned_loss=0.04634, over 3204594.71 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:36:18,872 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:36:32,601 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.390e+02 1.887e+02 2.183e+02 2.468e+02 4.688e+02, threshold=4.367e+02, percent-clipped=1.0 2023-05-02 13:36:53,947 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4022, 4.5822, 4.6749, 4.4974, 4.4930, 5.0314, 4.4950, 4.2040], device='cuda:5'), covar=tensor([0.1383, 0.1742, 0.1914, 0.1946, 0.2437, 0.0907, 0.1562, 0.2574], device='cuda:5'), in_proj_covar=tensor([0.0425, 0.0636, 0.0697, 0.0519, 0.0689, 0.0724, 0.0543, 0.0691], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 13:37:13,015 INFO [train.py:904] (5/8) Epoch 28, batch 4850, loss[loss=0.1943, simple_loss=0.2859, pruned_loss=0.0513, over 16757.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2711, pruned_loss=0.04543, over 3199253.17 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:38:26,587 INFO [train.py:904] (5/8) Epoch 28, batch 4900, loss[loss=0.1858, simple_loss=0.2862, pruned_loss=0.0427, over 16800.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2707, pruned_loss=0.04458, over 3171029.21 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:38:59,320 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 13:39:00,727 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 1.895e+02 2.133e+02 2.360e+02 3.975e+02, threshold=4.266e+02, percent-clipped=0.0 2023-05-02 13:39:04,551 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:39:36,092 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3722, 2.9754, 2.6774, 2.2576, 2.1969, 2.2883, 2.9903, 2.8026], device='cuda:5'), covar=tensor([0.2668, 0.0715, 0.1715, 0.2832, 0.2560, 0.2117, 0.0593, 0.1316], device='cuda:5'), in_proj_covar=tensor([0.0336, 0.0277, 0.0313, 0.0328, 0.0308, 0.0278, 0.0305, 0.0355], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 13:39:39,810 INFO [train.py:904] (5/8) Epoch 28, batch 4950, loss[loss=0.1821, simple_loss=0.276, pruned_loss=0.04417, over 16692.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2705, pruned_loss=0.04426, over 3147463.40 frames. ], batch size: 134, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:40:14,413 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:19,515 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:51,498 INFO [train.py:904] (5/8) Epoch 28, batch 5000, loss[loss=0.1849, simple_loss=0.2801, pruned_loss=0.04482, over 16949.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2718, pruned_loss=0.04435, over 3152790.11 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:41:26,572 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.098e+02 2.415e+02 2.843e+02 4.357e+02, threshold=4.831e+02, percent-clipped=1.0 2023-05-02 13:41:26,965 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0468, 4.1090, 4.2982, 4.2649, 4.3110, 4.0826, 4.0813, 4.0130], device='cuda:5'), covar=tensor([0.0300, 0.0526, 0.0422, 0.0464, 0.0504, 0.0397, 0.0842, 0.0540], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0488, 0.0471, 0.0435, 0.0517, 0.0496, 0.0571, 0.0400], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 13:41:28,686 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279078.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:42:04,201 INFO [train.py:904] (5/8) Epoch 28, batch 5050, loss[loss=0.1696, simple_loss=0.2639, pruned_loss=0.03764, over 16449.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2723, pruned_loss=0.04401, over 3176465.89 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:42:21,475 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 13:42:44,727 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 13:42:45,696 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:43:00,148 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5635, 3.5347, 3.5266, 2.7410, 3.2699, 2.0509, 3.1910, 2.8552], device='cuda:5'), covar=tensor([0.0211, 0.0189, 0.0180, 0.0354, 0.0132, 0.2617, 0.0172, 0.0339], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0175, 0.0212, 0.0186, 0.0189, 0.0217, 0.0201, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 13:43:14,239 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9367, 2.6214, 2.7953, 2.0585, 2.5997, 2.0657, 2.7108, 2.8551], device='cuda:5'), covar=tensor([0.0306, 0.0901, 0.0626, 0.1889, 0.0908, 0.0972, 0.0621, 0.0764], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0148, 0.0132, 0.0146, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 13:43:17,185 INFO [train.py:904] (5/8) Epoch 28, batch 5100, loss[loss=0.167, simple_loss=0.2474, pruned_loss=0.0433, over 16641.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2705, pruned_loss=0.04344, over 3177903.32 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:43:28,483 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2048, 2.3375, 2.3814, 3.8528, 2.2439, 2.7083, 2.4063, 2.5136], device='cuda:5'), covar=tensor([0.1502, 0.3539, 0.3062, 0.0630, 0.4164, 0.2583, 0.3765, 0.3216], device='cuda:5'), in_proj_covar=tensor([0.0424, 0.0479, 0.0387, 0.0340, 0.0448, 0.0547, 0.0449, 0.0557], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 13:43:37,564 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:43:52,647 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.882e+02 2.136e+02 2.468e+02 5.149e+02, threshold=4.272e+02, percent-clipped=1.0 2023-05-02 13:44:15,266 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:44:21,133 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4409, 4.4612, 4.6642, 4.3695, 4.5113, 5.0347, 4.5319, 4.0894], device='cuda:5'), covar=tensor([0.1380, 0.1883, 0.1771, 0.1999, 0.2388, 0.0922, 0.1443, 0.2624], device='cuda:5'), in_proj_covar=tensor([0.0422, 0.0628, 0.0688, 0.0512, 0.0682, 0.0716, 0.0538, 0.0683], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 13:44:30,680 INFO [train.py:904] (5/8) Epoch 28, batch 5150, loss[loss=0.1652, simple_loss=0.262, pruned_loss=0.03425, over 16515.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2698, pruned_loss=0.04234, over 3191938.09 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:44:49,081 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:45:42,237 INFO [train.py:904] (5/8) Epoch 28, batch 5200, loss[loss=0.1485, simple_loss=0.2438, pruned_loss=0.02662, over 16675.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2681, pruned_loss=0.04153, over 3203857.67 frames. ], batch size: 89, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:46:17,352 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.964e+02 2.228e+02 2.622e+02 8.480e+02, threshold=4.456e+02, percent-clipped=3.0 2023-05-02 13:46:53,565 INFO [train.py:904] (5/8) Epoch 28, batch 5250, loss[loss=0.1766, simple_loss=0.2579, pruned_loss=0.04766, over 16917.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2657, pruned_loss=0.04107, over 3209435.92 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:47:03,633 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0046, 4.8900, 4.7264, 3.0561, 4.0331, 4.6863, 3.9814, 2.7546], device='cuda:5'), covar=tensor([0.0550, 0.0031, 0.0044, 0.0451, 0.0109, 0.0113, 0.0136, 0.0487], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0089, 0.0091, 0.0135, 0.0103, 0.0115, 0.0099, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 13:47:16,666 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 13:47:57,908 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0408, 5.0663, 4.9237, 4.4901, 4.5288, 4.9791, 4.8646, 4.7028], device='cuda:5'), covar=tensor([0.0641, 0.0481, 0.0328, 0.0368, 0.1177, 0.0546, 0.0382, 0.0667], device='cuda:5'), in_proj_covar=tensor([0.0312, 0.0470, 0.0365, 0.0368, 0.0364, 0.0422, 0.0249, 0.0435], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 13:48:06,874 INFO [train.py:904] (5/8) Epoch 28, batch 5300, loss[loss=0.1724, simple_loss=0.2705, pruned_loss=0.03716, over 15366.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2624, pruned_loss=0.04013, over 3218962.35 frames. ], batch size: 191, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:48:41,225 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 1.963e+02 2.333e+02 2.759e+02 6.747e+02, threshold=4.665e+02, percent-clipped=2.0 2023-05-02 13:49:20,978 INFO [train.py:904] (5/8) Epoch 28, batch 5350, loss[loss=0.1813, simple_loss=0.2853, pruned_loss=0.03864, over 15326.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2613, pruned_loss=0.03948, over 3203126.14 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:49:42,441 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3456, 3.9216, 3.9254, 2.4811, 3.4464, 3.9205, 3.4966, 2.1969], device='cuda:5'), covar=tensor([0.0616, 0.0056, 0.0058, 0.0446, 0.0128, 0.0109, 0.0128, 0.0531], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0134, 0.0102, 0.0114, 0.0098, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 13:50:32,510 INFO [train.py:904] (5/8) Epoch 28, batch 5400, loss[loss=0.1734, simple_loss=0.2724, pruned_loss=0.03717, over 15488.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.264, pruned_loss=0.04055, over 3202411.24 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:51:08,328 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 1.928e+02 2.246e+02 2.506e+02 7.644e+02, threshold=4.493e+02, percent-clipped=1.0 2023-05-02 13:51:24,276 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:51:50,353 INFO [train.py:904] (5/8) Epoch 28, batch 5450, loss[loss=0.1868, simple_loss=0.2808, pruned_loss=0.04642, over 17110.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2661, pruned_loss=0.04123, over 3205143.26 frames. ], batch size: 49, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:52:31,332 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279529.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:52:52,368 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279542.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:53:09,378 INFO [train.py:904] (5/8) Epoch 28, batch 5500, loss[loss=0.2541, simple_loss=0.318, pruned_loss=0.0951, over 11610.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2742, pruned_loss=0.04636, over 3154000.19 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:53:18,711 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 13:53:47,386 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.730e+02 3.219e+02 3.961e+02 9.894e+02, threshold=6.439e+02, percent-clipped=12.0 2023-05-02 13:54:09,413 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:28,776 INFO [train.py:904] (5/8) Epoch 28, batch 5550, loss[loss=0.2087, simple_loss=0.2911, pruned_loss=0.06315, over 16629.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2812, pruned_loss=0.05152, over 3122100.96 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:54:29,467 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:32,961 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 13:55:15,786 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0394, 2.4011, 2.5519, 1.9905, 2.6755, 2.7597, 2.4089, 2.3855], device='cuda:5'), covar=tensor([0.0716, 0.0269, 0.0243, 0.0863, 0.0133, 0.0302, 0.0436, 0.0456], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0137, 0.0086, 0.0129, 0.0128, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 13:55:48,664 INFO [train.py:904] (5/8) Epoch 28, batch 5600, loss[loss=0.2138, simple_loss=0.2922, pruned_loss=0.0677, over 17037.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2857, pruned_loss=0.05553, over 3086412.40 frames. ], batch size: 53, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:56:28,894 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.204e+02 3.162e+02 3.859e+02 4.792e+02 1.536e+03, threshold=7.717e+02, percent-clipped=11.0 2023-05-02 13:57:11,937 INFO [train.py:904] (5/8) Epoch 28, batch 5650, loss[loss=0.1905, simple_loss=0.2804, pruned_loss=0.05029, over 16619.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2899, pruned_loss=0.05886, over 3063704.31 frames. ], batch size: 57, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:58:29,312 INFO [train.py:904] (5/8) Epoch 28, batch 5700, loss[loss=0.1974, simple_loss=0.2881, pruned_loss=0.05337, over 16368.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2909, pruned_loss=0.05993, over 3063233.37 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:58:52,075 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8881, 2.7261, 2.5777, 1.8802, 2.5733, 2.6951, 2.5345, 1.9785], device='cuda:5'), covar=tensor([0.0461, 0.0101, 0.0110, 0.0421, 0.0158, 0.0152, 0.0145, 0.0419], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0135, 0.0102, 0.0115, 0.0099, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 13:59:05,448 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.265e+02 3.251e+02 3.778e+02 4.612e+02 7.364e+02, threshold=7.556e+02, percent-clipped=0.0 2023-05-02 13:59:07,564 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-02 13:59:22,605 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:59:46,950 INFO [train.py:904] (5/8) Epoch 28, batch 5750, loss[loss=0.1934, simple_loss=0.291, pruned_loss=0.0479, over 16900.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2931, pruned_loss=0.06128, over 3030492.64 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:00:08,387 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9475, 2.9437, 3.2353, 1.5964, 3.3979, 3.5188, 2.8088, 2.4356], device='cuda:5'), covar=tensor([0.1323, 0.0326, 0.0253, 0.1504, 0.0127, 0.0223, 0.0483, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0137, 0.0086, 0.0130, 0.0129, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 14:00:39,833 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:01:07,310 INFO [train.py:904] (5/8) Epoch 28, batch 5800, loss[loss=0.1759, simple_loss=0.2642, pruned_loss=0.04384, over 17062.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2925, pruned_loss=0.0601, over 3029419.01 frames. ], batch size: 55, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:01:08,160 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-05-02 14:01:46,120 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.097e+02 2.839e+02 3.367e+02 4.183e+02 8.215e+02, threshold=6.734e+02, percent-clipped=2.0 2023-05-02 14:01:57,435 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:02:06,248 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3977, 2.6654, 2.2134, 2.4654, 3.0392, 2.6827, 3.0263, 3.2128], device='cuda:5'), covar=tensor([0.0160, 0.0443, 0.0623, 0.0470, 0.0285, 0.0376, 0.0268, 0.0262], device='cuda:5'), in_proj_covar=tensor([0.0231, 0.0245, 0.0233, 0.0233, 0.0245, 0.0243, 0.0242, 0.0243], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:02:16,994 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279898.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:02:25,210 INFO [train.py:904] (5/8) Epoch 28, batch 5850, loss[loss=0.2125, simple_loss=0.2996, pruned_loss=0.06275, over 16464.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2904, pruned_loss=0.05884, over 3031370.85 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:02:54,578 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1036, 3.9827, 4.1557, 4.2884, 4.3976, 3.9947, 4.3389, 4.4137], device='cuda:5'), covar=tensor([0.1790, 0.1124, 0.1404, 0.0734, 0.0593, 0.1346, 0.0894, 0.0713], device='cuda:5'), in_proj_covar=tensor([0.0673, 0.0822, 0.0953, 0.0839, 0.0640, 0.0670, 0.0700, 0.0809], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:03:44,213 INFO [train.py:904] (5/8) Epoch 28, batch 5900, loss[loss=0.1993, simple_loss=0.2911, pruned_loss=0.05379, over 17212.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2903, pruned_loss=0.05896, over 3035558.09 frames. ], batch size: 45, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:04:12,233 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 14:04:26,172 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.117e+02 2.561e+02 3.002e+02 3.588e+02 6.021e+02, threshold=6.005e+02, percent-clipped=0.0 2023-05-02 14:04:54,042 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3629, 3.5029, 3.6121, 3.5989, 3.6198, 3.4565, 3.4904, 3.5078], device='cuda:5'), covar=tensor([0.0421, 0.0712, 0.0576, 0.0494, 0.0510, 0.0621, 0.0803, 0.0561], device='cuda:5'), in_proj_covar=tensor([0.0431, 0.0487, 0.0472, 0.0433, 0.0516, 0.0496, 0.0572, 0.0399], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 14:05:07,650 INFO [train.py:904] (5/8) Epoch 28, batch 5950, loss[loss=0.2054, simple_loss=0.2944, pruned_loss=0.05821, over 16907.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2914, pruned_loss=0.05732, over 3063377.14 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:20,839 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280051.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:06:23,834 INFO [train.py:904] (5/8) Epoch 28, batch 6000, loss[loss=0.2002, simple_loss=0.2841, pruned_loss=0.05817, over 16838.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2902, pruned_loss=0.0567, over 3081776.77 frames. ], batch size: 42, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:23,834 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 14:06:34,277 INFO [train.py:938] (5/8) Epoch 28, validation: loss=0.148, simple_loss=0.2602, pruned_loss=0.0179, over 944034.00 frames. 2023-05-02 14:06:34,278 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 14:07:11,170 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 2.746e+02 3.426e+02 4.098e+02 7.990e+02, threshold=6.852e+02, percent-clipped=5.0 2023-05-02 14:07:25,953 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280088.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:07:27,236 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.5374, 5.8876, 5.6076, 5.6609, 5.2957, 5.2426, 5.3121, 5.9854], device='cuda:5'), covar=tensor([0.1372, 0.0948, 0.1001, 0.0857, 0.0892, 0.0710, 0.1202, 0.0858], device='cuda:5'), in_proj_covar=tensor([0.0718, 0.0875, 0.0715, 0.0673, 0.0550, 0.0547, 0.0726, 0.0678], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:07:51,277 INFO [train.py:904] (5/8) Epoch 28, batch 6050, loss[loss=0.2123, simple_loss=0.3017, pruned_loss=0.0615, over 16454.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2887, pruned_loss=0.05587, over 3109117.62 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:08:00,784 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9024, 3.7759, 3.8397, 4.0978, 4.1615, 3.8534, 4.1753, 4.2318], device='cuda:5'), covar=tensor([0.1864, 0.1336, 0.1935, 0.0899, 0.0898, 0.1738, 0.1119, 0.0870], device='cuda:5'), in_proj_covar=tensor([0.0669, 0.0817, 0.0948, 0.0832, 0.0636, 0.0665, 0.0696, 0.0803], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:08:06,576 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:09:02,453 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280149.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:09:08,880 INFO [train.py:904] (5/8) Epoch 28, batch 6100, loss[loss=0.199, simple_loss=0.288, pruned_loss=0.05507, over 16481.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2884, pruned_loss=0.05505, over 3130652.66 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:09:51,401 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.432e+02 2.900e+02 3.731e+02 6.785e+02, threshold=5.800e+02, percent-clipped=0.0 2023-05-02 14:10:02,005 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280185.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:22,469 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:29,422 INFO [train.py:904] (5/8) Epoch 28, batch 6150, loss[loss=0.2021, simple_loss=0.2919, pruned_loss=0.05621, over 16754.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2869, pruned_loss=0.05458, over 3138606.74 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:11:17,168 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280233.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:36,303 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280246.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:45,497 INFO [train.py:904] (5/8) Epoch 28, batch 6200, loss[loss=0.1941, simple_loss=0.2816, pruned_loss=0.05335, over 16999.00 frames. ], tot_loss[loss=0.197, simple_loss=0.285, pruned_loss=0.05451, over 3123543.46 frames. ], batch size: 41, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:12:24,024 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.646e+02 3.232e+02 3.925e+02 8.814e+02, threshold=6.464e+02, percent-clipped=6.0 2023-05-02 14:12:34,520 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3964, 3.1139, 3.5353, 1.7785, 3.6472, 3.6635, 2.9519, 2.7547], device='cuda:5'), covar=tensor([0.0830, 0.0319, 0.0207, 0.1320, 0.0101, 0.0213, 0.0457, 0.0504], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0086, 0.0131, 0.0130, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 14:13:00,713 INFO [train.py:904] (5/8) Epoch 28, batch 6250, loss[loss=0.1784, simple_loss=0.2706, pruned_loss=0.04311, over 16358.00 frames. ], tot_loss[loss=0.196, simple_loss=0.284, pruned_loss=0.05398, over 3125807.43 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:13:05,343 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 14:13:36,636 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:13:47,818 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3722, 3.5611, 3.2372, 3.0094, 2.9887, 3.4528, 3.2199, 3.1901], device='cuda:5'), covar=tensor([0.0910, 0.0876, 0.0420, 0.0443, 0.0955, 0.0624, 0.2132, 0.0646], device='cuda:5'), in_proj_covar=tensor([0.0313, 0.0472, 0.0365, 0.0368, 0.0363, 0.0422, 0.0249, 0.0436], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:14:14,430 INFO [train.py:904] (5/8) Epoch 28, batch 6300, loss[loss=0.1865, simple_loss=0.2777, pruned_loss=0.04761, over 16652.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2832, pruned_loss=0.05297, over 3130245.27 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:14:53,564 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 2.662e+02 3.040e+02 3.867e+02 7.586e+02, threshold=6.080e+02, percent-clipped=1.0 2023-05-02 14:15:09,909 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:15:31,097 INFO [train.py:904] (5/8) Epoch 28, batch 6350, loss[loss=0.2368, simple_loss=0.307, pruned_loss=0.08325, over 11160.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2841, pruned_loss=0.0541, over 3132620.08 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:15:36,784 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:31,361 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:43,549 INFO [train.py:904] (5/8) Epoch 28, batch 6400, loss[loss=0.1772, simple_loss=0.2615, pruned_loss=0.04646, over 16707.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2851, pruned_loss=0.05544, over 3120686.42 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:17:04,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5466, 1.7680, 2.1825, 2.4597, 2.4757, 2.7598, 1.8918, 2.7163], device='cuda:5'), covar=tensor([0.0241, 0.0585, 0.0363, 0.0395, 0.0395, 0.0240, 0.0633, 0.0178], device='cuda:5'), in_proj_covar=tensor([0.0199, 0.0199, 0.0188, 0.0193, 0.0209, 0.0167, 0.0203, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:17:19,317 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.958e+02 3.396e+02 3.952e+02 7.468e+02, threshold=6.793e+02, percent-clipped=6.0 2023-05-02 14:17:52,850 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 14:17:56,100 INFO [train.py:904] (5/8) Epoch 28, batch 6450, loss[loss=0.1923, simple_loss=0.2939, pruned_loss=0.04535, over 17216.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2849, pruned_loss=0.05495, over 3113480.05 frames. ], batch size: 44, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:18:18,338 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280518.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:18:41,628 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:19:00,048 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9035, 2.3277, 1.7884, 2.0653, 2.6946, 2.3226, 2.5760, 2.8773], device='cuda:5'), covar=tensor([0.0251, 0.0465, 0.0757, 0.0553, 0.0330, 0.0447, 0.0313, 0.0313], device='cuda:5'), in_proj_covar=tensor([0.0227, 0.0241, 0.0231, 0.0230, 0.0243, 0.0240, 0.0238, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:19:02,769 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:19:13,552 INFO [train.py:904] (5/8) Epoch 28, batch 6500, loss[loss=0.1795, simple_loss=0.2681, pruned_loss=0.04547, over 16953.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2833, pruned_loss=0.05373, over 3144539.28 frames. ], batch size: 41, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:19:28,718 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7698, 3.8316, 3.9158, 3.6990, 3.8361, 4.2330, 3.9171, 3.6112], device='cuda:5'), covar=tensor([0.2239, 0.2418, 0.2948, 0.2439, 0.2573, 0.1890, 0.1700, 0.2551], device='cuda:5'), in_proj_covar=tensor([0.0429, 0.0641, 0.0705, 0.0517, 0.0694, 0.0728, 0.0546, 0.0695], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 14:19:49,976 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.728e+02 2.616e+02 3.245e+02 4.274e+02 6.880e+02, threshold=6.490e+02, percent-clipped=1.0 2023-05-02 14:19:52,118 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:01,348 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2846, 4.2800, 4.2064, 3.1322, 4.1704, 1.5759, 3.9056, 3.6339], device='cuda:5'), covar=tensor([0.0263, 0.0217, 0.0271, 0.0524, 0.0214, 0.3756, 0.0269, 0.0458], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0186, 0.0189, 0.0217, 0.0201, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:20:15,915 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280594.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:28,522 INFO [train.py:904] (5/8) Epoch 28, batch 6550, loss[loss=0.2395, simple_loss=0.3086, pruned_loss=0.08514, over 11432.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2856, pruned_loss=0.05456, over 3138321.61 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:20:32,606 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4267, 2.1441, 1.8173, 1.8836, 2.4310, 2.1104, 2.0482, 2.5166], device='cuda:5'), covar=tensor([0.0269, 0.0457, 0.0594, 0.0585, 0.0329, 0.0454, 0.0276, 0.0350], device='cuda:5'), in_proj_covar=tensor([0.0228, 0.0241, 0.0231, 0.0231, 0.0243, 0.0240, 0.0239, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:20:37,499 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:40,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6619, 3.7637, 2.4160, 4.2711, 2.8535, 4.1590, 2.4908, 3.0796], device='cuda:5'), covar=tensor([0.0317, 0.0401, 0.1581, 0.0301, 0.0849, 0.0683, 0.1566, 0.0791], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0182, 0.0196, 0.0173, 0.0180, 0.0220, 0.0204, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 14:20:45,290 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:21:41,785 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5268, 3.6062, 3.3541, 3.0421, 3.2089, 3.5004, 3.3264, 3.2951], device='cuda:5'), covar=tensor([0.0606, 0.0679, 0.0287, 0.0286, 0.0507, 0.0490, 0.1335, 0.0489], device='cuda:5'), in_proj_covar=tensor([0.0309, 0.0467, 0.0361, 0.0365, 0.0360, 0.0417, 0.0247, 0.0432], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:21:44,262 INFO [train.py:904] (5/8) Epoch 28, batch 6600, loss[loss=0.1961, simple_loss=0.2866, pruned_loss=0.05278, over 16552.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2879, pruned_loss=0.055, over 3141413.88 frames. ], batch size: 68, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:22:12,047 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2668, 5.1208, 5.2709, 5.4770, 5.6539, 4.9153, 5.6301, 5.6775], device='cuda:5'), covar=tensor([0.2109, 0.1309, 0.1755, 0.0749, 0.0543, 0.0952, 0.0607, 0.0644], device='cuda:5'), in_proj_covar=tensor([0.0669, 0.0816, 0.0947, 0.0832, 0.0638, 0.0664, 0.0698, 0.0803], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:22:18,716 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280675.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:19,224 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 14:22:21,695 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.864e+02 2.836e+02 3.315e+02 4.097e+02 1.046e+03, threshold=6.629e+02, percent-clipped=1.0 2023-05-02 14:22:27,852 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5775, 2.2905, 1.9184, 2.0392, 2.6127, 2.2325, 2.3862, 2.7479], device='cuda:5'), covar=tensor([0.0285, 0.0478, 0.0596, 0.0547, 0.0317, 0.0452, 0.0266, 0.0320], device='cuda:5'), in_proj_covar=tensor([0.0227, 0.0241, 0.0231, 0.0231, 0.0243, 0.0240, 0.0239, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:22:28,863 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:53,822 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1974, 4.2684, 4.1070, 3.8323, 3.8253, 4.2033, 3.9024, 3.9643], device='cuda:5'), covar=tensor([0.0633, 0.0621, 0.0310, 0.0336, 0.0793, 0.0526, 0.0877, 0.0693], device='cuda:5'), in_proj_covar=tensor([0.0310, 0.0468, 0.0362, 0.0365, 0.0360, 0.0418, 0.0248, 0.0433], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:23:00,547 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:23:01,181 INFO [train.py:904] (5/8) Epoch 28, batch 6650, loss[loss=0.2067, simple_loss=0.2881, pruned_loss=0.06261, over 15275.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2874, pruned_loss=0.05563, over 3138804.16 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:23:07,196 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:23:22,927 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9128, 3.1884, 3.4745, 2.1370, 3.0230, 2.0244, 3.4996, 3.5644], device='cuda:5'), covar=tensor([0.0235, 0.0859, 0.0579, 0.2154, 0.0861, 0.1148, 0.0563, 0.0828], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 14:24:02,773 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280744.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:06,810 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 14:24:15,399 INFO [train.py:904] (5/8) Epoch 28, batch 6700, loss[loss=0.2471, simple_loss=0.3076, pruned_loss=0.09334, over 11569.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2863, pruned_loss=0.05586, over 3125722.47 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:24:18,762 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:30,581 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280763.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:51,655 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.083e+02 2.827e+02 3.469e+02 4.301e+02 6.613e+02, threshold=6.939e+02, percent-clipped=0.0 2023-05-02 14:24:58,761 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0771, 3.9539, 4.1320, 4.2640, 4.3633, 3.9898, 4.2917, 4.3934], device='cuda:5'), covar=tensor([0.1782, 0.1266, 0.1422, 0.0759, 0.0625, 0.1462, 0.0961, 0.0778], device='cuda:5'), in_proj_covar=tensor([0.0666, 0.0812, 0.0941, 0.0828, 0.0633, 0.0661, 0.0695, 0.0799], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:25:13,162 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280792.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:25:29,255 INFO [train.py:904] (5/8) Epoch 28, batch 6750, loss[loss=0.1676, simple_loss=0.2504, pruned_loss=0.04243, over 16438.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2857, pruned_loss=0.0561, over 3127678.34 frames. ], batch size: 75, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:25:34,738 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 14:26:41,172 INFO [train.py:904] (5/8) Epoch 28, batch 6800, loss[loss=0.2231, simple_loss=0.3, pruned_loss=0.07311, over 11700.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2867, pruned_loss=0.05694, over 3103215.94 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:00,897 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0743, 4.2671, 4.5351, 4.4579, 4.5107, 4.2336, 4.0261, 4.1775], device='cuda:5'), covar=tensor([0.0595, 0.0754, 0.0538, 0.0678, 0.0720, 0.0682, 0.1530, 0.0642], device='cuda:5'), in_proj_covar=tensor([0.0435, 0.0489, 0.0474, 0.0439, 0.0521, 0.0500, 0.0576, 0.0401], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 14:27:13,228 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:18,678 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.178e+02 2.733e+02 3.177e+02 3.942e+02 7.279e+02, threshold=6.355e+02, percent-clipped=1.0 2023-05-02 14:27:36,560 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280889.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:55,353 INFO [train.py:904] (5/8) Epoch 28, batch 6850, loss[loss=0.217, simple_loss=0.2901, pruned_loss=0.07196, over 11484.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2877, pruned_loss=0.05732, over 3100964.85 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:56,252 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:28:53,253 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9938, 2.2154, 2.2665, 3.5667, 2.1185, 2.5424, 2.2957, 2.3308], device='cuda:5'), covar=tensor([0.1616, 0.3676, 0.3067, 0.0644, 0.4094, 0.2515, 0.3654, 0.3208], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0474, 0.0385, 0.0336, 0.0446, 0.0543, 0.0445, 0.0554], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:29:06,760 INFO [train.py:904] (5/8) Epoch 28, batch 6900, loss[loss=0.1785, simple_loss=0.2795, pruned_loss=0.03878, over 16420.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2898, pruned_loss=0.05627, over 3114746.35 frames. ], batch size: 68, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:29:34,341 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280970.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:29:45,358 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.748e+02 3.135e+02 3.737e+02 6.416e+02, threshold=6.270e+02, percent-clipped=1.0 2023-05-02 14:29:52,195 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3995, 3.3704, 3.4135, 3.4940, 3.5251, 3.2893, 3.5095, 3.5620], device='cuda:5'), covar=tensor([0.1263, 0.0887, 0.1039, 0.0635, 0.0721, 0.2121, 0.1133, 0.0954], device='cuda:5'), in_proj_covar=tensor([0.0667, 0.0813, 0.0943, 0.0829, 0.0634, 0.0662, 0.0696, 0.0801], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:29:53,988 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:30:23,412 INFO [train.py:904] (5/8) Epoch 28, batch 6950, loss[loss=0.2018, simple_loss=0.287, pruned_loss=0.05828, over 17222.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2911, pruned_loss=0.05755, over 3096333.82 frames. ], batch size: 44, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:31:03,952 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:31:18,202 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 14:31:30,330 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281049.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:31:35,801 INFO [train.py:904] (5/8) Epoch 28, batch 7000, loss[loss=0.183, simple_loss=0.284, pruned_loss=0.04101, over 16441.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2906, pruned_loss=0.0569, over 3093940.72 frames. ], batch size: 68, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:31:44,074 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281058.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:32:12,985 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.882e+02 3.440e+02 4.031e+02 6.341e+02, threshold=6.881e+02, percent-clipped=1.0 2023-05-02 14:32:50,410 INFO [train.py:904] (5/8) Epoch 28, batch 7050, loss[loss=0.1963, simple_loss=0.2951, pruned_loss=0.04878, over 16866.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2913, pruned_loss=0.05624, over 3105681.01 frames. ], batch size: 109, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:33:01,806 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281110.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:34:04,176 INFO [train.py:904] (5/8) Epoch 28, batch 7100, loss[loss=0.1898, simple_loss=0.277, pruned_loss=0.05128, over 16805.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2896, pruned_loss=0.05607, over 3107391.24 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:34:29,305 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281169.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:37,900 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281174.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:43,392 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.877e+02 3.420e+02 4.394e+02 9.887e+02, threshold=6.841e+02, percent-clipped=1.0 2023-05-02 14:34:52,706 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5438, 3.3344, 3.8732, 1.9596, 4.0252, 4.0429, 3.0254, 2.9776], device='cuda:5'), covar=tensor([0.0895, 0.0354, 0.0224, 0.1298, 0.0093, 0.0213, 0.0479, 0.0502], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0087, 0.0132, 0.0131, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 14:34:59,629 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:35:20,726 INFO [train.py:904] (5/8) Epoch 28, batch 7150, loss[loss=0.2307, simple_loss=0.2958, pruned_loss=0.08277, over 11442.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2877, pruned_loss=0.05555, over 3123308.94 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:35:21,007 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281203.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:35:47,782 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:00,697 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281230.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 14:36:09,138 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:29,669 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281251.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:31,489 INFO [train.py:904] (5/8) Epoch 28, batch 7200, loss[loss=0.1724, simple_loss=0.2609, pruned_loss=0.04196, over 16420.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2855, pruned_loss=0.05402, over 3114685.43 frames. ], batch size: 68, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:36:51,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7355, 4.0112, 3.0176, 2.4271, 2.7563, 2.7479, 4.4267, 3.5819], device='cuda:5'), covar=tensor([0.3278, 0.0679, 0.2075, 0.3118, 0.2849, 0.2185, 0.0509, 0.1534], device='cuda:5'), in_proj_covar=tensor([0.0336, 0.0274, 0.0313, 0.0327, 0.0306, 0.0277, 0.0304, 0.0352], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 14:36:56,519 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281270.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:37:07,759 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.572e+02 3.176e+02 3.806e+02 6.231e+02, threshold=6.351e+02, percent-clipped=0.0 2023-05-02 14:37:23,471 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1072, 3.6477, 4.2977, 2.2217, 4.5254, 4.4809, 3.2947, 3.3915], device='cuda:5'), covar=tensor([0.0724, 0.0325, 0.0216, 0.1201, 0.0076, 0.0157, 0.0403, 0.0451], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0139, 0.0086, 0.0131, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 14:37:23,499 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1764, 2.1880, 2.6960, 3.0810, 2.9926, 3.6277, 2.2457, 3.5435], device='cuda:5'), covar=tensor([0.0241, 0.0564, 0.0373, 0.0333, 0.0343, 0.0170, 0.0637, 0.0141], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0191, 0.0207, 0.0165, 0.0202, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:37:46,529 INFO [train.py:904] (5/8) Epoch 28, batch 7250, loss[loss=0.1844, simple_loss=0.2666, pruned_loss=0.0511, over 16532.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2833, pruned_loss=0.05308, over 3115288.50 frames. ], batch size: 75, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:37:54,645 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-05-02 14:38:09,206 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:38:16,035 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 14:38:28,142 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 14:38:59,567 INFO [train.py:904] (5/8) Epoch 28, batch 7300, loss[loss=0.197, simple_loss=0.2903, pruned_loss=0.05184, over 16627.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2832, pruned_loss=0.05348, over 3106136.08 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:39:08,080 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281358.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:39:26,045 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 14:39:39,519 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 3.043e+02 3.727e+02 4.632e+02 1.394e+03, threshold=7.454e+02, percent-clipped=7.0 2023-05-02 14:40:13,656 INFO [train.py:904] (5/8) Epoch 28, batch 7350, loss[loss=0.2367, simple_loss=0.3048, pruned_loss=0.08429, over 11125.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2843, pruned_loss=0.05463, over 3084605.11 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:40:16,320 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281405.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 14:40:17,488 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:40:50,076 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:41:08,408 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6192, 3.5593, 3.9039, 1.8393, 4.0643, 4.1054, 3.2232, 3.0453], device='cuda:5'), covar=tensor([0.0914, 0.0302, 0.0253, 0.1446, 0.0111, 0.0192, 0.0440, 0.0554], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0111, 0.0103, 0.0139, 0.0086, 0.0131, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 14:41:28,055 INFO [train.py:904] (5/8) Epoch 28, batch 7400, loss[loss=0.2424, simple_loss=0.3069, pruned_loss=0.0889, over 11505.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2856, pruned_loss=0.05518, over 3080857.24 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:41:33,675 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281456.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:41:39,049 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3883, 3.1676, 3.6239, 1.7707, 3.7357, 3.7579, 2.9308, 2.7684], device='cuda:5'), covar=tensor([0.0921, 0.0352, 0.0220, 0.1408, 0.0115, 0.0235, 0.0491, 0.0568], device='cuda:5'), in_proj_covar=tensor([0.0148, 0.0111, 0.0103, 0.0139, 0.0086, 0.0131, 0.0129, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 14:41:51,916 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-02 14:42:08,139 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.701e+02 3.025e+02 3.387e+02 6.725e+02, threshold=6.049e+02, percent-clipped=0.0 2023-05-02 14:42:18,911 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:42:42,678 INFO [train.py:904] (5/8) Epoch 28, batch 7450, loss[loss=0.2117, simple_loss=0.3074, pruned_loss=0.05797, over 16790.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2872, pruned_loss=0.05655, over 3073760.91 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:43:06,521 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:43:12,990 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3946, 2.4948, 2.4609, 4.1052, 2.3133, 2.8073, 2.4788, 2.5989], device='cuda:5'), covar=tensor([0.1383, 0.3478, 0.2967, 0.0563, 0.4074, 0.2467, 0.3677, 0.3177], device='cuda:5'), in_proj_covar=tensor([0.0421, 0.0474, 0.0385, 0.0336, 0.0446, 0.0543, 0.0447, 0.0555], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:43:17,714 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281525.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:43:58,006 INFO [train.py:904] (5/8) Epoch 28, batch 7500, loss[loss=0.1928, simple_loss=0.2778, pruned_loss=0.05386, over 16497.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2872, pruned_loss=0.05596, over 3067664.13 frames. ], batch size: 68, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:44:36,766 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.813e+02 3.506e+02 4.192e+02 1.275e+03, threshold=7.011e+02, percent-clipped=6.0 2023-05-02 14:45:05,727 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4768, 1.7241, 2.1512, 2.4179, 2.4477, 2.7158, 1.9305, 2.6185], device='cuda:5'), covar=tensor([0.0269, 0.0591, 0.0360, 0.0369, 0.0359, 0.0233, 0.0587, 0.0170], device='cuda:5'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0191, 0.0207, 0.0165, 0.0202, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:45:11,673 INFO [train.py:904] (5/8) Epoch 28, batch 7550, loss[loss=0.1888, simple_loss=0.2736, pruned_loss=0.05197, over 16630.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2857, pruned_loss=0.05572, over 3070417.81 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:46:05,333 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 14:46:25,956 INFO [train.py:904] (5/8) Epoch 28, batch 7600, loss[loss=0.2057, simple_loss=0.2954, pruned_loss=0.05799, over 16250.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2851, pruned_loss=0.05588, over 3097740.83 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:47:04,823 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.856e+02 2.608e+02 3.017e+02 3.545e+02 6.203e+02, threshold=6.033e+02, percent-clipped=0.0 2023-05-02 14:47:40,202 INFO [train.py:904] (5/8) Epoch 28, batch 7650, loss[loss=0.211, simple_loss=0.2911, pruned_loss=0.06547, over 15392.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2856, pruned_loss=0.05642, over 3100943.34 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:47:43,578 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281705.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 14:48:51,548 INFO [train.py:904] (5/8) Epoch 28, batch 7700, loss[loss=0.188, simple_loss=0.2838, pruned_loss=0.04615, over 16815.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2852, pruned_loss=0.05687, over 3088641.17 frames. ], batch size: 102, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:48:52,555 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281753.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:49:04,614 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3086, 5.3000, 5.0986, 4.3813, 5.2452, 1.7982, 4.9651, 4.8031], device='cuda:5'), covar=tensor([0.0097, 0.0102, 0.0207, 0.0422, 0.0084, 0.3063, 0.0133, 0.0247], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0173, 0.0212, 0.0184, 0.0187, 0.0216, 0.0199, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:49:31,077 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.035e+02 3.039e+02 3.444e+02 4.107e+02 7.486e+02, threshold=6.887e+02, percent-clipped=3.0 2023-05-02 14:49:35,793 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:07,525 INFO [train.py:904] (5/8) Epoch 28, batch 7750, loss[loss=0.1957, simple_loss=0.291, pruned_loss=0.05018, over 16747.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2853, pruned_loss=0.05629, over 3097209.46 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:50:21,322 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:40,418 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281825.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:03,157 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4269, 4.4144, 4.2975, 3.4517, 4.3697, 1.6719, 4.1424, 3.9225], device='cuda:5'), covar=tensor([0.0125, 0.0129, 0.0212, 0.0378, 0.0101, 0.3200, 0.0157, 0.0297], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0173, 0.0212, 0.0184, 0.0187, 0.0216, 0.0199, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:51:21,175 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281852.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:21,934 INFO [train.py:904] (5/8) Epoch 28, batch 7800, loss[loss=0.2463, simple_loss=0.3106, pruned_loss=0.09102, over 11613.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2869, pruned_loss=0.05745, over 3085149.55 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:51:40,876 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1271, 2.1325, 2.6699, 2.9576, 2.8965, 3.4880, 2.2684, 3.4897], device='cuda:5'), covar=tensor([0.0227, 0.0538, 0.0360, 0.0368, 0.0347, 0.0189, 0.0599, 0.0150], device='cuda:5'), in_proj_covar=tensor([0.0195, 0.0197, 0.0185, 0.0190, 0.0207, 0.0164, 0.0202, 0.0165], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:51:52,696 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:52:03,258 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.715e+02 3.377e+02 4.149e+02 7.003e+02, threshold=6.754e+02, percent-clipped=1.0 2023-05-02 14:52:37,335 INFO [train.py:904] (5/8) Epoch 28, batch 7850, loss[loss=0.2041, simple_loss=0.294, pruned_loss=0.05715, over 16900.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2873, pruned_loss=0.0569, over 3085967.65 frames. ], batch size: 96, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:52:47,346 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281910.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:52:51,803 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281913.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:09,871 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3428, 2.4633, 1.8540, 2.0519, 2.8275, 2.4361, 2.9262, 3.0976], device='cuda:5'), covar=tensor([0.0199, 0.0631, 0.0848, 0.0731, 0.0396, 0.0553, 0.0310, 0.0350], device='cuda:5'), in_proj_covar=tensor([0.0226, 0.0242, 0.0232, 0.0232, 0.0243, 0.0241, 0.0238, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 14:53:38,013 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:48,944 INFO [train.py:904] (5/8) Epoch 28, batch 7900, loss[loss=0.2021, simple_loss=0.2773, pruned_loss=0.06342, over 11143.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2866, pruned_loss=0.0569, over 3065347.58 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:53:58,674 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8674, 5.1971, 5.3912, 5.1151, 5.2133, 5.7396, 5.1560, 4.9595], device='cuda:5'), covar=tensor([0.1043, 0.1771, 0.2271, 0.1962, 0.2245, 0.0895, 0.1641, 0.2395], device='cuda:5'), in_proj_covar=tensor([0.0426, 0.0636, 0.0700, 0.0515, 0.0690, 0.0724, 0.0546, 0.0693], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 14:54:16,299 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281971.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 14:54:29,909 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.510e+02 2.970e+02 3.651e+02 5.631e+02, threshold=5.940e+02, percent-clipped=0.0 2023-05-02 14:54:46,604 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3389, 3.4309, 3.7652, 2.2230, 3.2686, 2.4096, 3.7086, 3.7596], device='cuda:5'), covar=tensor([0.0248, 0.0888, 0.0535, 0.2194, 0.0839, 0.1019, 0.0620, 0.1004], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0158, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 14:54:55,402 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6199, 3.7921, 4.1059, 2.4693, 3.5424, 2.6000, 3.9774, 4.0851], device='cuda:5'), covar=tensor([0.0242, 0.0762, 0.0493, 0.2038, 0.0732, 0.0958, 0.0563, 0.0872], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0158, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 14:55:09,795 INFO [train.py:904] (5/8) Epoch 28, batch 7950, loss[loss=0.2304, simple_loss=0.2991, pruned_loss=0.08087, over 11590.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2868, pruned_loss=0.0571, over 3055285.93 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:55:14,714 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:56:27,143 INFO [train.py:904] (5/8) Epoch 28, batch 8000, loss[loss=0.2014, simple_loss=0.2897, pruned_loss=0.05657, over 16904.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2873, pruned_loss=0.05741, over 3062058.34 frames. ], batch size: 116, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:57:07,769 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.806e+02 3.354e+02 4.135e+02 1.069e+03, threshold=6.709e+02, percent-clipped=7.0 2023-05-02 14:57:11,195 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:24,405 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3938, 2.9408, 2.7022, 2.3000, 2.2620, 2.2793, 2.9998, 2.8804], device='cuda:5'), covar=tensor([0.2516, 0.0710, 0.1636, 0.2550, 0.2655, 0.2376, 0.0503, 0.1386], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0273, 0.0313, 0.0328, 0.0305, 0.0278, 0.0305, 0.0352], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 14:57:33,342 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282096.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:40,318 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 14:57:42,453 INFO [train.py:904] (5/8) Epoch 28, batch 8050, loss[loss=0.2124, simple_loss=0.3006, pruned_loss=0.06205, over 16509.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2877, pruned_loss=0.05764, over 3062085.55 frames. ], batch size: 75, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:57:46,586 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6345, 2.6142, 1.9404, 2.6893, 2.2063, 2.8062, 2.1680, 2.4074], device='cuda:5'), covar=tensor([0.0325, 0.0393, 0.1272, 0.0284, 0.0677, 0.0548, 0.1195, 0.0613], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0172, 0.0180, 0.0221, 0.0205, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 14:57:56,123 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:19,039 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 14:58:22,174 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:28,401 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.88 vs. limit=5.0 2023-05-02 14:58:57,914 INFO [train.py:904] (5/8) Epoch 28, batch 8100, loss[loss=0.1981, simple_loss=0.2826, pruned_loss=0.05677, over 16218.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2868, pruned_loss=0.05686, over 3065775.87 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:59:03,749 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282157.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:59:09,073 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:59:38,413 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.655e+02 3.116e+02 4.195e+02 8.434e+02, threshold=6.231e+02, percent-clipped=4.0 2023-05-02 15:00:13,681 INFO [train.py:904] (5/8) Epoch 28, batch 8150, loss[loss=0.1757, simple_loss=0.2575, pruned_loss=0.04693, over 17028.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2836, pruned_loss=0.05522, over 3099431.55 frames. ], batch size: 55, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:00:21,752 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282208.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:00:50,135 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2305, 4.2132, 4.1207, 3.2845, 4.1764, 1.7495, 3.9857, 3.6423], device='cuda:5'), covar=tensor([0.0111, 0.0105, 0.0202, 0.0326, 0.0094, 0.3013, 0.0137, 0.0344], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0174, 0.0213, 0.0185, 0.0188, 0.0217, 0.0200, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:01:01,035 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.8796, 6.2335, 5.9153, 6.0042, 5.5986, 5.6254, 5.7162, 6.3187], device='cuda:5'), covar=tensor([0.1133, 0.0757, 0.0931, 0.0818, 0.0780, 0.0544, 0.1055, 0.0785], device='cuda:5'), in_proj_covar=tensor([0.0716, 0.0866, 0.0710, 0.0669, 0.0546, 0.0550, 0.0723, 0.0673], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:01:29,255 INFO [train.py:904] (5/8) Epoch 28, batch 8200, loss[loss=0.1749, simple_loss=0.2709, pruned_loss=0.03939, over 16476.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2811, pruned_loss=0.05427, over 3116333.39 frames. ], batch size: 75, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:01:50,076 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282266.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:02:03,614 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9471, 1.8824, 1.7224, 1.5144, 2.0159, 1.6597, 1.5811, 1.9406], device='cuda:5'), covar=tensor([0.0303, 0.0322, 0.0458, 0.0375, 0.0273, 0.0301, 0.0246, 0.0241], device='cuda:5'), in_proj_covar=tensor([0.0227, 0.0241, 0.0232, 0.0232, 0.0242, 0.0240, 0.0238, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:02:13,182 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 2.646e+02 3.224e+02 3.830e+02 8.155e+02, threshold=6.449e+02, percent-clipped=2.0 2023-05-02 15:02:46,940 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282301.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:02:50,325 INFO [train.py:904] (5/8) Epoch 28, batch 8250, loss[loss=0.1785, simple_loss=0.267, pruned_loss=0.04503, over 12391.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2805, pruned_loss=0.05209, over 3094897.12 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:03:17,840 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 15:04:07,366 INFO [train.py:904] (5/8) Epoch 28, batch 8300, loss[loss=0.1836, simple_loss=0.2824, pruned_loss=0.04241, over 15253.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2779, pruned_loss=0.04904, over 3101914.72 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:04:31,772 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7528, 2.2347, 1.9358, 1.9655, 2.5437, 2.1601, 2.2611, 2.6332], device='cuda:5'), covar=tensor([0.0207, 0.0431, 0.0556, 0.0469, 0.0271, 0.0399, 0.0238, 0.0282], device='cuda:5'), in_proj_covar=tensor([0.0223, 0.0238, 0.0229, 0.0229, 0.0239, 0.0238, 0.0235, 0.0238], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:04:50,784 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.126e+02 2.473e+02 2.954e+02 5.391e+02, threshold=4.946e+02, percent-clipped=0.0 2023-05-02 15:05:25,581 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 15:05:26,438 INFO [train.py:904] (5/8) Epoch 28, batch 8350, loss[loss=0.1891, simple_loss=0.287, pruned_loss=0.04558, over 16825.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2779, pruned_loss=0.04787, over 3075040.72 frames. ], batch size: 116, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:06:33,099 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4524, 1.8759, 2.1876, 2.4875, 2.5727, 2.7933, 2.0140, 2.7467], device='cuda:5'), covar=tensor([0.0274, 0.0592, 0.0378, 0.0382, 0.0374, 0.0266, 0.0583, 0.0173], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0196, 0.0184, 0.0189, 0.0206, 0.0163, 0.0200, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:06:43,078 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282452.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:06:43,834 INFO [train.py:904] (5/8) Epoch 28, batch 8400, loss[loss=0.1593, simple_loss=0.2578, pruned_loss=0.03041, over 15338.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2753, pruned_loss=0.04574, over 3076672.05 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:07:26,992 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.221e+02 2.637e+02 3.332e+02 6.864e+02, threshold=5.273e+02, percent-clipped=5.0 2023-05-02 15:07:40,608 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:07:43,749 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 15:08:02,092 INFO [train.py:904] (5/8) Epoch 28, batch 8450, loss[loss=0.1757, simple_loss=0.2669, pruned_loss=0.0422, over 12242.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2736, pruned_loss=0.044, over 3079151.82 frames. ], batch size: 247, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:08:11,393 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:17,291 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:21,852 INFO [train.py:904] (5/8) Epoch 28, batch 8500, loss[loss=0.1616, simple_loss=0.2542, pruned_loss=0.0345, over 16774.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2699, pruned_loss=0.04203, over 3065646.08 frames. ], batch size: 124, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:09:28,043 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:43,299 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282566.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:10:06,467 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6562, 2.6426, 1.9809, 2.7758, 2.1869, 2.7972, 2.2214, 2.4695], device='cuda:5'), covar=tensor([0.0325, 0.0324, 0.1203, 0.0331, 0.0640, 0.0491, 0.1195, 0.0579], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0178, 0.0194, 0.0170, 0.0177, 0.0217, 0.0202, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 15:10:07,072 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.338e+02 2.128e+02 2.662e+02 3.202e+02 5.740e+02, threshold=5.324e+02, percent-clipped=2.0 2023-05-02 15:10:42,309 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282601.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:10:45,533 INFO [train.py:904] (5/8) Epoch 28, batch 8550, loss[loss=0.1543, simple_loss=0.2384, pruned_loss=0.03512, over 12028.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2672, pruned_loss=0.04088, over 3053407.88 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:11:05,503 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282614.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:12:14,466 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:12:21,561 INFO [train.py:904] (5/8) Epoch 28, batch 8600, loss[loss=0.1626, simple_loss=0.2507, pruned_loss=0.03719, over 12599.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2671, pruned_loss=0.03969, over 3046385.71 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:12:32,748 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282658.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:13:12,606 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282677.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:13:17,327 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.220e+02 2.463e+02 3.072e+02 6.042e+02, threshold=4.927e+02, percent-clipped=1.0 2023-05-02 15:13:50,851 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282698.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:13:50,940 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9359, 2.1363, 2.3422, 3.1673, 2.1620, 2.3314, 2.2912, 2.2019], device='cuda:5'), covar=tensor([0.1414, 0.3711, 0.2881, 0.0777, 0.4674, 0.2745, 0.3593, 0.3983], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0466, 0.0379, 0.0328, 0.0439, 0.0533, 0.0440, 0.0545], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:13:58,782 INFO [train.py:904] (5/8) Epoch 28, batch 8650, loss[loss=0.1621, simple_loss=0.2634, pruned_loss=0.0304, over 16820.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2654, pruned_loss=0.03837, over 3052834.57 frames. ], batch size: 124, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:14:20,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3728, 3.4429, 3.9990, 2.2020, 3.2854, 2.3923, 3.7919, 3.6663], device='cuda:5'), covar=tensor([0.0213, 0.0889, 0.0491, 0.2161, 0.0770, 0.1063, 0.0561, 0.0993], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0154, 0.0145, 0.0130, 0.0143, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 15:14:30,439 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2586, 1.6499, 1.9855, 2.2396, 2.3638, 2.5192, 1.9233, 2.4531], device='cuda:5'), covar=tensor([0.0302, 0.0663, 0.0383, 0.0424, 0.0412, 0.0238, 0.0574, 0.0182], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0196, 0.0184, 0.0188, 0.0205, 0.0163, 0.0199, 0.0163], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:14:36,891 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282719.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:14,389 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:34,457 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 15:15:40,283 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:40,926 INFO [train.py:904] (5/8) Epoch 28, batch 8700, loss[loss=0.1565, simple_loss=0.2541, pruned_loss=0.02943, over 16587.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2626, pruned_loss=0.03728, over 3035068.55 frames. ], batch size: 68, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:15:46,267 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7425, 3.4006, 3.6669, 2.0146, 3.8236, 3.9010, 3.1235, 3.0531], device='cuda:5'), covar=tensor([0.0615, 0.0284, 0.0224, 0.1159, 0.0096, 0.0199, 0.0409, 0.0447], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0108, 0.0099, 0.0135, 0.0083, 0.0127, 0.0126, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 15:15:53,198 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282759.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 15:16:21,802 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2151, 5.5257, 5.3141, 5.3008, 5.0252, 5.0347, 4.9122, 5.6096], device='cuda:5'), covar=tensor([0.1150, 0.0894, 0.0929, 0.0754, 0.0806, 0.0793, 0.1234, 0.0881], device='cuda:5'), in_proj_covar=tensor([0.0701, 0.0848, 0.0695, 0.0656, 0.0535, 0.0538, 0.0707, 0.0659], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:16:29,665 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.234e+02 2.610e+02 3.054e+02 5.139e+02, threshold=5.220e+02, percent-clipped=1.0 2023-05-02 15:16:33,539 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 15:17:06,891 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282800.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:17:12,009 INFO [train.py:904] (5/8) Epoch 28, batch 8750, loss[loss=0.1427, simple_loss=0.2348, pruned_loss=0.02524, over 12331.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2624, pruned_loss=0.03715, over 3030150.88 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:17:18,897 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282805.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:17:25,009 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5746, 4.7609, 4.8795, 4.7341, 4.7794, 5.2835, 4.7694, 4.4694], device='cuda:5'), covar=tensor([0.1152, 0.1702, 0.2069, 0.1886, 0.2363, 0.0829, 0.1488, 0.2340], device='cuda:5'), in_proj_covar=tensor([0.0416, 0.0621, 0.0684, 0.0502, 0.0673, 0.0710, 0.0535, 0.0674], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 15:18:47,228 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:19:02,846 INFO [train.py:904] (5/8) Epoch 28, batch 8800, loss[loss=0.1696, simple_loss=0.2663, pruned_loss=0.03642, over 16850.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2607, pruned_loss=0.03575, over 3052053.13 frames. ], batch size: 124, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:19:19,614 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 15:19:29,280 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:19:48,683 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6281, 4.9189, 5.0433, 4.8323, 4.8886, 5.3872, 4.8855, 4.6861], device='cuda:5'), covar=tensor([0.1233, 0.1544, 0.1704, 0.1911, 0.2294, 0.0876, 0.1644, 0.2370], device='cuda:5'), in_proj_covar=tensor([0.0415, 0.0617, 0.0681, 0.0500, 0.0671, 0.0708, 0.0532, 0.0671], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 15:20:00,728 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.123e+02 2.465e+02 3.098e+02 6.856e+02, threshold=4.929e+02, percent-clipped=4.0 2023-05-02 15:20:06,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9406, 2.0842, 2.0902, 3.5288, 1.9933, 2.3754, 2.2181, 2.1904], device='cuda:5'), covar=tensor([0.1575, 0.4074, 0.3507, 0.0686, 0.4772, 0.2944, 0.4028, 0.3943], device='cuda:5'), in_proj_covar=tensor([0.0412, 0.0466, 0.0380, 0.0328, 0.0439, 0.0533, 0.0439, 0.0544], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:20:47,892 INFO [train.py:904] (5/8) Epoch 28, batch 8850, loss[loss=0.1715, simple_loss=0.2791, pruned_loss=0.03191, over 16417.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2639, pruned_loss=0.03522, over 3065866.36 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:22:35,075 INFO [train.py:904] (5/8) Epoch 28, batch 8900, loss[loss=0.1547, simple_loss=0.2532, pruned_loss=0.02812, over 16595.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.264, pruned_loss=0.03458, over 3059583.17 frames. ], batch size: 62, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:23:38,032 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.086e+02 2.432e+02 2.867e+02 4.868e+02, threshold=4.864e+02, percent-clipped=0.0 2023-05-02 15:24:39,875 INFO [train.py:904] (5/8) Epoch 28, batch 8950, loss[loss=0.1396, simple_loss=0.2358, pruned_loss=0.02176, over 16767.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2633, pruned_loss=0.03465, over 3074790.93 frames. ], batch size: 76, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:25:04,690 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283014.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:25:45,371 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:26:27,358 INFO [train.py:904] (5/8) Epoch 28, batch 9000, loss[loss=0.1437, simple_loss=0.2425, pruned_loss=0.02246, over 16889.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2604, pruned_loss=0.03358, over 3075495.24 frames. ], batch size: 102, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:26:27,358 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 15:26:38,046 INFO [train.py:938] (5/8) Epoch 28, validation: loss=0.1436, simple_loss=0.2472, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 15:26:38,047 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 15:26:41,529 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283054.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 15:26:59,839 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1936, 3.6362, 3.6815, 2.4201, 3.2625, 3.7118, 3.4698, 2.0753], device='cuda:5'), covar=tensor([0.0572, 0.0060, 0.0057, 0.0445, 0.0134, 0.0096, 0.0090, 0.0533], device='cuda:5'), in_proj_covar=tensor([0.0133, 0.0086, 0.0087, 0.0130, 0.0099, 0.0111, 0.0094, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 15:27:36,920 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.008e+02 2.381e+02 2.804e+02 4.985e+02, threshold=4.761e+02, percent-clipped=1.0 2023-05-02 15:28:19,142 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9871, 2.9554, 2.9881, 1.7931, 3.1468, 3.3296, 2.7924, 2.4574], device='cuda:5'), covar=tensor([0.1132, 0.0249, 0.0220, 0.1345, 0.0137, 0.0211, 0.0484, 0.0651], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0107, 0.0098, 0.0135, 0.0083, 0.0126, 0.0125, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 15:28:21,094 INFO [train.py:904] (5/8) Epoch 28, batch 9050, loss[loss=0.1709, simple_loss=0.2599, pruned_loss=0.04094, over 15521.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2617, pruned_loss=0.03406, over 3073561.88 frames. ], batch size: 192, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:28:48,333 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:46,461 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:48,326 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:04,167 INFO [train.py:904] (5/8) Epoch 28, batch 9100, loss[loss=0.1593, simple_loss=0.2608, pruned_loss=0.02885, over 15553.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2611, pruned_loss=0.03414, over 3079553.42 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:30:20,703 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283161.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:36,187 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4089, 3.4901, 3.6780, 3.6634, 3.6795, 3.5090, 3.5297, 3.5559], device='cuda:5'), covar=tensor([0.0408, 0.0695, 0.0515, 0.0508, 0.0475, 0.0509, 0.0789, 0.0524], device='cuda:5'), in_proj_covar=tensor([0.0426, 0.0478, 0.0465, 0.0428, 0.0510, 0.0488, 0.0561, 0.0395], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 15:30:58,171 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:31:08,553 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.173e+02 2.551e+02 2.852e+02 5.102e+02, threshold=5.103e+02, percent-clipped=2.0 2023-05-02 15:31:18,459 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2535, 4.3417, 4.4847, 4.2097, 4.4014, 4.8365, 4.3976, 4.0173], device='cuda:5'), covar=tensor([0.1659, 0.1934, 0.1953, 0.2289, 0.2279, 0.0986, 0.1449, 0.2514], device='cuda:5'), in_proj_covar=tensor([0.0414, 0.0614, 0.0677, 0.0498, 0.0667, 0.0706, 0.0529, 0.0669], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 15:31:39,559 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283193.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:32:01,013 INFO [train.py:904] (5/8) Epoch 28, batch 9150, loss[loss=0.1623, simple_loss=0.2539, pruned_loss=0.03537, over 12049.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2615, pruned_loss=0.03402, over 3067251.28 frames. ], batch size: 250, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:32:07,125 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283205.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:33:29,191 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8473, 1.9588, 2.3748, 3.1137, 2.1174, 2.1012, 2.2236, 2.0383], device='cuda:5'), covar=tensor([0.1774, 0.4587, 0.3164, 0.0979, 0.6024, 0.3917, 0.3821, 0.5535], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0463, 0.0378, 0.0326, 0.0437, 0.0530, 0.0437, 0.0541], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:33:44,283 INFO [train.py:904] (5/8) Epoch 28, batch 9200, loss[loss=0.1596, simple_loss=0.2584, pruned_loss=0.03037, over 16246.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2575, pruned_loss=0.03347, over 3056815.84 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:34:34,271 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.118e+02 2.676e+02 3.077e+02 8.618e+02, threshold=5.351e+02, percent-clipped=1.0 2023-05-02 15:34:39,311 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:35:20,008 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-02 15:35:20,281 INFO [train.py:904] (5/8) Epoch 28, batch 9250, loss[loss=0.1528, simple_loss=0.2533, pruned_loss=0.02618, over 16705.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2575, pruned_loss=0.03384, over 3040066.23 frames. ], batch size: 83, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:35:44,175 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283314.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:35:54,089 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5122, 2.3838, 2.2685, 4.3160, 2.2972, 2.8081, 2.4626, 2.5457], device='cuda:5'), covar=tensor([0.1278, 0.3813, 0.3500, 0.0491, 0.4507, 0.2639, 0.3719, 0.3593], device='cuda:5'), in_proj_covar=tensor([0.0411, 0.0464, 0.0379, 0.0326, 0.0438, 0.0530, 0.0438, 0.0541], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:36:28,352 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:36:55,404 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:37:14,584 INFO [train.py:904] (5/8) Epoch 28, batch 9300, loss[loss=0.1654, simple_loss=0.2465, pruned_loss=0.04217, over 12254.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2555, pruned_loss=0.03315, over 3036078.72 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:37:16,993 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283354.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:37:34,864 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:38:16,509 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.947e+02 2.341e+02 2.771e+02 4.646e+02, threshold=4.683e+02, percent-clipped=0.0 2023-05-02 15:38:17,946 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283381.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:38:58,954 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283402.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:38:59,604 INFO [train.py:904] (5/8) Epoch 28, batch 9350, loss[loss=0.1924, simple_loss=0.2859, pruned_loss=0.04949, over 16301.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2552, pruned_loss=0.03313, over 3027468.40 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:40:02,348 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-02 15:40:41,032 INFO [train.py:904] (5/8) Epoch 28, batch 9400, loss[loss=0.1437, simple_loss=0.2319, pruned_loss=0.02774, over 12475.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2558, pruned_loss=0.03317, over 3031266.61 frames. ], batch size: 247, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:40:57,798 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9404, 4.8815, 4.6586, 4.1004, 4.7963, 1.8101, 4.5587, 4.3966], device='cuda:5'), covar=tensor([0.0094, 0.0113, 0.0208, 0.0308, 0.0102, 0.2842, 0.0128, 0.0266], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0170, 0.0207, 0.0179, 0.0184, 0.0213, 0.0196, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:41:00,062 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283461.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:19,450 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:20,025 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-05-02 15:41:39,547 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.116e+02 2.396e+02 2.984e+02 4.135e+02, threshold=4.791e+02, percent-clipped=0.0 2023-05-02 15:42:20,577 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283500.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:42:24,705 INFO [train.py:904] (5/8) Epoch 28, batch 9450, loss[loss=0.1761, simple_loss=0.2652, pruned_loss=0.04346, over 12606.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2577, pruned_loss=0.03362, over 3021592.44 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:42:36,846 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283509.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:43:23,215 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 15:43:48,966 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283544.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:44:06,799 INFO [train.py:904] (5/8) Epoch 28, batch 9500, loss[loss=0.1657, simple_loss=0.2561, pruned_loss=0.03771, over 17032.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2566, pruned_loss=0.03324, over 3039663.43 frames. ], batch size: 55, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:45:03,876 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.031e+02 2.419e+02 2.973e+02 5.266e+02, threshold=4.838e+02, percent-clipped=2.0 2023-05-02 15:45:53,231 INFO [train.py:904] (5/8) Epoch 28, batch 9550, loss[loss=0.1613, simple_loss=0.2659, pruned_loss=0.02836, over 15419.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2572, pruned_loss=0.03371, over 3051023.77 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:45:59,250 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283605.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 15:46:04,923 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 15:47:09,795 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283639.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:47:34,673 INFO [train.py:904] (5/8) Epoch 28, batch 9600, loss[loss=0.1867, simple_loss=0.2837, pruned_loss=0.04488, over 16198.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2577, pruned_loss=0.03426, over 3031876.82 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:48:11,570 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5943, 3.5475, 3.5212, 2.8465, 3.4806, 2.0691, 3.3226, 2.9006], device='cuda:5'), covar=tensor([0.0145, 0.0131, 0.0200, 0.0209, 0.0112, 0.2490, 0.0139, 0.0294], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0168, 0.0205, 0.0177, 0.0182, 0.0211, 0.0193, 0.0172], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:48:29,446 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.192e+02 2.621e+02 3.097e+02 5.622e+02, threshold=5.243e+02, percent-clipped=5.0 2023-05-02 15:49:23,006 INFO [train.py:904] (5/8) Epoch 28, batch 9650, loss[loss=0.166, simple_loss=0.2621, pruned_loss=0.03498, over 16369.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2589, pruned_loss=0.03441, over 3028987.57 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:49:31,888 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-02 15:51:01,687 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8413, 1.3934, 1.7535, 1.7179, 1.8718, 1.8560, 1.7062, 1.8786], device='cuda:5'), covar=tensor([0.0314, 0.0483, 0.0248, 0.0335, 0.0359, 0.0259, 0.0492, 0.0156], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0195, 0.0182, 0.0186, 0.0204, 0.0161, 0.0199, 0.0161], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:51:10,216 INFO [train.py:904] (5/8) Epoch 28, batch 9700, loss[loss=0.1682, simple_loss=0.264, pruned_loss=0.0362, over 12711.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2585, pruned_loss=0.03439, over 3032052.94 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:51:47,267 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283771.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:52:08,996 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.180e+02 2.423e+02 2.886e+02 5.203e+02, threshold=4.846e+02, percent-clipped=0.0 2023-05-02 15:52:48,208 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283800.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:52:52,813 INFO [train.py:904] (5/8) Epoch 28, batch 9750, loss[loss=0.163, simple_loss=0.2478, pruned_loss=0.03909, over 12683.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2569, pruned_loss=0.03425, over 3052168.96 frames. ], batch size: 250, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:53:24,794 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283819.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:13,361 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283842.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:24,073 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283848.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:31,912 INFO [train.py:904] (5/8) Epoch 28, batch 9800, loss[loss=0.1681, simple_loss=0.2724, pruned_loss=0.03191, over 16180.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2574, pruned_loss=0.03364, over 3068564.85 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:55:23,057 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.061e+02 2.416e+02 3.000e+02 5.773e+02, threshold=4.831e+02, percent-clipped=3.0 2023-05-02 15:56:11,012 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283900.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 15:56:11,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7054, 4.8183, 5.0848, 5.0630, 5.1179, 4.8773, 4.7097, 4.7148], device='cuda:5'), covar=tensor([0.0541, 0.0832, 0.0595, 0.0621, 0.0634, 0.0596, 0.1276, 0.0530], device='cuda:5'), in_proj_covar=tensor([0.0422, 0.0475, 0.0461, 0.0423, 0.0506, 0.0483, 0.0555, 0.0391], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 15:56:13,287 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8637, 3.7955, 3.9264, 4.0134, 4.0867, 3.6958, 4.0673, 4.1210], device='cuda:5'), covar=tensor([0.1582, 0.1103, 0.1257, 0.0709, 0.0596, 0.1922, 0.0768, 0.0776], device='cuda:5'), in_proj_covar=tensor([0.0640, 0.0783, 0.0903, 0.0799, 0.0607, 0.0633, 0.0668, 0.0773], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 15:56:15,652 INFO [train.py:904] (5/8) Epoch 28, batch 9850, loss[loss=0.1674, simple_loss=0.272, pruned_loss=0.03142, over 16999.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2589, pruned_loss=0.03335, over 3079644.94 frames. ], batch size: 41, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:56:17,427 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:57:04,869 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-05-02 15:57:14,612 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-05-02 15:57:36,961 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:58:06,907 INFO [train.py:904] (5/8) Epoch 28, batch 9900, loss[loss=0.1775, simple_loss=0.2839, pruned_loss=0.03561, over 16818.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2598, pruned_loss=0.03344, over 3088088.95 frames. ], batch size: 124, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:58:38,492 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:59:13,295 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 2.134e+02 2.412e+02 2.887e+02 8.278e+02, threshold=4.823e+02, percent-clipped=3.0 2023-05-02 15:59:30,177 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283987.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:59:56,987 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-05-02 16:00:07,765 INFO [train.py:904] (5/8) Epoch 28, batch 9950, loss[loss=0.1615, simple_loss=0.2608, pruned_loss=0.03109, over 16692.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2616, pruned_loss=0.03381, over 3079624.24 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:01:08,224 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:02:08,466 INFO [train.py:904] (5/8) Epoch 28, batch 10000, loss[loss=0.1633, simple_loss=0.2507, pruned_loss=0.03794, over 12748.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2604, pruned_loss=0.0334, over 3090733.12 frames. ], batch size: 250, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:02:54,329 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0628, 3.1415, 2.1480, 3.2985, 2.3923, 3.2949, 2.1692, 2.6176], device='cuda:5'), covar=tensor([0.0328, 0.0377, 0.1409, 0.0319, 0.0838, 0.0581, 0.1469, 0.0723], device='cuda:5'), in_proj_covar=tensor([0.0171, 0.0176, 0.0194, 0.0167, 0.0176, 0.0213, 0.0202, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 16:03:03,888 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.120e+02 2.490e+02 3.242e+02 7.022e+02, threshold=4.980e+02, percent-clipped=2.0 2023-05-02 16:03:50,393 INFO [train.py:904] (5/8) Epoch 28, batch 10050, loss[loss=0.1695, simple_loss=0.2645, pruned_loss=0.03721, over 16659.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2601, pruned_loss=0.0331, over 3083006.60 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:05:03,435 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 16:05:24,432 INFO [train.py:904] (5/8) Epoch 28, batch 10100, loss[loss=0.1737, simple_loss=0.2651, pruned_loss=0.04116, over 16885.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2607, pruned_loss=0.03339, over 3092076.21 frames. ], batch size: 109, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:06:20,722 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.222e+02 2.734e+02 3.195e+02 6.018e+02, threshold=5.468e+02, percent-clipped=8.0 2023-05-02 16:06:39,384 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:06:42,065 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284200.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:06:44,464 INFO [train.py:904] (5/8) Epoch 28, batch 10150, loss[loss=0.1524, simple_loss=0.2423, pruned_loss=0.03121, over 11962.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2595, pruned_loss=0.0335, over 3067221.67 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:07:10,318 INFO [train.py:904] (5/8) Epoch 29, batch 0, loss[loss=0.1777, simple_loss=0.2685, pruned_loss=0.04341, over 17120.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2685, pruned_loss=0.04341, over 17120.00 frames. ], batch size: 47, lr: 2.34e-03, grad_scale: 8.0 2023-05-02 16:07:10,319 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 16:07:17,744 INFO [train.py:938] (5/8) Epoch 29, validation: loss=0.1427, simple_loss=0.246, pruned_loss=0.0197, over 944034.00 frames. 2023-05-02 16:07:17,744 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 16:07:53,554 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 16:08:07,728 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4937, 3.7486, 3.9266, 2.7250, 3.5877, 4.0457, 3.6498, 2.0425], device='cuda:5'), covar=tensor([0.0632, 0.0400, 0.0090, 0.0489, 0.0156, 0.0123, 0.0139, 0.0670], device='cuda:5'), in_proj_covar=tensor([0.0135, 0.0087, 0.0088, 0.0132, 0.0100, 0.0111, 0.0095, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-02 16:08:10,880 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9988, 3.0759, 3.2680, 2.1605, 2.8987, 2.3298, 3.5444, 3.5219], device='cuda:5'), covar=tensor([0.0275, 0.1119, 0.0734, 0.2205, 0.0990, 0.1171, 0.0599, 0.0985], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0163, 0.0165, 0.0153, 0.0143, 0.0129, 0.0141, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 16:08:18,233 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284248.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:08:26,888 INFO [train.py:904] (5/8) Epoch 29, batch 50, loss[loss=0.1616, simple_loss=0.2544, pruned_loss=0.03436, over 16529.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2616, pruned_loss=0.04342, over 757939.89 frames. ], batch size: 68, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:08:28,825 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 16:09:08,279 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.422e+02 3.076e+02 3.842e+02 2.178e+03, threshold=6.152e+02, percent-clipped=5.0 2023-05-02 16:09:37,389 INFO [train.py:904] (5/8) Epoch 29, batch 100, loss[loss=0.168, simple_loss=0.2707, pruned_loss=0.03268, over 17136.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.258, pruned_loss=0.04218, over 1320922.64 frames. ], batch size: 48, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:10:02,645 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:10:46,104 INFO [train.py:904] (5/8) Epoch 29, batch 150, loss[loss=0.1499, simple_loss=0.2422, pruned_loss=0.02883, over 17098.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2578, pruned_loss=0.04135, over 1773802.15 frames. ], batch size: 47, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:11:25,632 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.248e+02 2.614e+02 3.025e+02 1.081e+03, threshold=5.228e+02, percent-clipped=2.0 2023-05-02 16:11:49,906 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-02 16:11:53,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5284, 4.2728, 4.4252, 4.7084, 4.7679, 4.3868, 4.7873, 4.7909], device='cuda:5'), covar=tensor([0.1719, 0.1578, 0.1935, 0.0932, 0.0922, 0.1309, 0.1940, 0.1143], device='cuda:5'), in_proj_covar=tensor([0.0652, 0.0797, 0.0918, 0.0811, 0.0617, 0.0642, 0.0681, 0.0787], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:11:55,153 INFO [train.py:904] (5/8) Epoch 29, batch 200, loss[loss=0.1707, simple_loss=0.2514, pruned_loss=0.04499, over 16442.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2585, pruned_loss=0.04168, over 2115271.65 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:12:51,484 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284443.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:13:04,671 INFO [train.py:904] (5/8) Epoch 29, batch 250, loss[loss=0.1621, simple_loss=0.2445, pruned_loss=0.03984, over 16488.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2562, pruned_loss=0.04079, over 2389419.27 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:13:47,580 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.177e+02 2.500e+02 3.084e+02 6.085e+02, threshold=5.001e+02, percent-clipped=1.0 2023-05-02 16:14:08,395 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284498.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:14:12,119 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8594, 4.6138, 4.9218, 5.1076, 5.2846, 4.6378, 5.2527, 5.2811], device='cuda:5'), covar=tensor([0.2163, 0.1484, 0.1871, 0.0812, 0.0646, 0.1135, 0.0689, 0.0752], device='cuda:5'), in_proj_covar=tensor([0.0657, 0.0802, 0.0925, 0.0817, 0.0621, 0.0647, 0.0686, 0.0792], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:14:17,002 INFO [train.py:904] (5/8) Epoch 29, batch 300, loss[loss=0.1607, simple_loss=0.2377, pruned_loss=0.04186, over 16796.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2546, pruned_loss=0.04004, over 2595686.26 frames. ], batch size: 83, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:14:17,416 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284504.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:14:48,865 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 16:15:14,307 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284546.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:15:23,227 INFO [train.py:904] (5/8) Epoch 29, batch 350, loss[loss=0.1391, simple_loss=0.233, pruned_loss=0.02262, over 17250.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2524, pruned_loss=0.03911, over 2758131.03 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:15:25,856 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 16:16:02,936 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.083e+02 2.374e+02 2.884e+02 4.611e+02, threshold=4.748e+02, percent-clipped=0.0 2023-05-02 16:16:31,992 INFO [train.py:904] (5/8) Epoch 29, batch 400, loss[loss=0.1733, simple_loss=0.2526, pruned_loss=0.04702, over 16493.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2515, pruned_loss=0.03839, over 2881059.45 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:16:57,144 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:17:41,209 INFO [train.py:904] (5/8) Epoch 29, batch 450, loss[loss=0.1819, simple_loss=0.258, pruned_loss=0.05287, over 16734.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2495, pruned_loss=0.03764, over 2980585.11 frames. ], batch size: 134, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:18:03,054 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284670.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:18:16,759 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8697, 4.4495, 3.1393, 2.3997, 2.5857, 2.6336, 4.7937, 3.4923], device='cuda:5'), covar=tensor([0.3041, 0.0549, 0.1915, 0.3247, 0.3250, 0.2294, 0.0325, 0.1579], device='cuda:5'), in_proj_covar=tensor([0.0336, 0.0272, 0.0312, 0.0326, 0.0302, 0.0278, 0.0303, 0.0351], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 16:18:18,521 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 1.978e+02 2.312e+02 2.858e+02 4.933e+02, threshold=4.623e+02, percent-clipped=2.0 2023-05-02 16:18:47,413 INFO [train.py:904] (5/8) Epoch 29, batch 500, loss[loss=0.1496, simple_loss=0.249, pruned_loss=0.02513, over 17136.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2491, pruned_loss=0.03694, over 3064448.21 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:19:56,203 INFO [train.py:904] (5/8) Epoch 29, batch 550, loss[loss=0.1843, simple_loss=0.2682, pruned_loss=0.05013, over 16912.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2491, pruned_loss=0.03679, over 3126175.62 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:20:35,719 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 1.886e+02 2.233e+02 2.634e+02 5.354e+02, threshold=4.465e+02, percent-clipped=1.0 2023-05-02 16:20:57,909 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:21:04,195 INFO [train.py:904] (5/8) Epoch 29, batch 600, loss[loss=0.1688, simple_loss=0.2578, pruned_loss=0.0399, over 17101.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2487, pruned_loss=0.0371, over 3159705.83 frames. ], batch size: 53, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:21:21,954 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:22:00,235 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 16:22:12,362 INFO [train.py:904] (5/8) Epoch 29, batch 650, loss[loss=0.2101, simple_loss=0.2772, pruned_loss=0.0715, over 16919.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2474, pruned_loss=0.03694, over 3191312.48 frames. ], batch size: 116, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:22:46,739 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284877.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:22:53,843 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.111e+02 2.518e+02 3.065e+02 7.660e+02, threshold=5.037e+02, percent-clipped=4.0 2023-05-02 16:23:22,624 INFO [train.py:904] (5/8) Epoch 29, batch 700, loss[loss=0.142, simple_loss=0.2327, pruned_loss=0.02565, over 17002.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2469, pruned_loss=0.03661, over 3216611.12 frames. ], batch size: 41, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:23:23,052 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1682, 4.8997, 5.1477, 5.3526, 5.5684, 4.8837, 5.5201, 5.5354], device='cuda:5'), covar=tensor([0.2078, 0.1407, 0.1889, 0.0821, 0.0615, 0.0869, 0.0595, 0.0718], device='cuda:5'), in_proj_covar=tensor([0.0679, 0.0828, 0.0956, 0.0843, 0.0641, 0.0665, 0.0710, 0.0817], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:23:25,846 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-02 16:24:12,245 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:24:30,119 INFO [train.py:904] (5/8) Epoch 29, batch 750, loss[loss=0.1669, simple_loss=0.2508, pruned_loss=0.04154, over 16794.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2476, pruned_loss=0.0373, over 3241258.86 frames. ], batch size: 83, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:25:13,141 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.091e+02 2.322e+02 2.629e+02 4.322e+02, threshold=4.644e+02, percent-clipped=0.0 2023-05-02 16:25:30,136 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284996.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:37,981 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:42,931 INFO [train.py:904] (5/8) Epoch 29, batch 800, loss[loss=0.1676, simple_loss=0.2466, pruned_loss=0.04435, over 16802.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2474, pruned_loss=0.03762, over 3262608.30 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:26:34,306 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285042.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:26:49,085 INFO [train.py:904] (5/8) Epoch 29, batch 850, loss[loss=0.1565, simple_loss=0.2554, pruned_loss=0.02885, over 17129.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2467, pruned_loss=0.03734, over 3272366.31 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:26:53,076 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285057.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:27:31,500 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.142e+02 2.515e+02 3.015e+02 5.883e+02, threshold=5.030e+02, percent-clipped=4.0 2023-05-02 16:27:49,582 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:27:55,035 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285103.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:27:55,655 INFO [train.py:904] (5/8) Epoch 29, batch 900, loss[loss=0.1826, simple_loss=0.261, pruned_loss=0.05208, over 16692.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2461, pruned_loss=0.03685, over 3282902.15 frames. ], batch size: 89, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:28:15,847 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 16:28:34,603 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2023-05-02 16:28:56,111 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285147.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:03,778 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5130, 1.7702, 2.1542, 2.3365, 2.5063, 2.4475, 1.9007, 2.5893], device='cuda:5'), covar=tensor([0.0243, 0.0557, 0.0380, 0.0359, 0.0368, 0.0394, 0.0588, 0.0208], device='cuda:5'), in_proj_covar=tensor([0.0200, 0.0201, 0.0189, 0.0194, 0.0212, 0.0168, 0.0205, 0.0167], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-05-02 16:29:05,602 INFO [train.py:904] (5/8) Epoch 29, batch 950, loss[loss=0.1664, simple_loss=0.2522, pruned_loss=0.04028, over 16524.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2464, pruned_loss=0.0372, over 3291298.97 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:29:30,531 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:38,350 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7634, 2.7026, 2.4435, 2.7093, 3.0548, 2.8567, 3.2959, 3.2225], device='cuda:5'), covar=tensor([0.0177, 0.0552, 0.0607, 0.0495, 0.0376, 0.0497, 0.0287, 0.0351], device='cuda:5'), in_proj_covar=tensor([0.0233, 0.0247, 0.0236, 0.0237, 0.0247, 0.0246, 0.0243, 0.0246], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:29:47,165 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.394e+02 2.122e+02 2.524e+02 3.029e+02 7.170e+02, threshold=5.047e+02, percent-clipped=3.0 2023-05-02 16:30:14,210 INFO [train.py:904] (5/8) Epoch 29, batch 1000, loss[loss=0.1496, simple_loss=0.2316, pruned_loss=0.03378, over 15433.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.2454, pruned_loss=0.03705, over 3286498.90 frames. ], batch size: 190, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:31:02,866 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3766, 3.4895, 4.0305, 2.1027, 3.2566, 2.5283, 3.7529, 3.7188], device='cuda:5'), covar=tensor([0.0264, 0.1039, 0.0467, 0.2224, 0.0818, 0.1030, 0.0619, 0.1095], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0158, 0.0148, 0.0133, 0.0146, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 16:31:21,192 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2692, 3.7003, 3.9878, 2.1629, 3.0914, 2.5822, 3.6069, 3.8426], device='cuda:5'), covar=tensor([0.0381, 0.1028, 0.0535, 0.2235, 0.0969, 0.1020, 0.0819, 0.1180], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0158, 0.0148, 0.0133, 0.0146, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 16:31:24,212 INFO [train.py:904] (5/8) Epoch 29, batch 1050, loss[loss=0.154, simple_loss=0.2513, pruned_loss=0.0284, over 17115.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.245, pruned_loss=0.03613, over 3303846.71 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:32:05,304 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.383e+02 1.938e+02 2.240e+02 2.554e+02 8.528e+02, threshold=4.479e+02, percent-clipped=1.0 2023-05-02 16:32:19,782 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:32:31,251 INFO [train.py:904] (5/8) Epoch 29, batch 1100, loss[loss=0.1349, simple_loss=0.2152, pruned_loss=0.0273, over 15946.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2441, pruned_loss=0.03614, over 3301290.89 frames. ], batch size: 35, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:33:37,443 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:33:37,919 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 16:33:40,188 INFO [train.py:904] (5/8) Epoch 29, batch 1150, loss[loss=0.135, simple_loss=0.2295, pruned_loss=0.02029, over 17212.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2441, pruned_loss=0.03594, over 3301794.33 frames. ], batch size: 44, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:34:00,682 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4361, 5.7884, 5.5489, 5.6367, 5.2425, 5.3890, 5.1837, 5.9465], device='cuda:5'), covar=tensor([0.1410, 0.0957, 0.1051, 0.0895, 0.0910, 0.0675, 0.1371, 0.0881], device='cuda:5'), in_proj_covar=tensor([0.0725, 0.0877, 0.0717, 0.0681, 0.0555, 0.0553, 0.0738, 0.0687], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:34:07,734 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2864, 2.3578, 2.4046, 4.0109, 2.2372, 2.7031, 2.4265, 2.4919], device='cuda:5'), covar=tensor([0.1471, 0.3836, 0.3405, 0.0669, 0.4434, 0.2827, 0.3809, 0.3794], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0477, 0.0391, 0.0338, 0.0449, 0.0546, 0.0450, 0.0560], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:34:22,233 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.496e+02 2.113e+02 2.491e+02 2.895e+02 1.258e+03, threshold=4.983e+02, percent-clipped=2.0 2023-05-02 16:34:40,301 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285398.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 16:34:47,460 INFO [train.py:904] (5/8) Epoch 29, batch 1200, loss[loss=0.1623, simple_loss=0.2389, pruned_loss=0.0428, over 16797.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2428, pruned_loss=0.03571, over 3307962.33 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:35:22,684 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 16:35:56,664 INFO [train.py:904] (5/8) Epoch 29, batch 1250, loss[loss=0.15, simple_loss=0.2365, pruned_loss=0.03175, over 17202.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2437, pruned_loss=0.0364, over 3315784.56 frames. ], batch size: 44, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:36:21,923 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:36:38,701 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.107e+02 2.455e+02 3.038e+02 4.730e+02, threshold=4.911e+02, percent-clipped=0.0 2023-05-02 16:37:05,043 INFO [train.py:904] (5/8) Epoch 29, batch 1300, loss[loss=0.1714, simple_loss=0.247, pruned_loss=0.04788, over 16905.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2438, pruned_loss=0.0366, over 3319595.64 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:37:17,104 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0220, 3.7335, 4.0456, 2.2803, 4.2445, 4.2386, 3.3476, 3.3682], device='cuda:5'), covar=tensor([0.0678, 0.0270, 0.0266, 0.1119, 0.0108, 0.0228, 0.0430, 0.0428], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0141, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 16:37:28,093 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285520.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:38:13,884 INFO [train.py:904] (5/8) Epoch 29, batch 1350, loss[loss=0.1776, simple_loss=0.2708, pruned_loss=0.04222, over 17045.00 frames. ], tot_loss[loss=0.1581, simple_loss=0.244, pruned_loss=0.03613, over 3326335.86 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:38:58,364 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.007e+02 2.357e+02 2.705e+02 4.553e+02, threshold=4.714e+02, percent-clipped=0.0 2023-05-02 16:39:11,773 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285595.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:39:24,838 INFO [train.py:904] (5/8) Epoch 29, batch 1400, loss[loss=0.1641, simple_loss=0.2367, pruned_loss=0.04571, over 16409.00 frames. ], tot_loss[loss=0.1584, simple_loss=0.2441, pruned_loss=0.03637, over 3316305.01 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:39:49,684 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:19,824 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:29,947 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2180, 5.2196, 4.9815, 4.3505, 5.0395, 1.9786, 4.8100, 4.7171], device='cuda:5'), covar=tensor([0.0107, 0.0107, 0.0252, 0.0513, 0.0119, 0.3103, 0.0164, 0.0297], device='cuda:5'), in_proj_covar=tensor([0.0184, 0.0177, 0.0215, 0.0185, 0.0192, 0.0220, 0.0203, 0.0182], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:40:31,636 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:33,563 INFO [train.py:904] (5/8) Epoch 29, batch 1450, loss[loss=0.1764, simple_loss=0.2518, pruned_loss=0.0505, over 16505.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2431, pruned_loss=0.03575, over 3316255.80 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:41:13,983 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:41:16,391 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.247e+02 2.721e+02 3.292e+02 6.741e+02, threshold=5.441e+02, percent-clipped=4.0 2023-05-02 16:41:34,321 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285698.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:41:37,507 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:41:43,181 INFO [train.py:904] (5/8) Epoch 29, batch 1500, loss[loss=0.1463, simple_loss=0.2498, pruned_loss=0.0214, over 17115.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2426, pruned_loss=0.0358, over 3325453.24 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:42:40,637 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285746.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:42:51,737 INFO [train.py:904] (5/8) Epoch 29, batch 1550, loss[loss=0.1546, simple_loss=0.2298, pruned_loss=0.0397, over 16426.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2444, pruned_loss=0.0372, over 3309830.59 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:43:34,664 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 2.209e+02 2.639e+02 3.034e+02 6.721e+02, threshold=5.278e+02, percent-clipped=2.0 2023-05-02 16:43:44,928 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 16:44:00,614 INFO [train.py:904] (5/8) Epoch 29, batch 1600, loss[loss=0.1802, simple_loss=0.2735, pruned_loss=0.04342, over 16627.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2463, pruned_loss=0.038, over 3313268.55 frames. ], batch size: 62, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:44:30,955 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 16:45:09,517 INFO [train.py:904] (5/8) Epoch 29, batch 1650, loss[loss=0.164, simple_loss=0.2551, pruned_loss=0.0365, over 17084.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2476, pruned_loss=0.03861, over 3311981.93 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:45:50,989 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.202e+02 2.447e+02 3.087e+02 4.596e+02, threshold=4.895e+02, percent-clipped=0.0 2023-05-02 16:46:16,884 INFO [train.py:904] (5/8) Epoch 29, batch 1700, loss[loss=0.1668, simple_loss=0.2693, pruned_loss=0.0321, over 17129.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2493, pruned_loss=0.03892, over 3303632.70 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:46:58,557 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-05-02 16:47:24,309 INFO [train.py:904] (5/8) Epoch 29, batch 1750, loss[loss=0.1551, simple_loss=0.2569, pruned_loss=0.02671, over 17033.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2496, pruned_loss=0.03843, over 3309280.07 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:47:24,880 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7865, 2.7157, 2.9172, 4.9372, 3.8393, 4.3163, 1.7653, 3.0935], device='cuda:5'), covar=tensor([0.1550, 0.0933, 0.1157, 0.0263, 0.0263, 0.0467, 0.1792, 0.0912], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0204, 0.0205, 0.0218, 0.0210, 0.0198], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 16:47:58,308 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:48:07,062 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.447e+02 2.051e+02 2.355e+02 2.951e+02 5.834e+02, threshold=4.710e+02, percent-clipped=3.0 2023-05-02 16:48:36,753 INFO [train.py:904] (5/8) Epoch 29, batch 1800, loss[loss=0.1622, simple_loss=0.2472, pruned_loss=0.03858, over 16752.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2507, pruned_loss=0.03837, over 3315051.56 frames. ], batch size: 83, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:49:43,966 INFO [train.py:904] (5/8) Epoch 29, batch 1850, loss[loss=0.1722, simple_loss=0.2552, pruned_loss=0.04466, over 15389.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2517, pruned_loss=0.0387, over 3316128.14 frames. ], batch size: 192, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:50:28,095 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.081e+02 2.345e+02 2.768e+02 6.474e+02, threshold=4.689e+02, percent-clipped=2.0 2023-05-02 16:50:53,101 INFO [train.py:904] (5/8) Epoch 29, batch 1900, loss[loss=0.1715, simple_loss=0.2534, pruned_loss=0.04478, over 16819.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.251, pruned_loss=0.03793, over 3315120.49 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:51:18,949 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2344, 5.7895, 5.9005, 5.5949, 5.7171, 6.2432, 5.7329, 5.4672], device='cuda:5'), covar=tensor([0.0863, 0.1932, 0.2309, 0.2051, 0.2651, 0.0974, 0.1604, 0.2145], device='cuda:5'), in_proj_covar=tensor([0.0436, 0.0647, 0.0721, 0.0529, 0.0709, 0.0740, 0.0557, 0.0705], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 16:51:20,302 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286123.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:51:29,283 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:52:04,245 INFO [train.py:904] (5/8) Epoch 29, batch 1950, loss[loss=0.1465, simple_loss=0.2389, pruned_loss=0.02707, over 16804.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2511, pruned_loss=0.03759, over 3318991.69 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:52:11,862 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 16:52:40,249 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2711, 3.2753, 2.1022, 3.4737, 2.6358, 3.4430, 2.2725, 2.7509], device='cuda:5'), covar=tensor([0.0329, 0.0504, 0.1732, 0.0396, 0.0867, 0.0909, 0.1505, 0.0791], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0184, 0.0200, 0.0178, 0.0183, 0.0225, 0.0209, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 16:52:45,770 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8829, 4.1035, 2.9371, 4.7035, 3.4010, 4.5978, 2.8018, 3.4183], device='cuda:5'), covar=tensor([0.0335, 0.0400, 0.1419, 0.0303, 0.0772, 0.0546, 0.1543, 0.0699], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0184, 0.0200, 0.0178, 0.0183, 0.0225, 0.0209, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 16:52:46,910 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286184.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:52:48,713 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.778e+02 2.194e+02 2.537e+02 3.082e+02 2.053e+03, threshold=5.073e+02, percent-clipped=1.0 2023-05-02 16:52:55,325 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286190.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:53:13,093 INFO [train.py:904] (5/8) Epoch 29, batch 2000, loss[loss=0.1645, simple_loss=0.2543, pruned_loss=0.03741, over 16404.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2506, pruned_loss=0.03724, over 3325451.62 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:53:58,147 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9147, 2.9425, 2.6139, 2.7692, 3.1357, 2.9629, 3.4394, 3.3851], device='cuda:5'), covar=tensor([0.0184, 0.0446, 0.0554, 0.0504, 0.0368, 0.0458, 0.0295, 0.0327], device='cuda:5'), in_proj_covar=tensor([0.0240, 0.0253, 0.0241, 0.0241, 0.0254, 0.0252, 0.0250, 0.0251], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:54:21,798 INFO [train.py:904] (5/8) Epoch 29, batch 2050, loss[loss=0.1668, simple_loss=0.2624, pruned_loss=0.03559, over 17013.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2503, pruned_loss=0.03768, over 3322734.91 frames. ], batch size: 50, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:54,863 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:55:04,678 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.125e+02 2.436e+02 2.931e+02 6.185e+02, threshold=4.871e+02, percent-clipped=3.0 2023-05-02 16:55:18,534 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:55:30,419 INFO [train.py:904] (5/8) Epoch 29, batch 2100, loss[loss=0.1872, simple_loss=0.2648, pruned_loss=0.05474, over 16866.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2511, pruned_loss=0.03786, over 3317494.11 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:00,357 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286326.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:26,322 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:40,455 INFO [train.py:904] (5/8) Epoch 29, batch 2150, loss[loss=0.1676, simple_loss=0.25, pruned_loss=0.04261, over 16466.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2518, pruned_loss=0.03837, over 3318837.37 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:44,691 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286356.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:52,417 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 16:56:57,440 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8135, 2.8268, 2.5086, 2.6814, 3.0858, 2.9315, 3.4008, 3.3797], device='cuda:5'), covar=tensor([0.0183, 0.0495, 0.0575, 0.0520, 0.0358, 0.0420, 0.0308, 0.0295], device='cuda:5'), in_proj_covar=tensor([0.0240, 0.0254, 0.0241, 0.0242, 0.0254, 0.0252, 0.0250, 0.0252], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:57:06,940 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6098, 5.0083, 5.3080, 5.2985, 5.3003, 4.9575, 4.6486, 4.7464], device='cuda:5'), covar=tensor([0.0753, 0.0807, 0.0664, 0.0684, 0.0825, 0.0701, 0.1599, 0.0599], device='cuda:5'), in_proj_covar=tensor([0.0452, 0.0510, 0.0492, 0.0454, 0.0541, 0.0516, 0.0592, 0.0416], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 16:57:11,628 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5333, 4.3631, 4.5898, 4.7357, 4.8313, 4.4141, 4.7098, 4.8161], device='cuda:5'), covar=tensor([0.1786, 0.1381, 0.1448, 0.0729, 0.0630, 0.1098, 0.1923, 0.0916], device='cuda:5'), in_proj_covar=tensor([0.0704, 0.0864, 0.0998, 0.0876, 0.0666, 0.0691, 0.0733, 0.0845], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 16:57:24,982 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.220e+02 2.758e+02 3.117e+02 6.077e+02, threshold=5.516e+02, percent-clipped=2.0 2023-05-02 16:57:50,677 INFO [train.py:904] (5/8) Epoch 29, batch 2200, loss[loss=0.163, simple_loss=0.2434, pruned_loss=0.04134, over 16879.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.253, pruned_loss=0.03892, over 3308716.59 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:57:53,018 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:58:59,518 INFO [train.py:904] (5/8) Epoch 29, batch 2250, loss[loss=0.1745, simple_loss=0.2546, pruned_loss=0.04719, over 16551.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.253, pruned_loss=0.03903, over 3319505.78 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:59:16,119 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8857, 2.0503, 2.4545, 2.7158, 2.7642, 2.8195, 2.1517, 3.0434], device='cuda:5'), covar=tensor([0.0224, 0.0577, 0.0413, 0.0375, 0.0403, 0.0380, 0.0638, 0.0210], device='cuda:5'), in_proj_covar=tensor([0.0203, 0.0203, 0.0190, 0.0197, 0.0214, 0.0170, 0.0207, 0.0170], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-05-02 16:59:34,554 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286479.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:59:43,651 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286485.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:59:45,657 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.269e+02 2.585e+02 3.162e+02 6.397e+02, threshold=5.169e+02, percent-clipped=1.0 2023-05-02 17:00:08,678 INFO [train.py:904] (5/8) Epoch 29, batch 2300, loss[loss=0.1781, simple_loss=0.2652, pruned_loss=0.0455, over 16554.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2538, pruned_loss=0.03934, over 3305780.39 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:00:42,469 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4212, 4.4136, 4.3206, 3.8075, 4.3841, 1.8091, 4.1415, 3.8915], device='cuda:5'), covar=tensor([0.0160, 0.0129, 0.0195, 0.0281, 0.0108, 0.2861, 0.0167, 0.0279], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0178, 0.0216, 0.0187, 0.0193, 0.0220, 0.0204, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:01:16,847 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3595, 4.3439, 4.2418, 3.6768, 4.3108, 1.9097, 4.1083, 3.7584], device='cuda:5'), covar=tensor([0.0164, 0.0131, 0.0199, 0.0283, 0.0098, 0.2817, 0.0142, 0.0292], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0178, 0.0216, 0.0187, 0.0193, 0.0220, 0.0204, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:01:17,613 INFO [train.py:904] (5/8) Epoch 29, batch 2350, loss[loss=0.1861, simple_loss=0.2751, pruned_loss=0.04855, over 16418.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2538, pruned_loss=0.03962, over 3306692.75 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:01:50,931 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 17:02:03,071 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.083e+02 2.366e+02 2.716e+02 6.133e+02, threshold=4.732e+02, percent-clipped=1.0 2023-05-02 17:02:04,875 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286588.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:02:27,448 INFO [train.py:904] (5/8) Epoch 29, batch 2400, loss[loss=0.161, simple_loss=0.2566, pruned_loss=0.03272, over 17281.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2543, pruned_loss=0.03944, over 3312961.39 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 17:03:31,021 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:34,193 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286651.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:37,431 INFO [train.py:904] (5/8) Epoch 29, batch 2450, loss[loss=0.1613, simple_loss=0.2528, pruned_loss=0.03493, over 16504.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.255, pruned_loss=0.03928, over 3317612.96 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:03:42,971 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3863, 5.3906, 5.1654, 4.6272, 5.2540, 2.3600, 5.0166, 5.0972], device='cuda:5'), covar=tensor([0.0138, 0.0128, 0.0248, 0.0460, 0.0117, 0.2653, 0.0165, 0.0235], device='cuda:5'), in_proj_covar=tensor([0.0185, 0.0178, 0.0216, 0.0188, 0.0194, 0.0221, 0.0205, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:04:00,717 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:04:23,750 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.201e+02 2.590e+02 3.137e+02 8.541e+02, threshold=5.179e+02, percent-clipped=2.0 2023-05-02 17:04:35,743 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9519, 2.1851, 2.3482, 3.5467, 2.1674, 2.4128, 2.2774, 2.3231], device='cuda:5'), covar=tensor([0.1692, 0.3827, 0.3157, 0.0796, 0.4085, 0.2809, 0.3913, 0.3304], device='cuda:5'), in_proj_covar=tensor([0.0426, 0.0481, 0.0392, 0.0341, 0.0449, 0.0551, 0.0453, 0.0563], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:04:41,085 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:04:46,163 INFO [train.py:904] (5/8) Epoch 29, batch 2500, loss[loss=0.1598, simple_loss=0.259, pruned_loss=0.03033, over 17267.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2546, pruned_loss=0.03874, over 3320689.27 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:05:21,886 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-02 17:05:24,997 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286732.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:05:30,220 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0095, 4.8962, 4.8931, 4.4791, 4.6272, 4.9162, 4.7675, 4.6208], device='cuda:5'), covar=tensor([0.0695, 0.0869, 0.0321, 0.0363, 0.0908, 0.0535, 0.0446, 0.0789], device='cuda:5'), in_proj_covar=tensor([0.0324, 0.0489, 0.0380, 0.0382, 0.0375, 0.0437, 0.0260, 0.0454], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 17:05:55,674 INFO [train.py:904] (5/8) Epoch 29, batch 2550, loss[loss=0.197, simple_loss=0.2799, pruned_loss=0.05709, over 16749.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2555, pruned_loss=0.03897, over 3325425.74 frames. ], batch size: 134, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:06:19,062 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:06:32,996 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286779.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:06:40,986 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:06:44,659 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.042e+02 2.418e+02 2.906e+02 6.403e+02, threshold=4.836e+02, percent-clipped=2.0 2023-05-02 17:07:07,740 INFO [train.py:904] (5/8) Epoch 29, batch 2600, loss[loss=0.1638, simple_loss=0.2616, pruned_loss=0.033, over 17016.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2555, pruned_loss=0.03898, over 3325682.94 frames. ], batch size: 50, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:07:39,059 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286827.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:07:45,105 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286831.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:07:47,353 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:08:15,910 INFO [train.py:904] (5/8) Epoch 29, batch 2650, loss[loss=0.1639, simple_loss=0.2469, pruned_loss=0.0405, over 16021.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2549, pruned_loss=0.03812, over 3330926.92 frames. ], batch size: 35, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:08:49,640 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-05-02 17:09:00,291 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.057e+02 2.433e+02 2.859e+02 4.818e+02, threshold=4.867e+02, percent-clipped=0.0 2023-05-02 17:09:22,343 INFO [train.py:904] (5/8) Epoch 29, batch 2700, loss[loss=0.1637, simple_loss=0.249, pruned_loss=0.03918, over 16262.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.03813, over 3334916.97 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:09:40,907 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9561, 5.0004, 5.3672, 5.3752, 5.4418, 5.0771, 5.0319, 4.8317], device='cuda:5'), covar=tensor([0.0349, 0.0488, 0.0405, 0.0402, 0.0460, 0.0369, 0.0960, 0.0490], device='cuda:5'), in_proj_covar=tensor([0.0451, 0.0510, 0.0493, 0.0451, 0.0539, 0.0516, 0.0594, 0.0416], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 17:10:17,513 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286944.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:10:26,957 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286951.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:10:30,040 INFO [train.py:904] (5/8) Epoch 29, batch 2750, loss[loss=0.1798, simple_loss=0.2644, pruned_loss=0.04757, over 16456.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2558, pruned_loss=0.03829, over 3335355.68 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:17,994 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 1.930e+02 2.341e+02 2.772e+02 4.818e+02, threshold=4.681e+02, percent-clipped=0.0 2023-05-02 17:11:32,923 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:35,317 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:40,755 INFO [train.py:904] (5/8) Epoch 29, batch 2800, loss[loss=0.1804, simple_loss=0.2578, pruned_loss=0.05153, over 16896.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2555, pruned_loss=0.03812, over 3334058.63 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:56,765 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0766, 4.5987, 4.4802, 3.4075, 3.7506, 4.4838, 4.0128, 2.8563], device='cuda:5'), covar=tensor([0.0450, 0.0077, 0.0058, 0.0342, 0.0167, 0.0105, 0.0108, 0.0419], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0091, 0.0093, 0.0137, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 17:12:12,999 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:19,221 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7582, 4.7181, 4.5891, 3.5826, 4.6438, 1.7880, 4.2660, 4.2301], device='cuda:5'), covar=tensor([0.0244, 0.0184, 0.0320, 0.0706, 0.0183, 0.3625, 0.0249, 0.0414], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0179, 0.0218, 0.0188, 0.0195, 0.0222, 0.0205, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:12:42,709 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287048.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:47,017 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7643, 6.0991, 5.8240, 5.9375, 5.5512, 5.5834, 5.4714, 6.2437], device='cuda:5'), covar=tensor([0.1397, 0.1007, 0.1148, 0.0967, 0.0893, 0.0642, 0.1323, 0.0896], device='cuda:5'), in_proj_covar=tensor([0.0733, 0.0888, 0.0727, 0.0688, 0.0561, 0.0559, 0.0745, 0.0698], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:12:50,055 INFO [train.py:904] (5/8) Epoch 29, batch 2850, loss[loss=0.1662, simple_loss=0.2494, pruned_loss=0.04144, over 16714.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2547, pruned_loss=0.03804, over 3326995.81 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:13:39,986 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.226e+02 2.685e+02 3.371e+02 7.491e+02, threshold=5.370e+02, percent-clipped=5.0 2023-05-02 17:14:00,585 INFO [train.py:904] (5/8) Epoch 29, batch 2900, loss[loss=0.1604, simple_loss=0.2515, pruned_loss=0.03469, over 17047.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2532, pruned_loss=0.03778, over 3322988.52 frames. ], batch size: 53, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:14:23,324 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0505, 4.0162, 3.9542, 3.3563, 3.9692, 1.8928, 3.7939, 3.4313], device='cuda:5'), covar=tensor([0.0165, 0.0179, 0.0219, 0.0285, 0.0124, 0.2971, 0.0167, 0.0307], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0179, 0.0217, 0.0189, 0.0195, 0.0222, 0.0205, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:14:31,460 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:15:07,443 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6937, 2.7186, 2.6503, 4.9380, 3.9159, 4.3070, 1.6999, 3.0346], device='cuda:5'), covar=tensor([0.1490, 0.0887, 0.1277, 0.0236, 0.0247, 0.0425, 0.1710, 0.0851], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0207, 0.0207, 0.0221, 0.0211, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 17:15:10,999 INFO [train.py:904] (5/8) Epoch 29, batch 2950, loss[loss=0.1707, simple_loss=0.2581, pruned_loss=0.04167, over 16798.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2528, pruned_loss=0.03872, over 3323994.47 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:15:59,396 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.092e+02 2.416e+02 3.171e+02 5.496e+02, threshold=4.831e+02, percent-clipped=1.0 2023-05-02 17:16:12,731 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 17:16:20,172 INFO [train.py:904] (5/8) Epoch 29, batch 3000, loss[loss=0.15, simple_loss=0.2511, pruned_loss=0.02439, over 17131.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2534, pruned_loss=0.03911, over 3320560.17 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:16:20,172 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 17:16:28,749 INFO [train.py:938] (5/8) Epoch 29, validation: loss=0.1336, simple_loss=0.2385, pruned_loss=0.01438, over 944034.00 frames. 2023-05-02 17:16:28,750 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 17:17:14,893 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7783, 4.6546, 4.6610, 4.3330, 4.3671, 4.7122, 4.4731, 4.4869], device='cuda:5'), covar=tensor([0.0852, 0.1165, 0.0379, 0.0387, 0.0945, 0.0696, 0.0627, 0.0749], device='cuda:5'), in_proj_covar=tensor([0.0326, 0.0491, 0.0382, 0.0384, 0.0378, 0.0441, 0.0261, 0.0456], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 17:17:25,972 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:17:40,004 INFO [train.py:904] (5/8) Epoch 29, batch 3050, loss[loss=0.1709, simple_loss=0.2558, pruned_loss=0.04302, over 16476.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2527, pruned_loss=0.03895, over 3312521.66 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:18:11,188 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 2023-05-02 17:18:29,495 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.177e+02 2.556e+02 3.004e+02 5.914e+02, threshold=5.112e+02, percent-clipped=3.0 2023-05-02 17:18:34,363 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287292.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:18:49,719 INFO [train.py:904] (5/8) Epoch 29, batch 3100, loss[loss=0.1369, simple_loss=0.2212, pruned_loss=0.0263, over 16758.00 frames. ], tot_loss[loss=0.165, simple_loss=0.252, pruned_loss=0.03896, over 3320252.16 frames. ], batch size: 39, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:19:22,342 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:00,331 INFO [train.py:904] (5/8) Epoch 29, batch 3150, loss[loss=0.208, simple_loss=0.2889, pruned_loss=0.06352, over 12415.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2517, pruned_loss=0.03905, over 3312564.85 frames. ], batch size: 246, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:20:30,290 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:38,272 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1583, 4.0103, 4.4298, 2.4658, 4.6518, 4.6940, 3.3953, 3.7056], device='cuda:5'), covar=tensor([0.0722, 0.0272, 0.0229, 0.1111, 0.0084, 0.0202, 0.0459, 0.0402], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0088, 0.0133, 0.0131, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 17:20:49,321 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.202e+02 2.547e+02 3.057e+02 5.331e+02, threshold=5.094e+02, percent-clipped=1.0 2023-05-02 17:21:10,114 INFO [train.py:904] (5/8) Epoch 29, batch 3200, loss[loss=0.179, simple_loss=0.2756, pruned_loss=0.04125, over 16723.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2513, pruned_loss=0.03843, over 3312822.56 frames. ], batch size: 57, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:21:20,068 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:21:40,383 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:18,324 INFO [train.py:904] (5/8) Epoch 29, batch 3250, loss[loss=0.1651, simple_loss=0.2479, pruned_loss=0.04112, over 16830.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2507, pruned_loss=0.03849, over 3309291.30 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:22:43,587 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:45,746 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287474.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:23:05,923 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.139e+02 2.558e+02 3.030e+02 4.692e+02, threshold=5.116e+02, percent-clipped=0.0 2023-05-02 17:23:12,604 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1468, 5.5751, 5.7551, 5.3981, 5.4941, 6.1147, 5.5713, 5.3018], device='cuda:5'), covar=tensor([0.1004, 0.1874, 0.2350, 0.2377, 0.2886, 0.1093, 0.1473, 0.2339], device='cuda:5'), in_proj_covar=tensor([0.0438, 0.0652, 0.0725, 0.0532, 0.0712, 0.0743, 0.0560, 0.0705], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 17:23:19,651 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7278, 3.4424, 3.8408, 2.1261, 3.9096, 3.9214, 3.1691, 2.9145], device='cuda:5'), covar=tensor([0.0753, 0.0279, 0.0193, 0.1140, 0.0128, 0.0236, 0.0425, 0.0502], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0088, 0.0133, 0.0131, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 17:23:26,858 INFO [train.py:904] (5/8) Epoch 29, batch 3300, loss[loss=0.1758, simple_loss=0.2699, pruned_loss=0.04085, over 17105.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.252, pruned_loss=0.03882, over 3316764.24 frames. ], batch size: 47, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:24:11,466 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287536.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:24:29,442 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 17:24:34,316 INFO [train.py:904] (5/8) Epoch 29, batch 3350, loss[loss=0.1621, simple_loss=0.2614, pruned_loss=0.03141, over 17110.00 frames. ], tot_loss[loss=0.165, simple_loss=0.253, pruned_loss=0.03848, over 3318799.32 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:24:46,541 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287562.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:24:55,657 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 17:25:00,570 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7581, 2.3367, 2.4601, 3.4169, 2.5562, 3.5942, 1.7338, 2.7403], device='cuda:5'), covar=tensor([0.1414, 0.0842, 0.1194, 0.0228, 0.0127, 0.0378, 0.1612, 0.0907], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0182, 0.0202, 0.0209, 0.0208, 0.0221, 0.0211, 0.0200], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 17:25:14,586 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287583.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:21,672 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 2.115e+02 2.472e+02 3.004e+02 5.841e+02, threshold=4.943e+02, percent-clipped=1.0 2023-05-02 17:25:33,265 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:43,308 INFO [train.py:904] (5/8) Epoch 29, batch 3400, loss[loss=0.1622, simple_loss=0.2406, pruned_loss=0.04194, over 16786.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2535, pruned_loss=0.03899, over 3322457.64 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:25:45,199 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-02 17:26:10,447 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:16,696 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5900, 4.4864, 4.4999, 4.1986, 4.2521, 4.5568, 4.3529, 4.2907], device='cuda:5'), covar=tensor([0.0760, 0.0950, 0.0367, 0.0353, 0.0854, 0.0598, 0.0598, 0.0722], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0496, 0.0385, 0.0387, 0.0380, 0.0444, 0.0263, 0.0460], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 17:26:40,251 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287644.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:53,541 INFO [train.py:904] (5/8) Epoch 29, batch 3450, loss[loss=0.1501, simple_loss=0.2307, pruned_loss=0.03472, over 16838.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2516, pruned_loss=0.03839, over 3322023.83 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:27:43,292 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 1.986e+02 2.391e+02 2.898e+02 4.066e+02, threshold=4.783e+02, percent-clipped=0.0 2023-05-02 17:28:04,238 INFO [train.py:904] (5/8) Epoch 29, batch 3500, loss[loss=0.1535, simple_loss=0.2522, pruned_loss=0.02742, over 17127.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2499, pruned_loss=0.03755, over 3330897.28 frames. ], batch size: 48, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:28:17,566 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5784, 3.8880, 4.0419, 2.8922, 3.6158, 4.1074, 3.7682, 2.3704], device='cuda:5'), covar=tensor([0.0593, 0.0433, 0.0077, 0.0431, 0.0144, 0.0123, 0.0125, 0.0576], device='cuda:5'), in_proj_covar=tensor([0.0142, 0.0093, 0.0094, 0.0140, 0.0106, 0.0120, 0.0102, 0.0135], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 17:28:26,811 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:14,562 INFO [train.py:904] (5/8) Epoch 29, batch 3550, loss[loss=0.151, simple_loss=0.2444, pruned_loss=0.02884, over 17013.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2494, pruned_loss=0.03728, over 3328133.66 frames. ], batch size: 50, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:29:24,355 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9749, 4.2206, 4.0587, 4.1225, 3.8188, 3.8742, 3.8739, 4.2072], device='cuda:5'), covar=tensor([0.1135, 0.0906, 0.1024, 0.0775, 0.0738, 0.1576, 0.1003, 0.0983], device='cuda:5'), in_proj_covar=tensor([0.0742, 0.0894, 0.0733, 0.0694, 0.0568, 0.0565, 0.0751, 0.0702], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:29:33,159 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:42,524 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:54,008 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287781.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:29:58,270 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4027, 5.3543, 5.2312, 4.6873, 4.8641, 5.2709, 5.2139, 4.8642], device='cuda:5'), covar=tensor([0.0606, 0.0561, 0.0336, 0.0403, 0.1127, 0.0528, 0.0291, 0.0836], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0499, 0.0387, 0.0389, 0.0383, 0.0447, 0.0265, 0.0462], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 17:30:07,188 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.118e+02 2.537e+02 3.010e+02 5.172e+02, threshold=5.074e+02, percent-clipped=1.0 2023-05-02 17:30:26,883 INFO [train.py:904] (5/8) Epoch 29, batch 3600, loss[loss=0.1494, simple_loss=0.2237, pruned_loss=0.03752, over 15789.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2483, pruned_loss=0.03689, over 3312315.69 frames. ], batch size: 35, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:31:11,626 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287834.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:31:40,153 INFO [train.py:904] (5/8) Epoch 29, batch 3650, loss[loss=0.1754, simple_loss=0.243, pruned_loss=0.05388, over 16794.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.248, pruned_loss=0.0376, over 3293011.71 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:32:08,490 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-05-02 17:32:35,504 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.232e+02 2.562e+02 3.214e+02 7.635e+02, threshold=5.123e+02, percent-clipped=1.0 2023-05-02 17:32:38,222 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287892.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:32:38,335 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8533, 4.7476, 4.9422, 5.0750, 5.2342, 4.7382, 5.2068, 5.2790], device='cuda:5'), covar=tensor([0.1706, 0.1079, 0.1458, 0.0680, 0.0478, 0.0854, 0.0608, 0.0527], device='cuda:5'), in_proj_covar=tensor([0.0712, 0.0869, 0.1002, 0.0882, 0.0673, 0.0697, 0.0736, 0.0852], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:32:38,382 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2449, 5.2235, 4.9405, 4.3235, 5.1706, 2.0439, 4.8766, 4.6856], device='cuda:5'), covar=tensor([0.0090, 0.0076, 0.0211, 0.0390, 0.0079, 0.2868, 0.0116, 0.0240], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0182, 0.0221, 0.0192, 0.0198, 0.0224, 0.0209, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:32:57,097 INFO [train.py:904] (5/8) Epoch 29, batch 3700, loss[loss=0.1748, simple_loss=0.2492, pruned_loss=0.05017, over 16779.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2469, pruned_loss=0.039, over 3273662.01 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:33:18,820 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:33:31,838 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-05-02 17:33:48,747 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:34:09,994 INFO [train.py:904] (5/8) Epoch 29, batch 3750, loss[loss=0.1699, simple_loss=0.2624, pruned_loss=0.03873, over 16594.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2474, pruned_loss=0.04035, over 3268337.04 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:34:46,616 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7470, 4.6849, 4.6433, 4.1145, 4.7182, 1.9599, 4.4781, 4.2502], device='cuda:5'), covar=tensor([0.0177, 0.0156, 0.0212, 0.0360, 0.0120, 0.2846, 0.0166, 0.0276], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0181, 0.0220, 0.0192, 0.0198, 0.0224, 0.0208, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:35:05,545 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.186e+02 2.475e+02 3.056e+02 8.739e+02, threshold=4.949e+02, percent-clipped=4.0 2023-05-02 17:35:29,316 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288003.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:35:30,048 INFO [train.py:904] (5/8) Epoch 29, batch 3800, loss[loss=0.1817, simple_loss=0.2673, pruned_loss=0.04806, over 17011.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2487, pruned_loss=0.04137, over 3268291.02 frames. ], batch size: 41, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:15,831 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1144, 3.1208, 2.1398, 3.2677, 2.5048, 3.3619, 2.2316, 2.6784], device='cuda:5'), covar=tensor([0.0343, 0.0447, 0.1496, 0.0316, 0.0810, 0.0620, 0.1426, 0.0696], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0185, 0.0200, 0.0180, 0.0183, 0.0227, 0.0208, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 17:36:40,385 INFO [train.py:904] (5/8) Epoch 29, batch 3850, loss[loss=0.1575, simple_loss=0.2431, pruned_loss=0.03602, over 16536.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2487, pruned_loss=0.04164, over 3277799.29 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:55,439 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288064.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:36:59,623 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288067.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:37:12,442 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288076.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:37:30,812 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.131e+02 2.477e+02 2.774e+02 5.510e+02, threshold=4.954e+02, percent-clipped=1.0 2023-05-02 17:37:50,143 INFO [train.py:904] (5/8) Epoch 29, batch 3900, loss[loss=0.1667, simple_loss=0.2397, pruned_loss=0.04685, over 16788.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2488, pruned_loss=0.04243, over 3276586.69 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:37:53,634 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288106.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:38:06,534 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:38:24,114 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 17:38:27,304 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:38:35,887 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7307, 2.6560, 2.0062, 2.7399, 2.1893, 2.8853, 2.1649, 2.4089], device='cuda:5'), covar=tensor([0.0321, 0.0353, 0.1283, 0.0232, 0.0699, 0.0350, 0.1195, 0.0664], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0184, 0.0199, 0.0179, 0.0183, 0.0226, 0.0207, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 17:39:00,613 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288152.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:39:03,102 INFO [train.py:904] (5/8) Epoch 29, batch 3950, loss[loss=0.1833, simple_loss=0.2635, pruned_loss=0.05151, over 16407.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.249, pruned_loss=0.04327, over 3276902.88 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:39:22,349 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288167.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:39:55,368 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.270e+02 2.585e+02 3.079e+02 6.589e+02, threshold=5.170e+02, percent-clipped=2.0 2023-05-02 17:39:58,735 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:40:16,248 INFO [train.py:904] (5/8) Epoch 29, batch 4000, loss[loss=0.1895, simple_loss=0.2728, pruned_loss=0.05311, over 16331.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2494, pruned_loss=0.04375, over 3287172.12 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:40:30,288 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288213.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:40:37,390 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:40:41,744 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4021, 3.5715, 3.6618, 3.6303, 3.6503, 3.4935, 3.5310, 3.5331], device='cuda:5'), covar=tensor([0.0402, 0.0618, 0.0479, 0.0464, 0.0571, 0.0534, 0.0726, 0.0577], device='cuda:5'), in_proj_covar=tensor([0.0452, 0.0512, 0.0494, 0.0454, 0.0542, 0.0520, 0.0598, 0.0419], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 17:40:54,387 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0896, 3.1298, 3.4887, 2.1901, 3.0283, 2.2567, 3.5598, 3.5878], device='cuda:5'), covar=tensor([0.0250, 0.0968, 0.0623, 0.2142, 0.0904, 0.1104, 0.0588, 0.0835], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0174, 0.0172, 0.0159, 0.0149, 0.0134, 0.0147, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 17:41:07,404 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288239.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:08,463 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:29,242 INFO [train.py:904] (5/8) Epoch 29, batch 4050, loss[loss=0.1562, simple_loss=0.2515, pruned_loss=0.03047, over 16852.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2502, pruned_loss=0.0432, over 3281253.61 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:41:47,213 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288266.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:17,244 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:21,067 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.858e+02 2.084e+02 2.415e+02 5.689e+02, threshold=4.167e+02, percent-clipped=1.0 2023-05-02 17:42:38,831 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9999, 4.2357, 4.0869, 4.1314, 3.8249, 3.8473, 3.8655, 4.2447], device='cuda:5'), covar=tensor([0.1133, 0.0823, 0.0882, 0.0744, 0.0732, 0.1675, 0.0929, 0.0886], device='cuda:5'), in_proj_covar=tensor([0.0741, 0.0891, 0.0730, 0.0693, 0.0568, 0.0565, 0.0750, 0.0701], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:42:42,073 INFO [train.py:904] (5/8) Epoch 29, batch 4100, loss[loss=0.1804, simple_loss=0.2679, pruned_loss=0.04639, over 17077.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2513, pruned_loss=0.04264, over 3277267.64 frames. ], batch size: 55, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:43:17,889 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:31,760 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5126, 4.5933, 4.8105, 4.7853, 4.8591, 4.5739, 4.5426, 4.3932], device='cuda:5'), covar=tensor([0.0333, 0.0488, 0.0367, 0.0426, 0.0426, 0.0347, 0.0854, 0.0528], device='cuda:5'), in_proj_covar=tensor([0.0450, 0.0510, 0.0492, 0.0452, 0.0540, 0.0518, 0.0596, 0.0418], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 17:43:45,893 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:59,220 INFO [train.py:904] (5/8) Epoch 29, batch 4150, loss[loss=0.1905, simple_loss=0.2775, pruned_loss=0.05173, over 16532.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2582, pruned_loss=0.04473, over 3253010.59 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:44:07,906 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:33,394 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:51,050 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:52,909 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.295e+02 2.546e+02 3.002e+02 5.285e+02, threshold=5.091e+02, percent-clipped=5.0 2023-05-02 17:45:02,596 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288396.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:02,674 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.5874, 2.4007, 2.3079, 3.4434, 2.3853, 3.6222, 1.5840, 2.7128], device='cuda:5'), covar=tensor([0.1436, 0.0879, 0.1342, 0.0182, 0.0161, 0.0370, 0.1747, 0.0867], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0181, 0.0201, 0.0207, 0.0207, 0.0219, 0.0211, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 17:45:13,940 INFO [train.py:904] (5/8) Epoch 29, batch 4200, loss[loss=0.2138, simple_loss=0.2917, pruned_loss=0.06795, over 11245.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2653, pruned_loss=0.04673, over 3208239.62 frames. ], batch size: 248, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:45:18,719 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:45,981 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:53,548 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288429.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:46:30,287 INFO [train.py:904] (5/8) Epoch 29, batch 4250, loss[loss=0.1628, simple_loss=0.263, pruned_loss=0.03127, over 16778.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2687, pruned_loss=0.04631, over 3185705.97 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:46:36,689 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:46:43,218 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288462.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:47:05,782 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:47:24,434 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.222e+02 2.492e+02 3.034e+02 5.623e+02, threshold=4.984e+02, percent-clipped=1.0 2023-05-02 17:47:45,993 INFO [train.py:904] (5/8) Epoch 29, batch 4300, loss[loss=0.1976, simple_loss=0.2881, pruned_loss=0.05356, over 16655.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2697, pruned_loss=0.04537, over 3181150.52 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:47:51,124 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288508.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:48:59,355 INFO [train.py:904] (5/8) Epoch 29, batch 4350, loss[loss=0.1726, simple_loss=0.2672, pruned_loss=0.03901, over 16746.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.273, pruned_loss=0.04641, over 3180953.08 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:49:53,413 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.193e+02 2.447e+02 2.913e+02 5.150e+02, threshold=4.894e+02, percent-clipped=1.0 2023-05-02 17:50:13,843 INFO [train.py:904] (5/8) Epoch 29, batch 4400, loss[loss=0.1868, simple_loss=0.2872, pruned_loss=0.04322, over 16842.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2753, pruned_loss=0.04747, over 3174797.20 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:50:22,371 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:51:26,313 INFO [train.py:904] (5/8) Epoch 29, batch 4450, loss[loss=0.1981, simple_loss=0.29, pruned_loss=0.05316, over 16755.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2791, pruned_loss=0.04891, over 3189071.80 frames. ], batch size: 89, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:51:34,143 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:51:49,654 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288670.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:52:08,450 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:18,790 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 1.892e+02 2.239e+02 2.604e+02 5.228e+02, threshold=4.478e+02, percent-clipped=1.0 2023-05-02 17:52:34,492 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288701.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:37,599 INFO [train.py:904] (5/8) Epoch 29, batch 4500, loss[loss=0.1774, simple_loss=0.2553, pruned_loss=0.04976, over 17056.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2797, pruned_loss=0.04988, over 3213651.62 frames. ], batch size: 53, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:52:42,949 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:01,886 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6275, 3.7506, 3.9788, 2.3567, 3.4865, 2.4175, 3.8638, 4.0811], device='cuda:5'), covar=tensor([0.0192, 0.0767, 0.0507, 0.2198, 0.0739, 0.1027, 0.0560, 0.0912], device='cuda:5'), in_proj_covar=tensor([0.0163, 0.0173, 0.0172, 0.0158, 0.0149, 0.0134, 0.0147, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 17:53:49,140 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:51,503 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 17:53:51,863 INFO [train.py:904] (5/8) Epoch 29, batch 4550, loss[loss=0.1755, simple_loss=0.2691, pruned_loss=0.04092, over 16842.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2802, pruned_loss=0.05063, over 3217419.51 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:54:03,018 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288762.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:54:14,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-05-02 17:54:40,991 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 17:54:44,114 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 1.747e+02 2.008e+02 2.421e+02 6.234e+02, threshold=4.017e+02, percent-clipped=2.0 2023-05-02 17:54:54,468 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:54:58,828 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5481, 4.6509, 4.8677, 4.5314, 4.6783, 5.2430, 4.6852, 4.3387], device='cuda:5'), covar=tensor([0.1329, 0.1934, 0.2087, 0.2067, 0.2485, 0.0972, 0.1586, 0.2454], device='cuda:5'), in_proj_covar=tensor([0.0431, 0.0640, 0.0706, 0.0519, 0.0696, 0.0728, 0.0548, 0.0690], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 17:55:04,704 INFO [train.py:904] (5/8) Epoch 29, batch 4600, loss[loss=0.197, simple_loss=0.2821, pruned_loss=0.05599, over 17089.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2809, pruned_loss=0.0508, over 3227538.18 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:55:11,584 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288808.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:55:14,480 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288810.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:55:14,732 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9163, 2.2060, 2.1798, 3.3974, 2.0424, 2.4761, 2.2992, 2.2932], device='cuda:5'), covar=tensor([0.1571, 0.3204, 0.3110, 0.0738, 0.4380, 0.2423, 0.3231, 0.3543], device='cuda:5'), in_proj_covar=tensor([0.0424, 0.0478, 0.0388, 0.0339, 0.0447, 0.0550, 0.0450, 0.0560], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:56:18,400 INFO [train.py:904] (5/8) Epoch 29, batch 4650, loss[loss=0.1781, simple_loss=0.2722, pruned_loss=0.04196, over 16739.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2803, pruned_loss=0.05088, over 3213679.52 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:56:19,940 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0784, 4.0290, 3.8657, 2.9014, 3.8729, 1.7235, 3.5836, 3.1285], device='cuda:5'), covar=tensor([0.0115, 0.0105, 0.0219, 0.0383, 0.0095, 0.3535, 0.0133, 0.0396], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0179, 0.0219, 0.0190, 0.0196, 0.0222, 0.0206, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 17:56:21,002 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288856.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:56:24,816 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:57:10,670 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.194e+02 1.769e+02 2.025e+02 2.417e+02 4.191e+02, threshold=4.050e+02, percent-clipped=1.0 2023-05-02 17:57:29,737 INFO [train.py:904] (5/8) Epoch 29, batch 4700, loss[loss=0.204, simple_loss=0.283, pruned_loss=0.06249, over 11550.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2779, pruned_loss=0.04998, over 3199127.38 frames. ], batch size: 248, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:57:37,809 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 17:58:21,515 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 17:58:21,680 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 17:58:41,601 INFO [train.py:904] (5/8) Epoch 29, batch 4750, loss[loss=0.1514, simple_loss=0.2441, pruned_loss=0.02934, over 17278.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.273, pruned_loss=0.04769, over 3206728.76 frames. ], batch size: 52, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:58:57,739 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288965.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:59:23,643 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:35,277 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 1.780e+02 1.987e+02 2.347e+02 4.131e+02, threshold=3.973e+02, percent-clipped=1.0 2023-05-02 17:59:50,967 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289001.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:54,503 INFO [train.py:904] (5/8) Epoch 29, batch 4800, loss[loss=0.1967, simple_loss=0.2817, pruned_loss=0.05583, over 12079.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.27, pruned_loss=0.04613, over 3194563.67 frames. ], batch size: 246, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:00:37,065 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:00,432 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289046.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:01:05,208 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:09,011 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:12,128 INFO [train.py:904] (5/8) Epoch 29, batch 4850, loss[loss=0.1785, simple_loss=0.2793, pruned_loss=0.03884, over 16404.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2708, pruned_loss=0.04506, over 3190465.78 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:08,680 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 1.829e+02 2.187e+02 2.631e+02 3.612e+02, threshold=4.373e+02, percent-clipped=0.0 2023-05-02 18:02:22,995 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289100.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:02:28,904 INFO [train.py:904] (5/8) Epoch 29, batch 4900, loss[loss=0.1512, simple_loss=0.2447, pruned_loss=0.02885, over 16847.00 frames. ], tot_loss[loss=0.178, simple_loss=0.269, pruned_loss=0.04345, over 3195139.53 frames. ], batch size: 96, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:33,611 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289107.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 18:03:16,452 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289136.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:03:42,159 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:03:43,052 INFO [train.py:904] (5/8) Epoch 29, batch 4950, loss[loss=0.1694, simple_loss=0.2656, pruned_loss=0.03661, over 16797.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2686, pruned_loss=0.04322, over 3176318.12 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:04:38,082 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.040e+02 2.385e+02 2.853e+02 5.286e+02, threshold=4.771e+02, percent-clipped=2.0 2023-05-02 18:04:40,852 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0595, 5.0323, 4.8788, 4.0137, 5.0030, 1.7936, 4.6605, 4.5321], device='cuda:5'), covar=tensor([0.0102, 0.0102, 0.0180, 0.0527, 0.0106, 0.3011, 0.0147, 0.0276], device='cuda:5'), in_proj_covar=tensor([0.0184, 0.0178, 0.0217, 0.0188, 0.0194, 0.0220, 0.0205, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:04:46,908 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289197.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:04:53,364 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 18:04:56,969 INFO [train.py:904] (5/8) Epoch 29, batch 5000, loss[loss=0.2059, simple_loss=0.2888, pruned_loss=0.06151, over 12323.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2699, pruned_loss=0.04336, over 3159445.35 frames. ], batch size: 246, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:04:58,261 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8149, 1.4145, 1.7167, 1.6651, 1.8994, 1.9962, 1.6440, 1.9255], device='cuda:5'), covar=tensor([0.0290, 0.0468, 0.0263, 0.0403, 0.0329, 0.0226, 0.0509, 0.0155], device='cuda:5'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0195, 0.0212, 0.0169, 0.0206, 0.0169], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-05-02 18:06:09,975 INFO [train.py:904] (5/8) Epoch 29, batch 5050, loss[loss=0.1664, simple_loss=0.2632, pruned_loss=0.03475, over 16769.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2708, pruned_loss=0.0431, over 3173837.90 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:06:18,604 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9273, 2.9087, 2.6330, 5.1237, 3.6489, 4.0861, 1.7332, 3.0670], device='cuda:5'), covar=tensor([0.1398, 0.0946, 0.1484, 0.0140, 0.0387, 0.0517, 0.1834, 0.0989], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0181, 0.0200, 0.0204, 0.0206, 0.0217, 0.0210, 0.0199], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 18:06:27,018 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289265.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:07:03,095 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 1.941e+02 2.202e+02 2.540e+02 3.761e+02, threshold=4.404e+02, percent-clipped=0.0 2023-05-02 18:07:22,122 INFO [train.py:904] (5/8) Epoch 29, batch 5100, loss[loss=0.1521, simple_loss=0.2475, pruned_loss=0.02834, over 16495.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2693, pruned_loss=0.04253, over 3170977.30 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:07:24,280 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3391, 4.3217, 4.2353, 3.3956, 4.2610, 1.6070, 3.9964, 3.7352], device='cuda:5'), covar=tensor([0.0133, 0.0132, 0.0179, 0.0489, 0.0113, 0.3302, 0.0151, 0.0352], device='cuda:5'), in_proj_covar=tensor([0.0184, 0.0179, 0.0217, 0.0188, 0.0194, 0.0220, 0.0205, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:07:34,913 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:08:36,374 INFO [train.py:904] (5/8) Epoch 29, batch 5150, loss[loss=0.1909, simple_loss=0.2907, pruned_loss=0.04557, over 16359.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2694, pruned_loss=0.04162, over 3196982.45 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:08:45,301 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 18:09:29,042 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.894e+02 2.237e+02 2.654e+02 4.017e+02, threshold=4.473e+02, percent-clipped=0.0 2023-05-02 18:09:45,678 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289402.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:09:47,653 INFO [train.py:904] (5/8) Epoch 29, batch 5200, loss[loss=0.1706, simple_loss=0.2633, pruned_loss=0.03893, over 16685.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2675, pruned_loss=0.04083, over 3207148.91 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:10:46,589 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2700, 4.1078, 4.3425, 4.4756, 4.6154, 4.2454, 4.5351, 4.6442], device='cuda:5'), covar=tensor([0.1806, 0.1228, 0.1523, 0.0736, 0.0531, 0.1152, 0.0851, 0.0685], device='cuda:5'), in_proj_covar=tensor([0.0687, 0.0837, 0.0966, 0.0853, 0.0649, 0.0671, 0.0709, 0.0824], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:10:59,671 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:11:00,521 INFO [train.py:904] (5/8) Epoch 29, batch 5250, loss[loss=0.1634, simple_loss=0.2423, pruned_loss=0.04221, over 16869.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2651, pruned_loss=0.04064, over 3200219.92 frames. ], batch size: 42, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:11:34,817 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4261, 4.3877, 4.3039, 3.4725, 4.3361, 1.6547, 4.0701, 3.8665], device='cuda:5'), covar=tensor([0.0111, 0.0122, 0.0189, 0.0413, 0.0114, 0.3174, 0.0143, 0.0307], device='cuda:5'), in_proj_covar=tensor([0.0183, 0.0178, 0.0216, 0.0187, 0.0193, 0.0220, 0.0204, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:11:55,944 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 1.894e+02 2.157e+02 2.543e+02 4.571e+02, threshold=4.315e+02, percent-clipped=1.0 2023-05-02 18:11:58,123 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:10,715 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:14,377 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289503.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:15,674 INFO [train.py:904] (5/8) Epoch 29, batch 5300, loss[loss=0.1621, simple_loss=0.2502, pruned_loss=0.03695, over 16786.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2613, pruned_loss=0.03969, over 3206929.78 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:13:26,902 INFO [train.py:904] (5/8) Epoch 29, batch 5350, loss[loss=0.1891, simple_loss=0.2799, pruned_loss=0.04913, over 17024.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.26, pruned_loss=0.0392, over 3216233.92 frames. ], batch size: 55, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:13:42,104 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289564.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:13:53,325 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9000, 5.1680, 4.9505, 5.0232, 4.7240, 4.7143, 4.5947, 5.2839], device='cuda:5'), covar=tensor([0.1292, 0.0891, 0.0963, 0.0826, 0.0841, 0.0954, 0.1188, 0.0853], device='cuda:5'), in_proj_covar=tensor([0.0727, 0.0874, 0.0718, 0.0676, 0.0557, 0.0554, 0.0731, 0.0687], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:14:21,404 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.853e+02 2.169e+02 2.585e+02 3.843e+02, threshold=4.338e+02, percent-clipped=0.0 2023-05-02 18:14:41,096 INFO [train.py:904] (5/8) Epoch 29, batch 5400, loss[loss=0.1835, simple_loss=0.2726, pruned_loss=0.04717, over 16805.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2626, pruned_loss=0.0399, over 3201017.33 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:15:23,734 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 18:15:57,439 INFO [train.py:904] (5/8) Epoch 29, batch 5450, loss[loss=0.1873, simple_loss=0.2818, pruned_loss=0.04641, over 16492.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2643, pruned_loss=0.04048, over 3179814.88 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:16:38,824 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4908, 4.2079, 4.2192, 2.8551, 3.6211, 4.2336, 3.6862, 2.3336], device='cuda:5'), covar=tensor([0.0607, 0.0063, 0.0060, 0.0437, 0.0143, 0.0107, 0.0122, 0.0526], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0136, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 18:16:53,639 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.272e+02 2.915e+02 3.734e+02 7.781e+02, threshold=5.831e+02, percent-clipped=14.0 2023-05-02 18:16:57,249 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1045, 2.8712, 3.1134, 1.8410, 3.2433, 3.2758, 2.6255, 2.5701], device='cuda:5'), covar=tensor([0.0888, 0.0313, 0.0229, 0.1177, 0.0117, 0.0234, 0.0501, 0.0483], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0087, 0.0131, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 18:17:12,076 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289702.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:17:14,654 INFO [train.py:904] (5/8) Epoch 29, batch 5500, loss[loss=0.253, simple_loss=0.3262, pruned_loss=0.08989, over 11886.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2713, pruned_loss=0.04461, over 3142121.04 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:17:28,372 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0455, 4.1198, 3.9562, 3.6658, 3.7069, 4.0396, 3.7161, 3.8530], device='cuda:5'), covar=tensor([0.0600, 0.0631, 0.0307, 0.0282, 0.0677, 0.0481, 0.1114, 0.0538], device='cuda:5'), in_proj_covar=tensor([0.0313, 0.0475, 0.0368, 0.0369, 0.0364, 0.0425, 0.0252, 0.0439], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:18:26,457 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289750.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:18:31,709 INFO [train.py:904] (5/8) Epoch 29, batch 5550, loss[loss=0.1792, simple_loss=0.2562, pruned_loss=0.05111, over 16310.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.278, pruned_loss=0.04923, over 3121373.31 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:19:30,746 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.953e+02 3.616e+02 4.343e+02 9.960e+02, threshold=7.231e+02, percent-clipped=12.0 2023-05-02 18:19:34,144 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289792.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:19:53,194 INFO [train.py:904] (5/8) Epoch 29, batch 5600, loss[loss=0.2101, simple_loss=0.2967, pruned_loss=0.06176, over 16808.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2829, pruned_loss=0.05324, over 3087570.97 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:20:47,986 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9742, 3.8914, 4.0116, 4.1299, 4.2412, 3.8450, 4.1908, 4.2683], device='cuda:5'), covar=tensor([0.1776, 0.1184, 0.1434, 0.0777, 0.0657, 0.1615, 0.0946, 0.0840], device='cuda:5'), in_proj_covar=tensor([0.0688, 0.0838, 0.0966, 0.0853, 0.0650, 0.0672, 0.0708, 0.0825], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:20:54,482 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289840.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:21:16,131 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1511, 5.1953, 5.0323, 4.6745, 4.6856, 5.0772, 5.0105, 4.8195], device='cuda:5'), covar=tensor([0.0622, 0.0442, 0.0285, 0.0310, 0.1017, 0.0464, 0.0329, 0.0633], device='cuda:5'), in_proj_covar=tensor([0.0311, 0.0472, 0.0365, 0.0366, 0.0362, 0.0423, 0.0251, 0.0437], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:21:16,889 INFO [train.py:904] (5/8) Epoch 29, batch 5650, loss[loss=0.1953, simple_loss=0.2824, pruned_loss=0.05409, over 16685.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2879, pruned_loss=0.05772, over 3045407.58 frames. ], batch size: 57, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:21:25,654 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289859.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:22:15,731 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 3.088e+02 3.696e+02 4.617e+02 9.164e+02, threshold=7.392e+02, percent-clipped=2.0 2023-05-02 18:22:35,929 INFO [train.py:904] (5/8) Epoch 29, batch 5700, loss[loss=0.1886, simple_loss=0.2874, pruned_loss=0.04496, over 16463.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2896, pruned_loss=0.0592, over 3039404.05 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:23:02,272 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289920.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:23:55,709 INFO [train.py:904] (5/8) Epoch 29, batch 5750, loss[loss=0.2119, simple_loss=0.2985, pruned_loss=0.06266, over 15201.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2926, pruned_loss=0.06097, over 3014951.96 frames. ], batch size: 191, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:24:41,584 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289981.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 18:24:56,848 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.660e+02 3.372e+02 4.165e+02 9.302e+02, threshold=6.745e+02, percent-clipped=2.0 2023-05-02 18:25:08,681 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289998.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:25:20,991 INFO [train.py:904] (5/8) Epoch 29, batch 5800, loss[loss=0.1923, simple_loss=0.2829, pruned_loss=0.0508, over 17017.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2923, pruned_loss=0.06005, over 3012287.82 frames. ], batch size: 55, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:25:51,866 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3419, 2.1715, 1.7492, 1.8622, 2.4071, 2.0738, 2.0501, 2.4917], device='cuda:5'), covar=tensor([0.0276, 0.0471, 0.0641, 0.0575, 0.0312, 0.0449, 0.0247, 0.0339], device='cuda:5'), in_proj_covar=tensor([0.0233, 0.0243, 0.0233, 0.0234, 0.0244, 0.0243, 0.0240, 0.0244], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:26:34,366 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290050.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:26:39,159 INFO [train.py:904] (5/8) Epoch 29, batch 5850, loss[loss=0.1785, simple_loss=0.2711, pruned_loss=0.04291, over 16594.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2899, pruned_loss=0.05819, over 3029811.59 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:26:48,293 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290059.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:27:39,838 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.628e+02 3.224e+02 3.685e+02 6.132e+02, threshold=6.447e+02, percent-clipped=0.0 2023-05-02 18:28:01,364 INFO [train.py:904] (5/8) Epoch 29, batch 5900, loss[loss=0.2003, simple_loss=0.298, pruned_loss=0.05131, over 16454.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2897, pruned_loss=0.0581, over 3029376.26 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:28:15,116 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:29:20,960 INFO [train.py:904] (5/8) Epoch 29, batch 5950, loss[loss=0.1756, simple_loss=0.2718, pruned_loss=0.03971, over 15360.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2905, pruned_loss=0.05672, over 3055483.61 frames. ], batch size: 191, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:29:30,163 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:30:18,752 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.594e+02 3.044e+02 3.971e+02 8.061e+02, threshold=6.088e+02, percent-clipped=1.0 2023-05-02 18:30:38,300 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2527, 2.9125, 2.6209, 2.2775, 2.1805, 2.2516, 2.9603, 2.8148], device='cuda:5'), covar=tensor([0.2678, 0.0728, 0.1839, 0.2539, 0.2500, 0.2334, 0.0584, 0.1492], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0276, 0.0315, 0.0330, 0.0308, 0.0279, 0.0308, 0.0354], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 18:30:40,229 INFO [train.py:904] (5/8) Epoch 29, batch 6000, loss[loss=0.2176, simple_loss=0.3012, pruned_loss=0.06697, over 15329.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2892, pruned_loss=0.05604, over 3066624.01 frames. ], batch size: 191, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:30:40,229 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 18:30:50,252 INFO [train.py:938] (5/8) Epoch 29, validation: loss=0.1475, simple_loss=0.2594, pruned_loss=0.01778, over 944034.00 frames. 2023-05-02 18:30:50,253 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 18:30:55,864 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:31:08,023 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5536, 1.9108, 2.2180, 2.4509, 2.5769, 2.8138, 1.9339, 2.8341], device='cuda:5'), covar=tensor([0.0280, 0.0569, 0.0397, 0.0427, 0.0372, 0.0263, 0.0614, 0.0200], device='cuda:5'), in_proj_covar=tensor([0.0198, 0.0197, 0.0186, 0.0193, 0.0210, 0.0167, 0.0203, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:31:54,789 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3637, 4.2169, 4.3902, 4.5772, 4.7362, 4.3047, 4.6684, 4.7658], device='cuda:5'), covar=tensor([0.2157, 0.1451, 0.1806, 0.0883, 0.0749, 0.1297, 0.1007, 0.0767], device='cuda:5'), in_proj_covar=tensor([0.0684, 0.0835, 0.0961, 0.0847, 0.0646, 0.0667, 0.0707, 0.0823], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:32:11,213 INFO [train.py:904] (5/8) Epoch 29, batch 6050, loss[loss=0.2066, simple_loss=0.2956, pruned_loss=0.05884, over 16891.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2871, pruned_loss=0.0547, over 3089007.33 frames. ], batch size: 109, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:32:15,062 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0615, 3.7483, 4.2779, 1.9016, 4.4846, 4.4564, 3.3425, 3.3748], device='cuda:5'), covar=tensor([0.0686, 0.0270, 0.0164, 0.1413, 0.0065, 0.0157, 0.0398, 0.0477], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0132, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 18:32:42,695 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290276.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:33:04,447 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.468e+02 2.905e+02 3.476e+02 8.072e+02, threshold=5.810e+02, percent-clipped=1.0 2023-05-02 18:33:29,600 INFO [train.py:904] (5/8) Epoch 29, batch 6100, loss[loss=0.1909, simple_loss=0.2804, pruned_loss=0.05073, over 16234.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2863, pruned_loss=0.05361, over 3111210.94 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:33:53,044 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:34:21,544 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 18:34:46,616 INFO [train.py:904] (5/8) Epoch 29, batch 6150, loss[loss=0.202, simple_loss=0.2786, pruned_loss=0.06274, over 11446.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2853, pruned_loss=0.05348, over 3110008.69 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:34:47,114 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:05,040 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5198, 3.7428, 2.8463, 2.3007, 2.4529, 2.5274, 4.0168, 3.3271], device='cuda:5'), covar=tensor([0.3170, 0.0620, 0.1861, 0.2767, 0.2715, 0.2116, 0.0452, 0.1383], device='cuda:5'), in_proj_covar=tensor([0.0338, 0.0277, 0.0315, 0.0331, 0.0308, 0.0280, 0.0308, 0.0355], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 18:35:27,513 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290379.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:44,986 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.708e+02 2.649e+02 3.165e+02 3.863e+02 6.871e+02, threshold=6.329e+02, percent-clipped=1.0 2023-05-02 18:36:04,351 INFO [train.py:904] (5/8) Epoch 29, batch 6200, loss[loss=0.2006, simple_loss=0.2732, pruned_loss=0.06397, over 12032.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2841, pruned_loss=0.05334, over 3117212.58 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:36:08,280 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:37:21,107 INFO [train.py:904] (5/8) Epoch 29, batch 6250, loss[loss=0.2065, simple_loss=0.2866, pruned_loss=0.06322, over 11209.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2839, pruned_loss=0.05284, over 3123811.71 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:37:41,271 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9405, 2.2033, 2.3173, 3.4460, 2.0947, 2.4526, 2.3092, 2.3392], device='cuda:5'), covar=tensor([0.1515, 0.3534, 0.3123, 0.0736, 0.4235, 0.2629, 0.3668, 0.3289], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0475, 0.0387, 0.0337, 0.0446, 0.0547, 0.0449, 0.0558], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:38:15,687 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.626e+02 3.170e+02 3.673e+02 5.508e+02, threshold=6.341e+02, percent-clipped=0.0 2023-05-02 18:38:34,378 INFO [train.py:904] (5/8) Epoch 29, batch 6300, loss[loss=0.1866, simple_loss=0.28, pruned_loss=0.04661, over 16777.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2835, pruned_loss=0.05208, over 3140426.81 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:38:38,258 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1003, 2.5590, 2.0594, 2.3310, 2.8905, 2.5000, 2.8263, 3.0203], device='cuda:5'), covar=tensor([0.0220, 0.0485, 0.0665, 0.0508, 0.0306, 0.0420, 0.0265, 0.0294], device='cuda:5'), in_proj_covar=tensor([0.0232, 0.0244, 0.0234, 0.0235, 0.0245, 0.0244, 0.0240, 0.0244], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:39:03,959 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6074, 3.2895, 3.7383, 1.8796, 3.8908, 3.8798, 2.9538, 2.9475], device='cuda:5'), covar=tensor([0.0779, 0.0319, 0.0204, 0.1306, 0.0087, 0.0202, 0.0477, 0.0491], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0133, 0.0130, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 18:39:52,815 INFO [train.py:904] (5/8) Epoch 29, batch 6350, loss[loss=0.1861, simple_loss=0.2722, pruned_loss=0.04999, over 17067.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.284, pruned_loss=0.05307, over 3123417.97 frames. ], batch size: 53, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:40:28,397 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290576.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:40:50,036 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.799e+02 3.423e+02 4.274e+02 8.913e+02, threshold=6.847e+02, percent-clipped=6.0 2023-05-02 18:41:09,303 INFO [train.py:904] (5/8) Epoch 29, batch 6400, loss[loss=0.1911, simple_loss=0.2769, pruned_loss=0.0526, over 16650.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2845, pruned_loss=0.05445, over 3107394.85 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:41:17,184 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0083, 3.2345, 3.5556, 1.9255, 3.0703, 2.2151, 3.4410, 3.5677], device='cuda:5'), covar=tensor([0.0262, 0.0842, 0.0579, 0.2358, 0.0837, 0.1060, 0.0677, 0.0948], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 18:41:39,271 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290624.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:42:04,946 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-02 18:42:23,304 INFO [train.py:904] (5/8) Epoch 29, batch 6450, loss[loss=0.1784, simple_loss=0.2777, pruned_loss=0.0396, over 16838.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2848, pruned_loss=0.05406, over 3111501.15 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:42:23,862 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:42:53,367 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:43:19,913 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.663e+02 2.912e+02 3.666e+02 7.754e+02, threshold=5.824e+02, percent-clipped=2.0 2023-05-02 18:43:37,902 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:43:40,051 INFO [train.py:904] (5/8) Epoch 29, batch 6500, loss[loss=0.1761, simple_loss=0.265, pruned_loss=0.04362, over 16717.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2832, pruned_loss=0.05369, over 3113499.66 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:43:44,266 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290706.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:44:05,224 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9400, 4.2010, 4.0102, 4.0732, 3.7130, 3.8335, 3.8463, 4.1721], device='cuda:5'), covar=tensor([0.1098, 0.0860, 0.1032, 0.0907, 0.0778, 0.1681, 0.0993, 0.1053], device='cuda:5'), in_proj_covar=tensor([0.0722, 0.0865, 0.0714, 0.0674, 0.0552, 0.0553, 0.0727, 0.0681], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 18:44:31,084 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 18:44:59,510 INFO [train.py:904] (5/8) Epoch 29, batch 6550, loss[loss=0.2047, simple_loss=0.3046, pruned_loss=0.05243, over 16339.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2852, pruned_loss=0.0546, over 3109059.57 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:44:59,830 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:45:58,159 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.774e+02 3.264e+02 3.867e+02 7.263e+02, threshold=6.527e+02, percent-clipped=1.0 2023-05-02 18:46:19,479 INFO [train.py:904] (5/8) Epoch 29, batch 6600, loss[loss=0.2454, simple_loss=0.3155, pruned_loss=0.08768, over 11384.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.287, pruned_loss=0.05476, over 3094116.39 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:47:38,816 INFO [train.py:904] (5/8) Epoch 29, batch 6650, loss[loss=0.1868, simple_loss=0.2689, pruned_loss=0.05235, over 16899.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.287, pruned_loss=0.05528, over 3095347.69 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:48:35,970 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.668e+02 3.227e+02 3.787e+02 6.260e+02, threshold=6.453e+02, percent-clipped=0.0 2023-05-02 18:48:55,756 INFO [train.py:904] (5/8) Epoch 29, batch 6700, loss[loss=0.1803, simple_loss=0.2709, pruned_loss=0.04483, over 16838.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2863, pruned_loss=0.05562, over 3094088.48 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:50:12,985 INFO [train.py:904] (5/8) Epoch 29, batch 6750, loss[loss=0.1741, simple_loss=0.2662, pruned_loss=0.04098, over 16901.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2854, pruned_loss=0.05579, over 3084709.21 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:50:43,279 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290974.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:51:10,932 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 2.623e+02 3.095e+02 3.882e+02 8.048e+02, threshold=6.190e+02, percent-clipped=2.0 2023-05-02 18:51:28,430 INFO [train.py:904] (5/8) Epoch 29, batch 6800, loss[loss=0.1852, simple_loss=0.2798, pruned_loss=0.04528, over 16400.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2858, pruned_loss=0.05597, over 3091134.17 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:51:57,163 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291022.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:52:03,264 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 18:52:46,136 INFO [train.py:904] (5/8) Epoch 29, batch 6850, loss[loss=0.1891, simple_loss=0.2921, pruned_loss=0.04309, over 16879.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2864, pruned_loss=0.05619, over 3088618.97 frames. ], batch size: 109, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:53:09,463 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291069.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:53:28,407 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2128, 3.2405, 2.0487, 3.4627, 2.5021, 3.4679, 2.1597, 2.6187], device='cuda:5'), covar=tensor([0.0316, 0.0456, 0.1678, 0.0337, 0.0916, 0.0721, 0.1605, 0.0854], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0181, 0.0196, 0.0174, 0.0180, 0.0221, 0.0204, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 18:53:42,657 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.621e+02 3.195e+02 3.750e+02 8.276e+02, threshold=6.389e+02, percent-clipped=2.0 2023-05-02 18:54:02,730 INFO [train.py:904] (5/8) Epoch 29, batch 6900, loss[loss=0.1907, simple_loss=0.2931, pruned_loss=0.04422, over 16847.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2886, pruned_loss=0.05509, over 3120447.36 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:54:44,636 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:55:20,068 INFO [train.py:904] (5/8) Epoch 29, batch 6950, loss[loss=0.1899, simple_loss=0.2745, pruned_loss=0.05267, over 16497.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2896, pruned_loss=0.05638, over 3110548.59 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:56:18,591 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 3.069e+02 3.631e+02 4.410e+02 8.633e+02, threshold=7.261e+02, percent-clipped=4.0 2023-05-02 18:56:36,626 INFO [train.py:904] (5/8) Epoch 29, batch 7000, loss[loss=0.1975, simple_loss=0.2977, pruned_loss=0.04866, over 16420.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2896, pruned_loss=0.05571, over 3104779.70 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:56:37,082 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8668, 5.2332, 5.4647, 5.1171, 5.2148, 5.8036, 5.2161, 4.9503], device='cuda:5'), covar=tensor([0.1141, 0.1709, 0.2344, 0.1841, 0.2128, 0.0803, 0.1678, 0.2228], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0643, 0.0713, 0.0519, 0.0696, 0.0731, 0.0552, 0.0696], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 18:57:09,151 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2859, 2.9447, 2.6744, 2.2870, 2.2232, 2.3229, 2.9404, 2.8537], device='cuda:5'), covar=tensor([0.2663, 0.0712, 0.1740, 0.2684, 0.2505, 0.2297, 0.0599, 0.1475], device='cuda:5'), in_proj_covar=tensor([0.0337, 0.0276, 0.0315, 0.0329, 0.0307, 0.0279, 0.0306, 0.0352], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 18:57:50,297 INFO [train.py:904] (5/8) Epoch 29, batch 7050, loss[loss=0.236, simple_loss=0.3093, pruned_loss=0.08133, over 16706.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2907, pruned_loss=0.05595, over 3098523.74 frames. ], batch size: 57, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:58:49,452 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.753e+02 3.384e+02 4.060e+02 1.222e+03, threshold=6.768e+02, percent-clipped=4.0 2023-05-02 18:59:07,325 INFO [train.py:904] (5/8) Epoch 29, batch 7100, loss[loss=0.1964, simple_loss=0.2889, pruned_loss=0.05192, over 16981.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2894, pruned_loss=0.05606, over 3096543.03 frames. ], batch size: 41, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:59:19,911 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9092, 2.7102, 2.7623, 2.1816, 2.6350, 2.1303, 2.7057, 2.9227], device='cuda:5'), covar=tensor([0.0284, 0.0861, 0.0550, 0.1856, 0.0893, 0.1037, 0.0593, 0.0809], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0149, 0.0133, 0.0146, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 19:00:25,896 INFO [train.py:904] (5/8) Epoch 29, batch 7150, loss[loss=0.2131, simple_loss=0.2919, pruned_loss=0.0672, over 11535.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2878, pruned_loss=0.05629, over 3076101.54 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:01:23,572 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.798e+02 3.396e+02 4.128e+02 9.099e+02, threshold=6.792e+02, percent-clipped=1.0 2023-05-02 19:01:41,845 INFO [train.py:904] (5/8) Epoch 29, batch 7200, loss[loss=0.1982, simple_loss=0.2846, pruned_loss=0.05595, over 12086.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2852, pruned_loss=0.05454, over 3068485.87 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:02:14,456 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=291425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:03:00,076 INFO [train.py:904] (5/8) Epoch 29, batch 7250, loss[loss=0.1662, simple_loss=0.2575, pruned_loss=0.0375, over 16848.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2829, pruned_loss=0.05353, over 3055232.38 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:03:58,889 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.627e+02 2.585e+02 3.000e+02 3.629e+02 8.129e+02, threshold=6.000e+02, percent-clipped=2.0 2023-05-02 19:04:05,766 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4151, 3.3821, 3.4404, 3.5116, 3.5482, 3.3128, 3.5195, 3.5871], device='cuda:5'), covar=tensor([0.1230, 0.0966, 0.1002, 0.0628, 0.0643, 0.2219, 0.1112, 0.0873], device='cuda:5'), in_proj_covar=tensor([0.0676, 0.0827, 0.0950, 0.0836, 0.0640, 0.0659, 0.0701, 0.0813], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:04:16,131 INFO [train.py:904] (5/8) Epoch 29, batch 7300, loss[loss=0.2101, simple_loss=0.2882, pruned_loss=0.06597, over 16570.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2824, pruned_loss=0.05319, over 3074452.68 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:05:25,675 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 19:05:32,036 INFO [train.py:904] (5/8) Epoch 29, batch 7350, loss[loss=0.2416, simple_loss=0.3076, pruned_loss=0.08779, over 11296.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2837, pruned_loss=0.05417, over 3066372.21 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:06:32,138 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.867e+02 3.369e+02 3.998e+02 9.840e+02, threshold=6.738e+02, percent-clipped=8.0 2023-05-02 19:06:49,564 INFO [train.py:904] (5/8) Epoch 29, batch 7400, loss[loss=0.2086, simple_loss=0.3034, pruned_loss=0.0569, over 16471.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2844, pruned_loss=0.05463, over 3056577.38 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:06:53,651 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 19:08:07,289 INFO [train.py:904] (5/8) Epoch 29, batch 7450, loss[loss=0.195, simple_loss=0.2915, pruned_loss=0.04932, over 16740.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2864, pruned_loss=0.05624, over 3040665.16 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:09:05,178 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5409, 3.5279, 3.5024, 2.7469, 3.3987, 2.0405, 3.2308, 2.9690], device='cuda:5'), covar=tensor([0.0186, 0.0176, 0.0215, 0.0263, 0.0137, 0.2562, 0.0153, 0.0276], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0176, 0.0215, 0.0186, 0.0191, 0.0218, 0.0202, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:09:10,903 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.789e+02 3.286e+02 3.823e+02 6.627e+02, threshold=6.571e+02, percent-clipped=0.0 2023-05-02 19:09:24,846 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 19:09:25,619 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7447, 1.8195, 1.5757, 1.4499, 1.9550, 1.5563, 1.5346, 1.8776], device='cuda:5'), covar=tensor([0.0295, 0.0333, 0.0506, 0.0417, 0.0252, 0.0332, 0.0210, 0.0262], device='cuda:5'), in_proj_covar=tensor([0.0230, 0.0244, 0.0234, 0.0234, 0.0244, 0.0241, 0.0239, 0.0242], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:09:28,114 INFO [train.py:904] (5/8) Epoch 29, batch 7500, loss[loss=0.1813, simple_loss=0.2738, pruned_loss=0.0444, over 16421.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2865, pruned_loss=0.05534, over 3050101.78 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:10:01,695 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=291725.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:10:25,214 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 19:10:42,357 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 19:10:45,768 INFO [train.py:904] (5/8) Epoch 29, batch 7550, loss[loss=0.1829, simple_loss=0.2682, pruned_loss=0.04878, over 16526.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2864, pruned_loss=0.05632, over 3038434.79 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:11:15,435 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:19,708 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:28,359 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4095, 2.5467, 2.5724, 4.2364, 2.4173, 2.9008, 2.5898, 2.6760], device='cuda:5'), covar=tensor([0.1446, 0.3449, 0.2881, 0.0534, 0.3979, 0.2363, 0.3314, 0.3107], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0476, 0.0386, 0.0337, 0.0446, 0.0546, 0.0449, 0.0558], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:11:44,714 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.637e+02 3.162e+02 3.774e+02 1.302e+03, threshold=6.323e+02, percent-clipped=2.0 2023-05-02 19:12:01,417 INFO [train.py:904] (5/8) Epoch 29, batch 7600, loss[loss=0.2226, simple_loss=0.293, pruned_loss=0.07605, over 11461.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2853, pruned_loss=0.05604, over 3072317.94 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:12:52,604 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291837.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:13:13,509 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8150, 1.8296, 1.6117, 1.4930, 1.9709, 1.5897, 1.5765, 1.9354], device='cuda:5'), covar=tensor([0.0204, 0.0323, 0.0430, 0.0382, 0.0240, 0.0310, 0.0195, 0.0229], device='cuda:5'), in_proj_covar=tensor([0.0229, 0.0243, 0.0233, 0.0233, 0.0243, 0.0240, 0.0238, 0.0241], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:13:18,978 INFO [train.py:904] (5/8) Epoch 29, batch 7650, loss[loss=0.2113, simple_loss=0.3045, pruned_loss=0.05899, over 15384.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2855, pruned_loss=0.05599, over 3080458.23 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:14:00,662 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-02 19:14:20,912 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.915e+02 3.313e+02 4.275e+02 8.487e+02, threshold=6.627e+02, percent-clipped=5.0 2023-05-02 19:14:36,026 INFO [train.py:904] (5/8) Epoch 29, batch 7700, loss[loss=0.2287, simple_loss=0.3007, pruned_loss=0.07837, over 11243.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2855, pruned_loss=0.05632, over 3071993.58 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:15:53,503 INFO [train.py:904] (5/8) Epoch 29, batch 7750, loss[loss=0.2031, simple_loss=0.2946, pruned_loss=0.05587, over 16787.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2856, pruned_loss=0.05648, over 3067877.78 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:16:37,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4079, 2.5493, 2.4062, 4.1982, 2.3781, 2.8877, 2.5540, 2.6941], device='cuda:5'), covar=tensor([0.1456, 0.3551, 0.3080, 0.0543, 0.4043, 0.2392, 0.3475, 0.3110], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0476, 0.0387, 0.0337, 0.0447, 0.0547, 0.0449, 0.0558], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:16:42,177 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8552, 4.9073, 4.7275, 4.3646, 4.3903, 4.8049, 4.6141, 4.5180], device='cuda:5'), covar=tensor([0.0732, 0.0724, 0.0371, 0.0389, 0.1067, 0.0640, 0.0487, 0.0800], device='cuda:5'), in_proj_covar=tensor([0.0309, 0.0468, 0.0361, 0.0362, 0.0358, 0.0417, 0.0249, 0.0434], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:16:55,757 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.177e+02 2.797e+02 3.314e+02 3.963e+02 7.468e+02, threshold=6.627e+02, percent-clipped=1.0 2023-05-02 19:17:14,578 INFO [train.py:904] (5/8) Epoch 29, batch 7800, loss[loss=0.2033, simple_loss=0.2869, pruned_loss=0.05987, over 15409.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2862, pruned_loss=0.05652, over 3075303.26 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:18:30,025 INFO [train.py:904] (5/8) Epoch 29, batch 7850, loss[loss=0.2146, simple_loss=0.2931, pruned_loss=0.06806, over 11187.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.287, pruned_loss=0.05635, over 3065613.32 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:18:30,523 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:19:30,098 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.705e+02 3.148e+02 3.782e+02 7.710e+02, threshold=6.297e+02, percent-clipped=1.0 2023-05-02 19:19:43,461 INFO [train.py:904] (5/8) Epoch 29, batch 7900, loss[loss=0.2009, simple_loss=0.2946, pruned_loss=0.05362, over 16242.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2858, pruned_loss=0.05583, over 3070174.38 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:19:59,943 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=292115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:20:26,275 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:20:57,194 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 19:21:01,246 INFO [train.py:904] (5/8) Epoch 29, batch 7950, loss[loss=0.2245, simple_loss=0.3012, pruned_loss=0.07388, over 16851.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2864, pruned_loss=0.05602, over 3081459.82 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:22:01,837 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9297, 2.0715, 2.1697, 3.4813, 2.0296, 2.3113, 2.2184, 2.2081], device='cuda:5'), covar=tensor([0.1712, 0.4058, 0.3285, 0.0716, 0.4761, 0.2986, 0.3971, 0.3793], device='cuda:5'), in_proj_covar=tensor([0.0423, 0.0477, 0.0387, 0.0338, 0.0448, 0.0549, 0.0450, 0.0559], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:22:03,659 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.577e+02 3.154e+02 3.832e+02 7.252e+02, threshold=6.308e+02, percent-clipped=1.0 2023-05-02 19:22:18,483 INFO [train.py:904] (5/8) Epoch 29, batch 8000, loss[loss=0.2061, simple_loss=0.2966, pruned_loss=0.05783, over 16394.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.287, pruned_loss=0.05656, over 3077724.83 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:23:31,128 INFO [train.py:904] (5/8) Epoch 29, batch 8050, loss[loss=0.1866, simple_loss=0.2797, pruned_loss=0.04677, over 16429.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2871, pruned_loss=0.05657, over 3065829.37 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:24:32,482 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.613e+02 3.145e+02 3.848e+02 6.977e+02, threshold=6.289e+02, percent-clipped=2.0 2023-05-02 19:24:46,384 INFO [train.py:904] (5/8) Epoch 29, batch 8100, loss[loss=0.1878, simple_loss=0.2803, pruned_loss=0.0476, over 16606.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2864, pruned_loss=0.05606, over 3077113.61 frames. ], batch size: 62, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:26:01,422 INFO [train.py:904] (5/8) Epoch 29, batch 8150, loss[loss=0.2085, simple_loss=0.2992, pruned_loss=0.05896, over 15451.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2848, pruned_loss=0.05562, over 3065239.79 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:26:40,471 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-02 19:27:01,242 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.736e+02 3.221e+02 3.909e+02 8.294e+02, threshold=6.443e+02, percent-clipped=4.0 2023-05-02 19:27:15,048 INFO [train.py:904] (5/8) Epoch 29, batch 8200, loss[loss=0.1988, simple_loss=0.2919, pruned_loss=0.0529, over 15284.00 frames. ], tot_loss[loss=0.196, simple_loss=0.282, pruned_loss=0.05498, over 3062645.14 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:27:25,931 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292410.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:27:51,078 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3977, 4.5158, 4.6797, 4.4598, 4.5926, 5.0517, 4.5874, 4.2845], device='cuda:5'), covar=tensor([0.1595, 0.2032, 0.2437, 0.2045, 0.2560, 0.1023, 0.1695, 0.2651], device='cuda:5'), in_proj_covar=tensor([0.0430, 0.0640, 0.0709, 0.0517, 0.0692, 0.0730, 0.0550, 0.0694], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 19:28:00,670 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292432.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:28:18,308 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1584, 2.9176, 2.9834, 1.7636, 3.1637, 3.2392, 2.7935, 2.6759], device='cuda:5'), covar=tensor([0.0880, 0.0296, 0.0345, 0.1362, 0.0154, 0.0295, 0.0476, 0.0513], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0139, 0.0088, 0.0132, 0.0130, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 19:28:34,810 INFO [train.py:904] (5/8) Epoch 29, batch 8250, loss[loss=0.1656, simple_loss=0.2656, pruned_loss=0.0328, over 16841.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.281, pruned_loss=0.05258, over 3059106.81 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:29:17,989 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:29:41,439 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.407e+02 2.184e+02 2.675e+02 3.477e+02 6.200e+02, threshold=5.351e+02, percent-clipped=0.0 2023-05-02 19:29:45,234 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9060, 2.6446, 2.5779, 1.9344, 2.4714, 2.7017, 2.5931, 1.9173], device='cuda:5'), covar=tensor([0.0451, 0.0140, 0.0109, 0.0370, 0.0186, 0.0143, 0.0126, 0.0459], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0135, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 19:29:47,833 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 19:29:55,808 INFO [train.py:904] (5/8) Epoch 29, batch 8300, loss[loss=0.1837, simple_loss=0.2781, pruned_loss=0.0447, over 16776.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2783, pruned_loss=0.04968, over 3053859.95 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:31:15,576 INFO [train.py:904] (5/8) Epoch 29, batch 8350, loss[loss=0.1855, simple_loss=0.2731, pruned_loss=0.04897, over 12314.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2782, pruned_loss=0.04791, over 3062886.82 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:32:20,506 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.032e+02 2.659e+02 3.191e+02 5.260e+02, threshold=5.318e+02, percent-clipped=0.0 2023-05-02 19:32:36,083 INFO [train.py:904] (5/8) Epoch 29, batch 8400, loss[loss=0.163, simple_loss=0.2568, pruned_loss=0.03466, over 16231.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2757, pruned_loss=0.04578, over 3059779.87 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:33:14,765 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1985, 3.2798, 2.0644, 3.4896, 2.4461, 3.5213, 2.2492, 2.7015], device='cuda:5'), covar=tensor([0.0325, 0.0343, 0.1501, 0.0307, 0.0828, 0.0488, 0.1430, 0.0708], device='cuda:5'), in_proj_covar=tensor([0.0173, 0.0177, 0.0192, 0.0170, 0.0177, 0.0215, 0.0200, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 19:33:56,347 INFO [train.py:904] (5/8) Epoch 29, batch 8450, loss[loss=0.185, simple_loss=0.2774, pruned_loss=0.04628, over 16752.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2736, pruned_loss=0.04385, over 3053506.54 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:35:03,482 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.196e+02 2.552e+02 2.926e+02 6.173e+02, threshold=5.103e+02, percent-clipped=1.0 2023-05-02 19:35:19,255 INFO [train.py:904] (5/8) Epoch 29, batch 8500, loss[loss=0.167, simple_loss=0.2627, pruned_loss=0.03568, over 16936.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2702, pruned_loss=0.04184, over 3059776.76 frames. ], batch size: 90, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:35:28,927 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292710.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:36:42,999 INFO [train.py:904] (5/8) Epoch 29, batch 8550, loss[loss=0.1726, simple_loss=0.2644, pruned_loss=0.04044, over 16634.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2675, pruned_loss=0.0404, over 3057937.45 frames. ], batch size: 69, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:36:52,496 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292758.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:36:54,455 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-02 19:38:01,144 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6294, 3.0003, 3.3582, 1.9598, 2.8405, 2.1758, 3.1659, 3.1890], device='cuda:5'), covar=tensor([0.0324, 0.0921, 0.0504, 0.2315, 0.0876, 0.1115, 0.0720, 0.1013], device='cuda:5'), in_proj_covar=tensor([0.0159, 0.0168, 0.0168, 0.0155, 0.0146, 0.0132, 0.0144, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 19:38:04,313 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.162e+02 2.570e+02 3.257e+02 5.024e+02, threshold=5.139e+02, percent-clipped=0.0 2023-05-02 19:38:21,510 INFO [train.py:904] (5/8) Epoch 29, batch 8600, loss[loss=0.1715, simple_loss=0.2717, pruned_loss=0.03565, over 16413.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2667, pruned_loss=0.03915, over 3055487.25 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:40:01,874 INFO [train.py:904] (5/8) Epoch 29, batch 8650, loss[loss=0.1527, simple_loss=0.2583, pruned_loss=0.02357, over 16891.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2645, pruned_loss=0.03779, over 3045468.02 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:41:31,355 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.024e+02 2.393e+02 2.870e+02 4.192e+02, threshold=4.785e+02, percent-clipped=0.0 2023-05-02 19:41:48,481 INFO [train.py:904] (5/8) Epoch 29, batch 8700, loss[loss=0.1667, simple_loss=0.2668, pruned_loss=0.03324, over 16474.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.262, pruned_loss=0.03677, over 3053730.95 frames. ], batch size: 75, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:43:26,858 INFO [train.py:904] (5/8) Epoch 29, batch 8750, loss[loss=0.1774, simple_loss=0.2772, pruned_loss=0.03884, over 16307.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2623, pruned_loss=0.03651, over 3056906.16 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:44:01,878 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:45:00,945 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.165e+02 2.592e+02 3.174e+02 6.684e+02, threshold=5.184e+02, percent-clipped=4.0 2023-05-02 19:45:17,875 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5206, 3.6937, 2.4881, 2.1783, 2.2097, 2.2139, 3.7374, 3.0585], device='cuda:5'), covar=tensor([0.3432, 0.0668, 0.2388, 0.3402, 0.3246, 0.2668, 0.0614, 0.1590], device='cuda:5'), in_proj_covar=tensor([0.0330, 0.0269, 0.0308, 0.0322, 0.0299, 0.0273, 0.0299, 0.0343], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 19:45:20,147 INFO [train.py:904] (5/8) Epoch 29, batch 8800, loss[loss=0.175, simple_loss=0.2679, pruned_loss=0.04111, over 16786.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2608, pruned_loss=0.0355, over 3066438.59 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:46:12,268 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:46:15,918 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8691, 3.9470, 4.0775, 3.8515, 4.0039, 4.3752, 4.0051, 3.7554], device='cuda:5'), covar=tensor([0.2313, 0.1950, 0.1817, 0.2400, 0.2481, 0.1522, 0.1563, 0.2669], device='cuda:5'), in_proj_covar=tensor([0.0417, 0.0623, 0.0693, 0.0506, 0.0676, 0.0712, 0.0538, 0.0676], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 19:46:28,932 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0337, 4.1308, 4.3383, 4.3327, 4.3535, 4.1520, 4.1369, 4.0897], device='cuda:5'), covar=tensor([0.0313, 0.0540, 0.0421, 0.0356, 0.0335, 0.0381, 0.0743, 0.0452], device='cuda:5'), in_proj_covar=tensor([0.0430, 0.0488, 0.0469, 0.0434, 0.0513, 0.0494, 0.0569, 0.0397], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 19:47:05,868 INFO [train.py:904] (5/8) Epoch 29, batch 8850, loss[loss=0.1802, simple_loss=0.2847, pruned_loss=0.03782, over 16669.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2635, pruned_loss=0.03514, over 3059764.50 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:48:05,784 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:48:35,604 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.279e+02 2.601e+02 3.166e+02 4.771e+02, threshold=5.201e+02, percent-clipped=0.0 2023-05-02 19:48:55,730 INFO [train.py:904] (5/8) Epoch 29, batch 8900, loss[loss=0.1426, simple_loss=0.2442, pruned_loss=0.02044, over 16780.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2637, pruned_loss=0.03453, over 3048274.80 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:49:23,414 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293117.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:50:34,630 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:50:57,035 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 19:51:01,963 INFO [train.py:904] (5/8) Epoch 29, batch 8950, loss[loss=0.1659, simple_loss=0.2566, pruned_loss=0.03765, over 17024.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2634, pruned_loss=0.03502, over 3067029.94 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:51:52,648 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293178.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:52:32,344 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 2.109e+02 2.425e+02 3.053e+02 5.475e+02, threshold=4.851e+02, percent-clipped=2.0 2023-05-02 19:52:53,152 INFO [train.py:904] (5/8) Epoch 29, batch 9000, loss[loss=0.1637, simple_loss=0.2553, pruned_loss=0.03602, over 16843.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2607, pruned_loss=0.034, over 3075237.29 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:52:53,153 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 19:53:02,752 INFO [train.py:938] (5/8) Epoch 29, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01987, over 944034.00 frames. 2023-05-02 19:53:02,753 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 19:53:49,446 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2042, 1.6001, 1.9278, 2.1389, 2.2158, 2.3543, 1.7903, 2.3307], device='cuda:5'), covar=tensor([0.0271, 0.0581, 0.0359, 0.0368, 0.0384, 0.0246, 0.0615, 0.0171], device='cuda:5'), in_proj_covar=tensor([0.0193, 0.0194, 0.0183, 0.0187, 0.0205, 0.0163, 0.0200, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 19:54:47,961 INFO [train.py:904] (5/8) Epoch 29, batch 9050, loss[loss=0.1475, simple_loss=0.2346, pruned_loss=0.03016, over 16712.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.262, pruned_loss=0.03441, over 3088862.96 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:56:03,699 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-05-02 19:56:15,030 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.113e+02 2.508e+02 2.955e+02 4.534e+02, threshold=5.015e+02, percent-clipped=0.0 2023-05-02 19:56:34,928 INFO [train.py:904] (5/8) Epoch 29, batch 9100, loss[loss=0.1628, simple_loss=0.267, pruned_loss=0.02927, over 16661.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2618, pruned_loss=0.03494, over 3103295.75 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:56:36,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8483, 3.5035, 3.8889, 1.9943, 4.0796, 4.1504, 3.2181, 3.2102], device='cuda:5'), covar=tensor([0.0692, 0.0283, 0.0236, 0.1207, 0.0081, 0.0146, 0.0390, 0.0435], device='cuda:5'), in_proj_covar=tensor([0.0145, 0.0108, 0.0099, 0.0136, 0.0085, 0.0128, 0.0127, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 19:57:17,602 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:57:35,691 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2974, 4.3452, 4.4915, 4.2576, 4.4224, 4.8598, 4.4091, 4.0545], device='cuda:5'), covar=tensor([0.1593, 0.1939, 0.2143, 0.2158, 0.2316, 0.1051, 0.1456, 0.2584], device='cuda:5'), in_proj_covar=tensor([0.0414, 0.0620, 0.0687, 0.0503, 0.0673, 0.0709, 0.0533, 0.0671], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 19:58:34,126 INFO [train.py:904] (5/8) Epoch 29, batch 9150, loss[loss=0.1529, simple_loss=0.2435, pruned_loss=0.03118, over 16634.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2624, pruned_loss=0.03465, over 3095853.83 frames. ], batch size: 57, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:58:51,981 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293361.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:56,826 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:00:04,473 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 2.108e+02 2.614e+02 3.370e+02 9.014e+02, threshold=5.228e+02, percent-clipped=3.0 2023-05-02 20:00:20,600 INFO [train.py:904] (5/8) Epoch 29, batch 9200, loss[loss=0.1727, simple_loss=0.2679, pruned_loss=0.03877, over 16679.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2579, pruned_loss=0.0336, over 3087219.21 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:00:55,044 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293422.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:00,694 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:23,546 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293437.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:58,344 INFO [train.py:904] (5/8) Epoch 29, batch 9250, loss[loss=0.1611, simple_loss=0.263, pruned_loss=0.02961, over 15293.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2577, pruned_loss=0.03381, over 3068559.54 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:02:39,076 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:03:28,488 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.175e+02 2.514e+02 3.115e+02 5.451e+02, threshold=5.029e+02, percent-clipped=2.0 2023-05-02 20:03:38,232 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9537, 2.6669, 3.0095, 2.1591, 2.7679, 2.2018, 2.7233, 2.8552], device='cuda:5'), covar=tensor([0.0318, 0.1042, 0.0464, 0.1905, 0.0743, 0.1004, 0.0602, 0.0881], device='cuda:5'), in_proj_covar=tensor([0.0156, 0.0165, 0.0166, 0.0153, 0.0144, 0.0130, 0.0142, 0.0177], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:03:49,272 INFO [train.py:904] (5/8) Epoch 29, batch 9300, loss[loss=0.1347, simple_loss=0.2301, pruned_loss=0.01959, over 16682.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2559, pruned_loss=0.0331, over 3057491.23 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:05:33,322 INFO [train.py:904] (5/8) Epoch 29, batch 9350, loss[loss=0.1618, simple_loss=0.2568, pruned_loss=0.03338, over 16224.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2551, pruned_loss=0.03271, over 3074036.63 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:06:56,974 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 1.973e+02 2.358e+02 2.873e+02 5.255e+02, threshold=4.716e+02, percent-clipped=1.0 2023-05-02 20:07:01,149 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9238, 3.9239, 4.1570, 4.1374, 4.1576, 3.9711, 3.9438, 3.9961], device='cuda:5'), covar=tensor([0.0430, 0.0962, 0.0592, 0.0722, 0.0791, 0.0673, 0.0994, 0.0498], device='cuda:5'), in_proj_covar=tensor([0.0426, 0.0484, 0.0467, 0.0430, 0.0511, 0.0491, 0.0564, 0.0394], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 20:07:15,896 INFO [train.py:904] (5/8) Epoch 29, batch 9400, loss[loss=0.1504, simple_loss=0.2643, pruned_loss=0.01823, over 16862.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2554, pruned_loss=0.03262, over 3068274.63 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:07:53,888 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:08:23,072 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 20:08:55,057 INFO [train.py:904] (5/8) Epoch 29, batch 9450, loss[loss=0.1571, simple_loss=0.2489, pruned_loss=0.03263, over 12545.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2574, pruned_loss=0.0329, over 3064148.88 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:09:29,882 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:10:18,287 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.101e+02 2.359e+02 2.998e+02 5.430e+02, threshold=4.718e+02, percent-clipped=2.0 2023-05-02 20:10:34,123 INFO [train.py:904] (5/8) Epoch 29, batch 9500, loss[loss=0.1607, simple_loss=0.246, pruned_loss=0.03774, over 12989.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2571, pruned_loss=0.03304, over 3056721.47 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:11:02,476 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:09,000 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:31,650 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 20:11:38,324 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2096, 2.5679, 2.5758, 1.9065, 2.7917, 2.8452, 2.5583, 2.5286], device='cuda:5'), covar=tensor([0.0619, 0.0283, 0.0265, 0.1019, 0.0126, 0.0234, 0.0434, 0.0433], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0108, 0.0099, 0.0135, 0.0084, 0.0127, 0.0126, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 20:11:42,001 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:12:00,077 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6470, 3.7013, 3.5189, 3.1621, 3.3609, 3.5881, 3.3608, 3.3982], device='cuda:5'), covar=tensor([0.0580, 0.0661, 0.0333, 0.0328, 0.0578, 0.0530, 0.1448, 0.0482], device='cuda:5'), in_proj_covar=tensor([0.0302, 0.0455, 0.0353, 0.0353, 0.0348, 0.0405, 0.0243, 0.0420], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:12:19,512 INFO [train.py:904] (5/8) Epoch 29, batch 9550, loss[loss=0.1533, simple_loss=0.2588, pruned_loss=0.02396, over 16923.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2566, pruned_loss=0.03302, over 3076957.87 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:12:53,428 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1824, 3.2360, 1.7259, 3.4467, 2.3632, 3.4217, 2.0077, 2.6400], device='cuda:5'), covar=tensor([0.0344, 0.0392, 0.1991, 0.0322, 0.0935, 0.0662, 0.1823, 0.0819], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0175, 0.0191, 0.0167, 0.0176, 0.0213, 0.0201, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:12:54,865 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2035, 3.2554, 1.9294, 3.4985, 2.4316, 3.4843, 2.1810, 2.6763], device='cuda:5'), covar=tensor([0.0373, 0.0432, 0.1862, 0.0325, 0.0956, 0.0627, 0.1699, 0.0861], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0175, 0.0191, 0.0167, 0.0176, 0.0213, 0.0201, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:12:59,229 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:25,492 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:25,598 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8653, 3.7522, 3.9296, 4.0152, 4.0993, 3.6965, 4.0806, 4.1509], device='cuda:5'), covar=tensor([0.1685, 0.1126, 0.1246, 0.0727, 0.0584, 0.1851, 0.0732, 0.0715], device='cuda:5'), in_proj_covar=tensor([0.0655, 0.0799, 0.0917, 0.0817, 0.0621, 0.0640, 0.0681, 0.0791], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:13:28,679 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-05-02 20:13:43,638 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.327e+02 2.273e+02 2.766e+02 3.522e+02 6.419e+02, threshold=5.532e+02, percent-clipped=5.0 2023-05-02 20:13:59,142 INFO [train.py:904] (5/8) Epoch 29, batch 9600, loss[loss=0.2088, simple_loss=0.3065, pruned_loss=0.05554, over 16750.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2578, pruned_loss=0.0338, over 3065398.63 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:14:32,483 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:15:45,751 INFO [train.py:904] (5/8) Epoch 29, batch 9650, loss[loss=0.1665, simple_loss=0.257, pruned_loss=0.03801, over 16956.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2604, pruned_loss=0.03428, over 3073962.37 frames. ], batch size: 109, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:16:38,139 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9679, 2.1360, 2.2443, 2.9360, 1.5178, 3.2209, 1.7819, 2.6752], device='cuda:5'), covar=tensor([0.1410, 0.0959, 0.1293, 0.0218, 0.0097, 0.0410, 0.1781, 0.0838], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0179, 0.0198, 0.0199, 0.0201, 0.0214, 0.0209, 0.0197], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:17:16,480 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.571e+02 2.271e+02 2.644e+02 3.266e+02 5.877e+02, threshold=5.288e+02, percent-clipped=1.0 2023-05-02 20:17:35,797 INFO [train.py:904] (5/8) Epoch 29, batch 9700, loss[loss=0.1575, simple_loss=0.2543, pruned_loss=0.03032, over 16748.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2594, pruned_loss=0.03418, over 3068505.48 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:18:08,949 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.7760, 3.1673, 3.4204, 2.0483, 2.9110, 2.2715, 3.3225, 3.4136], device='cuda:5'), covar=tensor([0.0306, 0.0912, 0.0608, 0.2337, 0.0870, 0.1092, 0.0637, 0.0930], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0166, 0.0167, 0.0154, 0.0145, 0.0130, 0.0142, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:19:17,227 INFO [train.py:904] (5/8) Epoch 29, batch 9750, loss[loss=0.1445, simple_loss=0.2465, pruned_loss=0.02125, over 16884.00 frames. ], tot_loss[loss=0.163, simple_loss=0.258, pruned_loss=0.03405, over 3060796.36 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:19:21,453 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7794, 2.2105, 1.9388, 1.9244, 2.4997, 2.1421, 2.0826, 2.6073], device='cuda:5'), covar=tensor([0.0209, 0.0557, 0.0617, 0.0579, 0.0348, 0.0472, 0.0220, 0.0321], device='cuda:5'), in_proj_covar=tensor([0.0220, 0.0236, 0.0225, 0.0226, 0.0235, 0.0234, 0.0229, 0.0232], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:20:15,722 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0144, 1.8745, 1.6744, 1.4868, 2.0124, 1.5956, 1.4800, 1.9622], device='cuda:5'), covar=tensor([0.0202, 0.0359, 0.0474, 0.0409, 0.0283, 0.0344, 0.0174, 0.0245], device='cuda:5'), in_proj_covar=tensor([0.0219, 0.0236, 0.0225, 0.0225, 0.0234, 0.0233, 0.0229, 0.0232], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:20:25,319 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293987.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:20:38,044 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.578e+02 2.282e+02 2.658e+02 3.259e+02 1.207e+03, threshold=5.315e+02, percent-clipped=3.0 2023-05-02 20:20:56,842 INFO [train.py:904] (5/8) Epoch 29, batch 9800, loss[loss=0.1671, simple_loss=0.2696, pruned_loss=0.03235, over 16307.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2579, pruned_loss=0.03339, over 3059275.39 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:21:22,942 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:27,500 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294020.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:47,351 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8665, 2.1353, 2.2337, 3.4122, 2.0669, 2.3616, 2.2731, 2.2508], device='cuda:5'), covar=tensor([0.1546, 0.3989, 0.3380, 0.0725, 0.4614, 0.2762, 0.3798, 0.3759], device='cuda:5'), in_proj_covar=tensor([0.0414, 0.0467, 0.0382, 0.0329, 0.0439, 0.0534, 0.0441, 0.0547], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:21:51,638 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6969, 4.6688, 4.4286, 3.8280, 4.5757, 1.8473, 4.3737, 4.1959], device='cuda:5'), covar=tensor([0.0082, 0.0078, 0.0209, 0.0267, 0.0098, 0.2862, 0.0119, 0.0255], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0170, 0.0207, 0.0178, 0.0185, 0.0214, 0.0196, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:22:29,334 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294048.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:22:39,772 INFO [train.py:904] (5/8) Epoch 29, batch 9850, loss[loss=0.152, simple_loss=0.2549, pruned_loss=0.02448, over 17285.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2585, pruned_loss=0.03283, over 3056895.64 frames. ], batch size: 52, lr: 2.30e-03, grad_scale: 16.0 2023-05-02 20:23:02,391 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:23:04,972 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 20:23:09,658 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:23:24,811 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 20:24:08,266 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4676, 2.0442, 1.7398, 1.7079, 2.2832, 1.8964, 1.7297, 2.3797], device='cuda:5'), covar=tensor([0.0233, 0.0505, 0.0650, 0.0614, 0.0343, 0.0468, 0.0207, 0.0299], device='cuda:5'), in_proj_covar=tensor([0.0219, 0.0236, 0.0225, 0.0225, 0.0234, 0.0233, 0.0228, 0.0232], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:24:14,911 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.054e+02 2.475e+02 2.961e+02 4.915e+02, threshold=4.950e+02, percent-clipped=0.0 2023-05-02 20:24:32,439 INFO [train.py:904] (5/8) Epoch 29, batch 9900, loss[loss=0.1694, simple_loss=0.2746, pruned_loss=0.03211, over 16345.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2591, pruned_loss=0.03277, over 3056920.34 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:25:36,186 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 20:26:27,782 INFO [train.py:904] (5/8) Epoch 29, batch 9950, loss[loss=0.1517, simple_loss=0.2506, pruned_loss=0.02635, over 16544.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2612, pruned_loss=0.03332, over 3057020.98 frames. ], batch size: 68, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:27:06,840 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:28:08,141 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.053e+02 2.347e+02 2.830e+02 4.908e+02, threshold=4.695e+02, percent-clipped=0.0 2023-05-02 20:28:27,683 INFO [train.py:904] (5/8) Epoch 29, batch 10000, loss[loss=0.155, simple_loss=0.2604, pruned_loss=0.02478, over 16869.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2602, pruned_loss=0.03313, over 3067981.18 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:28:40,009 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4073, 3.0921, 3.2907, 1.9249, 3.4461, 3.5428, 2.9003, 2.8677], device='cuda:5'), covar=tensor([0.0737, 0.0283, 0.0253, 0.1139, 0.0111, 0.0171, 0.0433, 0.0443], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0108, 0.0098, 0.0135, 0.0084, 0.0126, 0.0126, 0.0126], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 20:29:21,438 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294231.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:30:08,535 INFO [train.py:904] (5/8) Epoch 29, batch 10050, loss[loss=0.1618, simple_loss=0.265, pruned_loss=0.0293, over 16338.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2605, pruned_loss=0.03323, over 3080490.87 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:31:27,888 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.082e+02 2.498e+02 3.055e+02 5.920e+02, threshold=4.997e+02, percent-clipped=2.0 2023-05-02 20:31:35,347 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6252, 4.8040, 4.9890, 4.6838, 4.8226, 5.3114, 4.7951, 4.5136], device='cuda:5'), covar=tensor([0.1302, 0.1721, 0.1900, 0.2000, 0.2287, 0.0825, 0.1608, 0.2329], device='cuda:5'), in_proj_covar=tensor([0.0410, 0.0616, 0.0686, 0.0500, 0.0666, 0.0707, 0.0531, 0.0663], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 20:31:41,189 INFO [train.py:904] (5/8) Epoch 29, batch 10100, loss[loss=0.1627, simple_loss=0.2482, pruned_loss=0.03862, over 12800.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2606, pruned_loss=0.03333, over 3059924.34 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:31:59,158 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8720, 3.7829, 3.9428, 4.0228, 4.0883, 3.6950, 4.0732, 4.1374], device='cuda:5'), covar=tensor([0.1679, 0.1284, 0.1209, 0.0759, 0.0626, 0.2076, 0.0818, 0.0793], device='cuda:5'), in_proj_covar=tensor([0.0653, 0.0797, 0.0915, 0.0815, 0.0620, 0.0638, 0.0678, 0.0788], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:32:42,418 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:32:46,887 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294343.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:33:20,747 INFO [train.py:904] (5/8) Epoch 30, batch 0, loss[loss=0.1892, simple_loss=0.2902, pruned_loss=0.04415, over 17039.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2902, pruned_loss=0.04415, over 17039.00 frames. ], batch size: 50, lr: 2.26e-03, grad_scale: 8.0 2023-05-02 20:33:20,747 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 20:33:28,213 INFO [train.py:938] (5/8) Epoch 30, validation: loss=0.1426, simple_loss=0.2458, pruned_loss=0.01974, over 944034.00 frames. 2023-05-02 20:33:28,214 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 20:34:10,675 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0331, 3.0936, 2.5578, 4.6778, 3.6439, 4.2611, 1.7780, 3.3535], device='cuda:5'), covar=tensor([0.1390, 0.0690, 0.1367, 0.0180, 0.0199, 0.0441, 0.1714, 0.0752], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0178, 0.0198, 0.0199, 0.0200, 0.0214, 0.0209, 0.0196], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:34:28,811 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.426e+02 2.887e+02 3.382e+02 7.415e+02, threshold=5.774e+02, percent-clipped=4.0 2023-05-02 20:34:29,634 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294400.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:34:36,218 INFO [train.py:904] (5/8) Epoch 30, batch 50, loss[loss=0.1946, simple_loss=0.2791, pruned_loss=0.05502, over 15493.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2609, pruned_loss=0.04325, over 753796.48 frames. ], batch size: 190, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:35:00,554 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 20:35:02,126 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294423.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:35:43,894 INFO [train.py:904] (5/8) Epoch 30, batch 100, loss[loss=0.1534, simple_loss=0.2414, pruned_loss=0.03268, over 16823.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2608, pruned_loss=0.04262, over 1321305.66 frames. ], batch size: 42, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:36:24,637 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294484.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:32,221 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:46,207 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.392e+02 2.778e+02 3.445e+02 1.255e+03, threshold=5.557e+02, percent-clipped=5.0 2023-05-02 20:36:51,465 INFO [train.py:904] (5/8) Epoch 30, batch 150, loss[loss=0.1843, simple_loss=0.2817, pruned_loss=0.04344, over 17041.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2593, pruned_loss=0.04234, over 1763079.39 frames. ], batch size: 55, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:37:20,779 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294526.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:22,122 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5402, 2.3385, 1.8956, 2.1088, 2.6549, 2.4012, 2.4648, 2.7246], device='cuda:5'), covar=tensor([0.0299, 0.0476, 0.0611, 0.0541, 0.0294, 0.0429, 0.0246, 0.0346], device='cuda:5'), in_proj_covar=tensor([0.0228, 0.0243, 0.0232, 0.0233, 0.0241, 0.0240, 0.0236, 0.0240], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:37:38,714 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-02 20:37:43,815 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0938, 2.3121, 2.7611, 3.0796, 2.9763, 3.6276, 2.6545, 3.6140], device='cuda:5'), covar=tensor([0.0309, 0.0557, 0.0408, 0.0396, 0.0391, 0.0243, 0.0534, 0.0221], device='cuda:5'), in_proj_covar=tensor([0.0194, 0.0195, 0.0184, 0.0188, 0.0207, 0.0163, 0.0202, 0.0164], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:37:53,399 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:59,957 INFO [train.py:904] (5/8) Epoch 30, batch 200, loss[loss=0.1505, simple_loss=0.2429, pruned_loss=0.02906, over 17212.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2585, pruned_loss=0.0414, over 2116716.61 frames. ], batch size: 44, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:38:23,470 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:39:01,689 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.189e+02 2.574e+02 3.148e+02 2.009e+03, threshold=5.148e+02, percent-clipped=2.0 2023-05-02 20:39:06,664 INFO [train.py:904] (5/8) Epoch 30, batch 250, loss[loss=0.1264, simple_loss=0.2091, pruned_loss=0.02188, over 16792.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2561, pruned_loss=0.04051, over 2385148.10 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:39:43,094 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2163, 3.9744, 4.3795, 2.3592, 4.5782, 4.7314, 3.4907, 3.7354], device='cuda:5'), covar=tensor([0.0706, 0.0293, 0.0297, 0.1138, 0.0106, 0.0165, 0.0447, 0.0391], device='cuda:5'), in_proj_covar=tensor([0.0147, 0.0109, 0.0100, 0.0137, 0.0086, 0.0129, 0.0128, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 20:39:45,446 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:01,651 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:15,324 INFO [train.py:904] (5/8) Epoch 30, batch 300, loss[loss=0.185, simple_loss=0.257, pruned_loss=0.05652, over 16489.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2536, pruned_loss=0.03965, over 2595994.62 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:41:06,782 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:12,847 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:17,966 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7465, 6.1390, 5.8578, 5.9453, 5.5036, 5.6053, 5.4939, 6.2455], device='cuda:5'), covar=tensor([0.1453, 0.0937, 0.1114, 0.0922, 0.0960, 0.0636, 0.1362, 0.0949], device='cuda:5'), in_proj_covar=tensor([0.0716, 0.0862, 0.0710, 0.0671, 0.0552, 0.0548, 0.0724, 0.0677], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:41:19,995 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.370e+02 1.963e+02 2.186e+02 2.662e+02 5.332e+02, threshold=4.373e+02, percent-clipped=1.0 2023-05-02 20:41:25,726 INFO [train.py:904] (5/8) Epoch 30, batch 350, loss[loss=0.1669, simple_loss=0.251, pruned_loss=0.04139, over 12559.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2513, pruned_loss=0.03848, over 2757315.67 frames. ], batch size: 247, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:42:33,015 INFO [train.py:904] (5/8) Epoch 30, batch 400, loss[loss=0.1654, simple_loss=0.2404, pruned_loss=0.04515, over 16789.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2512, pruned_loss=0.03845, over 2887393.21 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:43:01,596 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6243, 3.7368, 3.9364, 2.8027, 3.5689, 4.0010, 3.6738, 2.3461], device='cuda:5'), covar=tensor([0.0555, 0.0254, 0.0070, 0.0433, 0.0142, 0.0128, 0.0121, 0.0523], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 20:43:06,909 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294779.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:43:29,495 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6275, 3.4671, 4.1378, 2.2640, 3.2196, 2.5902, 3.9584, 3.7450], device='cuda:5'), covar=tensor([0.0215, 0.1045, 0.0453, 0.2115, 0.0839, 0.0980, 0.0550, 0.1188], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0157, 0.0147, 0.0132, 0.0145, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:43:34,531 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 2.158e+02 2.521e+02 2.985e+02 1.602e+03, threshold=5.041e+02, percent-clipped=2.0 2023-05-02 20:43:41,856 INFO [train.py:904] (5/8) Epoch 30, batch 450, loss[loss=0.1699, simple_loss=0.2661, pruned_loss=0.03689, over 17032.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2501, pruned_loss=0.03754, over 2981147.36 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:44:12,164 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294826.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:36,123 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:49,600 INFO [train.py:904] (5/8) Epoch 30, batch 500, loss[loss=0.1378, simple_loss=0.2202, pruned_loss=0.0277, over 16998.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2485, pruned_loss=0.03683, over 3059736.00 frames. ], batch size: 41, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:44:53,056 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9106, 2.8812, 3.2188, 2.1550, 2.8067, 2.1986, 3.4245, 3.2951], device='cuda:5'), covar=tensor([0.0235, 0.1068, 0.0625, 0.2055, 0.0923, 0.1088, 0.0548, 0.0937], device='cuda:5'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0157, 0.0147, 0.0132, 0.0145, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:45:16,887 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5484, 2.4405, 2.4277, 4.2828, 2.4664, 2.7736, 2.5199, 2.6127], device='cuda:5'), covar=tensor([0.1331, 0.3813, 0.3402, 0.0611, 0.4205, 0.2841, 0.3740, 0.3985], device='cuda:5'), in_proj_covar=tensor([0.0425, 0.0478, 0.0391, 0.0339, 0.0448, 0.0547, 0.0451, 0.0560], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:45:17,741 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:45:52,587 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.033e+02 2.399e+02 2.919e+02 5.551e+02, threshold=4.798e+02, percent-clipped=2.0 2023-05-02 20:45:57,299 INFO [train.py:904] (5/8) Epoch 30, batch 550, loss[loss=0.171, simple_loss=0.2552, pruned_loss=0.04342, over 16468.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2478, pruned_loss=0.03688, over 3119052.78 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:46:29,054 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:46:50,981 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.6855, 6.0740, 5.8323, 5.8732, 5.3858, 5.5026, 5.4390, 6.1789], device='cuda:5'), covar=tensor([0.1537, 0.0979, 0.1152, 0.0977, 0.1103, 0.0677, 0.1310, 0.0985], device='cuda:5'), in_proj_covar=tensor([0.0723, 0.0869, 0.0716, 0.0678, 0.0556, 0.0554, 0.0732, 0.0684], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:47:05,471 INFO [train.py:904] (5/8) Epoch 30, batch 600, loss[loss=0.1616, simple_loss=0.2375, pruned_loss=0.04286, over 16825.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2472, pruned_loss=0.03682, over 3164039.30 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:47:52,965 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:01,941 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:07,358 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.157e+02 2.572e+02 3.014e+02 4.443e+02, threshold=5.143e+02, percent-clipped=0.0 2023-05-02 20:48:13,124 INFO [train.py:904] (5/8) Epoch 30, batch 650, loss[loss=0.1296, simple_loss=0.2125, pruned_loss=0.02329, over 16772.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2459, pruned_loss=0.03667, over 3194305.24 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:48:42,002 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9883, 5.3364, 5.1173, 5.1168, 4.8275, 4.8162, 4.7413, 5.4384], device='cuda:5'), covar=tensor([0.1321, 0.0921, 0.1034, 0.0910, 0.0877, 0.1071, 0.1372, 0.0902], device='cuda:5'), in_proj_covar=tensor([0.0723, 0.0868, 0.0716, 0.0677, 0.0555, 0.0553, 0.0732, 0.0684], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 20:49:04,987 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295043.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:15,928 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295050.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 20:49:20,788 INFO [train.py:904] (5/8) Epoch 30, batch 700, loss[loss=0.1729, simple_loss=0.2545, pruned_loss=0.04563, over 16855.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2449, pruned_loss=0.03658, over 3213341.76 frames. ], batch size: 109, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:49:39,365 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:54,495 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:50:03,936 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 20:50:23,317 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 2.027e+02 2.322e+02 2.790e+02 6.383e+02, threshold=4.644e+02, percent-clipped=2.0 2023-05-02 20:50:27,900 INFO [train.py:904] (5/8) Epoch 30, batch 750, loss[loss=0.1752, simple_loss=0.2724, pruned_loss=0.03903, over 17073.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2458, pruned_loss=0.03659, over 3232326.71 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:50:56,747 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2210, 4.6143, 4.5871, 3.4523, 3.8685, 4.5445, 4.0518, 2.7754], device='cuda:5'), covar=tensor([0.0475, 0.0067, 0.0055, 0.0362, 0.0144, 0.0105, 0.0101, 0.0504], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0137, 0.0104, 0.0117, 0.0100, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 20:51:00,103 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:03,378 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:24,568 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:38,205 INFO [train.py:904] (5/8) Epoch 30, batch 800, loss[loss=0.1503, simple_loss=0.2472, pruned_loss=0.02676, over 17257.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.246, pruned_loss=0.03675, over 3243713.91 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:51:39,913 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8897, 2.1146, 2.5531, 2.8977, 2.7947, 3.4027, 2.3177, 3.3338], device='cuda:5'), covar=tensor([0.0324, 0.0597, 0.0392, 0.0408, 0.0423, 0.0230, 0.0592, 0.0221], device='cuda:5'), in_proj_covar=tensor([0.0199, 0.0199, 0.0187, 0.0193, 0.0212, 0.0168, 0.0205, 0.0168], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-05-02 20:51:49,423 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-05-02 20:51:51,588 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6988, 3.8597, 2.6023, 4.4249, 3.1599, 4.3851, 2.7326, 3.2985], device='cuda:5'), covar=tensor([0.0392, 0.0470, 0.1653, 0.0413, 0.0849, 0.0611, 0.1536, 0.0785], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0175, 0.0181, 0.0222, 0.0207, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:52:27,619 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3108, 4.7648, 4.7258, 3.5846, 4.0064, 4.6740, 4.1300, 3.1283], device='cuda:5'), covar=tensor([0.0441, 0.0061, 0.0047, 0.0342, 0.0148, 0.0116, 0.0101, 0.0412], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 20:52:29,871 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:52:42,515 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.068e+02 2.393e+02 2.794e+02 7.464e+02, threshold=4.786e+02, percent-clipped=2.0 2023-05-02 20:52:44,709 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4754, 3.5775, 4.1187, 2.2433, 3.2388, 2.3620, 3.9743, 3.8614], device='cuda:5'), covar=tensor([0.0278, 0.1107, 0.0460, 0.2252, 0.0902, 0.1118, 0.0548, 0.1165], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0158, 0.0148, 0.0133, 0.0145, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 20:52:48,759 INFO [train.py:904] (5/8) Epoch 30, batch 850, loss[loss=0.1656, simple_loss=0.273, pruned_loss=0.02908, over 17130.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2457, pruned_loss=0.03615, over 3270836.38 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:53:18,818 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295225.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:21,322 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295227.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:56,733 INFO [train.py:904] (5/8) Epoch 30, batch 900, loss[loss=0.1696, simple_loss=0.2475, pruned_loss=0.04584, over 12553.00 frames. ], tot_loss[loss=0.1581, simple_loss=0.2447, pruned_loss=0.03579, over 3281683.34 frames. ], batch size: 248, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:54:27,486 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:31,273 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5743, 3.6320, 3.3817, 2.9819, 3.2659, 3.5114, 3.3474, 3.3635], device='cuda:5'), covar=tensor([0.0577, 0.0620, 0.0317, 0.0303, 0.0505, 0.0453, 0.1358, 0.0493], device='cuda:5'), in_proj_covar=tensor([0.0320, 0.0483, 0.0373, 0.0376, 0.0370, 0.0432, 0.0258, 0.0448], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 20:54:34,108 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295279.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:41,840 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8536, 4.3410, 4.3799, 3.1414, 3.6001, 4.3262, 3.9188, 2.6633], device='cuda:5'), covar=tensor([0.0508, 0.0081, 0.0053, 0.0390, 0.0165, 0.0113, 0.0107, 0.0489], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0136, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 20:54:43,629 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295286.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:55:02,822 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.063e+02 2.465e+02 2.923e+02 4.814e+02, threshold=4.929e+02, percent-clipped=1.0 2023-05-02 20:55:08,082 INFO [train.py:904] (5/8) Epoch 30, batch 950, loss[loss=0.1464, simple_loss=0.2304, pruned_loss=0.03123, over 16800.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.245, pruned_loss=0.03609, over 3289002.24 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:55:48,118 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3583, 3.4051, 3.9321, 2.0804, 3.2930, 2.4713, 3.8349, 3.7427], device='cuda:5'), covar=tensor([0.0292, 0.1064, 0.0501, 0.2246, 0.0847, 0.1016, 0.0588, 0.1099], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0158, 0.0149, 0.0133, 0.0146, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 20:55:49,189 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:55:59,246 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295340.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:06,376 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295345.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 20:56:17,479 INFO [train.py:904] (5/8) Epoch 30, batch 1000, loss[loss=0.1608, simple_loss=0.2361, pruned_loss=0.04277, over 16913.00 frames. ], tot_loss[loss=0.1579, simple_loss=0.244, pruned_loss=0.03588, over 3298599.50 frames. ], batch size: 116, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:56:17,939 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9073, 4.1716, 2.4567, 4.6431, 3.3090, 4.5987, 2.8012, 3.4935], device='cuda:5'), covar=tensor([0.0370, 0.0403, 0.1758, 0.0332, 0.0803, 0.0599, 0.1552, 0.0719], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0176, 0.0181, 0.0223, 0.0207, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 20:56:24,302 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:39,518 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 20:56:53,442 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 20:57:12,884 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295394.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:20,047 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.023e+02 2.384e+02 2.929e+02 5.067e+02, threshold=4.769e+02, percent-clipped=1.0 2023-05-02 20:57:26,455 INFO [train.py:904] (5/8) Epoch 30, batch 1050, loss[loss=0.1421, simple_loss=0.2374, pruned_loss=0.02337, over 17146.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2436, pruned_loss=0.03578, over 3304191.03 frames. ], batch size: 47, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:57:48,348 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295420.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:54,265 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:58:36,218 INFO [train.py:904] (5/8) Epoch 30, batch 1100, loss[loss=0.1527, simple_loss=0.2374, pruned_loss=0.03399, over 16486.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2432, pruned_loss=0.03554, over 3310623.76 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:59:32,063 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6259, 3.7497, 3.9375, 2.8503, 3.5914, 4.0118, 3.7219, 2.3914], device='cuda:5'), covar=tensor([0.0556, 0.0238, 0.0065, 0.0405, 0.0128, 0.0106, 0.0108, 0.0548], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0136, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 20:59:33,160 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0346, 4.0949, 4.2776, 4.2996, 4.3288, 4.0981, 4.1064, 4.0524], device='cuda:5'), covar=tensor([0.0449, 0.0748, 0.0595, 0.0451, 0.0609, 0.0567, 0.0807, 0.0652], device='cuda:5'), in_proj_covar=tensor([0.0446, 0.0509, 0.0488, 0.0450, 0.0535, 0.0517, 0.0592, 0.0415], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 20:59:38,543 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.061e+02 2.413e+02 2.897e+02 1.352e+03, threshold=4.827e+02, percent-clipped=9.0 2023-05-02 20:59:43,315 INFO [train.py:904] (5/8) Epoch 30, batch 1150, loss[loss=0.1445, simple_loss=0.2265, pruned_loss=0.0312, over 16677.00 frames. ], tot_loss[loss=0.1563, simple_loss=0.2425, pruned_loss=0.03501, over 3308018.09 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:00:30,579 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8342, 4.0468, 2.7092, 4.5996, 3.2925, 4.5960, 2.8983, 3.5094], device='cuda:5'), covar=tensor([0.0369, 0.0420, 0.1606, 0.0391, 0.0814, 0.0531, 0.1415, 0.0675], device='cuda:5'), in_proj_covar=tensor([0.0178, 0.0182, 0.0197, 0.0175, 0.0181, 0.0222, 0.0207, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:00:41,659 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-05-02 21:00:52,306 INFO [train.py:904] (5/8) Epoch 30, batch 1200, loss[loss=0.1764, simple_loss=0.2685, pruned_loss=0.04211, over 17091.00 frames. ], tot_loss[loss=0.1556, simple_loss=0.2415, pruned_loss=0.03484, over 3315682.49 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:00:54,984 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:00:58,028 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295558.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:01:29,885 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295581.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:01:55,308 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.025e+02 2.431e+02 2.981e+02 5.369e+02, threshold=4.863e+02, percent-clipped=2.0 2023-05-02 21:02:00,809 INFO [train.py:904] (5/8) Epoch 30, batch 1250, loss[loss=0.1721, simple_loss=0.2501, pruned_loss=0.04707, over 16436.00 frames. ], tot_loss[loss=0.1562, simple_loss=0.2423, pruned_loss=0.0351, over 3324535.80 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:02:19,210 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295617.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:02:21,405 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295619.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:02:34,271 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295629.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:02:39,634 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6570, 2.6162, 2.2518, 2.4456, 2.9291, 2.6558, 3.2628, 3.2136], device='cuda:5'), covar=tensor([0.0253, 0.0636, 0.0811, 0.0675, 0.0428, 0.0613, 0.0349, 0.0412], device='cuda:5'), in_proj_covar=tensor([0.0239, 0.0252, 0.0239, 0.0240, 0.0250, 0.0249, 0.0247, 0.0250], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:02:43,117 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2534, 4.2375, 4.4171, 4.2017, 4.3170, 4.8271, 4.3563, 4.0126], device='cuda:5'), covar=tensor([0.2153, 0.2465, 0.2714, 0.2517, 0.2855, 0.1404, 0.1929, 0.2986], device='cuda:5'), in_proj_covar=tensor([0.0435, 0.0653, 0.0727, 0.0528, 0.0704, 0.0745, 0.0558, 0.0702], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:02:43,128 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:02:46,382 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.6248, 2.6586, 2.3196, 2.4008, 2.9569, 2.6185, 3.1906, 3.1068], device='cuda:5'), covar=tensor([0.0188, 0.0566, 0.0674, 0.0641, 0.0396, 0.0571, 0.0330, 0.0378], device='cuda:5'), in_proj_covar=tensor([0.0239, 0.0252, 0.0239, 0.0240, 0.0250, 0.0249, 0.0247, 0.0250], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:02:58,112 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295645.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:08,457 INFO [train.py:904] (5/8) Epoch 30, batch 1300, loss[loss=0.1412, simple_loss=0.2226, pruned_loss=0.02991, over 16810.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2426, pruned_loss=0.0351, over 3315533.40 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:03:52,676 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295685.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:57,775 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:59,102 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295690.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:04:02,885 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295693.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:11,939 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.150e+02 2.443e+02 2.917e+02 6.484e+02, threshold=4.887e+02, percent-clipped=2.0 2023-05-02 21:04:18,066 INFO [train.py:904] (5/8) Epoch 30, batch 1350, loss[loss=0.1638, simple_loss=0.2439, pruned_loss=0.0419, over 16748.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2426, pruned_loss=0.0348, over 3290775.83 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:04:29,328 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3915, 3.5292, 4.0144, 2.2079, 3.2777, 2.5040, 3.8002, 3.7150], device='cuda:5'), covar=tensor([0.0293, 0.1056, 0.0490, 0.2212, 0.0842, 0.1085, 0.0617, 0.1124], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 21:04:33,156 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:46,305 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295724.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:17,249 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295746.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:26,459 INFO [train.py:904] (5/8) Epoch 30, batch 1400, loss[loss=0.1792, simple_loss=0.2538, pruned_loss=0.05233, over 12599.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2428, pruned_loss=0.035, over 3299466.68 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:05:52,149 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295772.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:06:20,382 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-05-02 21:06:30,030 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 2.078e+02 2.399e+02 2.795e+02 5.908e+02, threshold=4.799e+02, percent-clipped=2.0 2023-05-02 21:06:35,127 INFO [train.py:904] (5/8) Epoch 30, batch 1450, loss[loss=0.1555, simple_loss=0.2291, pruned_loss=0.04097, over 16478.00 frames. ], tot_loss[loss=0.1557, simple_loss=0.2421, pruned_loss=0.03467, over 3304787.04 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:06:57,740 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295820.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:07:15,634 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4504, 3.9843, 4.4315, 2.4013, 4.6427, 4.7368, 3.5062, 3.7128], device='cuda:5'), covar=tensor([0.0614, 0.0281, 0.0226, 0.1140, 0.0084, 0.0173, 0.0426, 0.0422], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0140, 0.0088, 0.0134, 0.0131, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:07:18,831 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-05-02 21:07:44,322 INFO [train.py:904] (5/8) Epoch 30, batch 1500, loss[loss=0.1777, simple_loss=0.2464, pruned_loss=0.05449, over 16903.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2421, pruned_loss=0.03503, over 3313980.08 frames. ], batch size: 90, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:08:06,491 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 21:08:11,985 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:22,737 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:22,771 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:47,915 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.156e+02 2.462e+02 3.031e+02 5.873e+02, threshold=4.924e+02, percent-clipped=1.0 2023-05-02 21:08:54,106 INFO [train.py:904] (5/8) Epoch 30, batch 1550, loss[loss=0.1629, simple_loss=0.2466, pruned_loss=0.0396, over 12556.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2427, pruned_loss=0.03623, over 3295460.63 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:08:55,847 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7077, 3.3586, 3.7296, 2.0527, 3.7921, 3.8370, 3.1556, 2.8456], device='cuda:5'), covar=tensor([0.0713, 0.0269, 0.0213, 0.1126, 0.0125, 0.0217, 0.0422, 0.0458], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0113, 0.0104, 0.0140, 0.0089, 0.0135, 0.0132, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:09:05,574 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295912.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:09:08,353 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295914.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:09:29,569 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295929.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:09:36,317 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295934.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:09:37,392 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295935.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:09:39,121 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4927, 4.3160, 4.5356, 4.6596, 4.7807, 4.3347, 4.6517, 4.7709], device='cuda:5'), covar=tensor([0.1665, 0.1467, 0.1448, 0.0768, 0.0592, 0.1182, 0.2398, 0.0754], device='cuda:5'), in_proj_covar=tensor([0.0709, 0.0861, 0.0994, 0.0879, 0.0667, 0.0688, 0.0728, 0.0851], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:09:59,299 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3334, 5.7038, 5.4491, 5.5326, 5.1770, 5.1635, 5.0872, 5.8474], device='cuda:5'), covar=tensor([0.1572, 0.1015, 0.1181, 0.0902, 0.1015, 0.0820, 0.1339, 0.0922], device='cuda:5'), in_proj_covar=tensor([0.0743, 0.0895, 0.0737, 0.0698, 0.0572, 0.0568, 0.0753, 0.0706], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:10:02,944 INFO [train.py:904] (5/8) Epoch 30, batch 1600, loss[loss=0.1907, simple_loss=0.2792, pruned_loss=0.05104, over 16555.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2455, pruned_loss=0.03751, over 3298038.94 frames. ], batch size: 68, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:10:36,722 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7586, 4.3692, 2.7440, 2.3178, 2.4011, 2.4325, 4.5983, 3.2564], device='cuda:5'), covar=tensor([0.3366, 0.0619, 0.2336, 0.3359, 0.3406, 0.2584, 0.0511, 0.1862], device='cuda:5'), in_proj_covar=tensor([0.0340, 0.0278, 0.0317, 0.0331, 0.0309, 0.0283, 0.0307, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:10:42,972 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:10:45,979 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295985.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:10:49,443 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9551, 5.2870, 5.0518, 5.0891, 4.8180, 4.7699, 4.7732, 5.3951], device='cuda:5'), covar=tensor([0.1349, 0.0931, 0.1090, 0.0939, 0.0870, 0.1115, 0.1313, 0.0924], device='cuda:5'), in_proj_covar=tensor([0.0743, 0.0894, 0.0737, 0.0697, 0.0572, 0.0568, 0.0753, 0.0705], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:10:51,156 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:11:05,446 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.204e+02 2.497e+02 2.999e+02 5.396e+02, threshold=4.994e+02, percent-clipped=1.0 2023-05-02 21:11:14,074 INFO [train.py:904] (5/8) Epoch 30, batch 1650, loss[loss=0.177, simple_loss=0.2541, pruned_loss=0.04994, over 16370.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2466, pruned_loss=0.03777, over 3295925.04 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:11:30,109 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296015.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:00,955 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:06,279 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:16,959 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:18,412 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-02 21:12:24,450 INFO [train.py:904] (5/8) Epoch 30, batch 1700, loss[loss=0.2528, simple_loss=0.325, pruned_loss=0.0903, over 12035.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2489, pruned_loss=0.03876, over 3283795.41 frames. ], batch size: 247, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:12:36,863 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296063.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:48,989 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8820, 2.8979, 2.6584, 5.0821, 3.9820, 4.4210, 1.7374, 3.2601], device='cuda:5'), covar=tensor([0.1406, 0.0866, 0.1348, 0.0214, 0.0217, 0.0394, 0.1708, 0.0786], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0184, 0.0204, 0.0209, 0.0208, 0.0221, 0.0214, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:13:26,988 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.214e+02 2.660e+02 3.287e+02 5.248e+02, threshold=5.320e+02, percent-clipped=2.0 2023-05-02 21:13:32,343 INFO [train.py:904] (5/8) Epoch 30, batch 1750, loss[loss=0.1684, simple_loss=0.2708, pruned_loss=0.03303, over 17030.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.25, pruned_loss=0.03872, over 3292047.01 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:13:40,777 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:13:42,353 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-05-02 21:14:24,239 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9375, 2.9641, 2.7096, 5.0320, 3.9657, 4.3282, 1.6702, 3.2340], device='cuda:5'), covar=tensor([0.1359, 0.0841, 0.1227, 0.0221, 0.0236, 0.0464, 0.1694, 0.0795], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0184, 0.0203, 0.0209, 0.0208, 0.0221, 0.0214, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:14:32,738 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8342, 2.8113, 2.6781, 4.7935, 3.7229, 4.2190, 1.7945, 3.1182], device='cuda:5'), covar=tensor([0.1426, 0.0885, 0.1292, 0.0236, 0.0188, 0.0428, 0.1683, 0.0842], device='cuda:5'), in_proj_covar=tensor([0.0176, 0.0185, 0.0203, 0.0209, 0.0208, 0.0221, 0.0214, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:14:41,098 INFO [train.py:904] (5/8) Epoch 30, batch 1800, loss[loss=0.1557, simple_loss=0.2501, pruned_loss=0.03061, over 17271.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2512, pruned_loss=0.03859, over 3293092.09 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:15:12,950 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:15:18,253 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 21:15:47,287 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.165e+02 2.480e+02 2.953e+02 5.408e+02, threshold=4.959e+02, percent-clipped=1.0 2023-05-02 21:15:51,606 INFO [train.py:904] (5/8) Epoch 30, batch 1850, loss[loss=0.1552, simple_loss=0.242, pruned_loss=0.03422, over 16741.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2518, pruned_loss=0.03841, over 3298690.50 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:16:03,846 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296212.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:16:06,033 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296214.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:16:28,004 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296229.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:16:57,139 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.4320, 5.8271, 5.5413, 5.6130, 5.2802, 5.3521, 5.1281, 5.9638], device='cuda:5'), covar=tensor([0.1707, 0.0961, 0.1361, 0.0970, 0.0999, 0.0700, 0.1410, 0.0909], device='cuda:5'), in_proj_covar=tensor([0.0740, 0.0891, 0.0734, 0.0695, 0.0570, 0.0564, 0.0750, 0.0702], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:17:02,021 INFO [train.py:904] (5/8) Epoch 30, batch 1900, loss[loss=0.1618, simple_loss=0.2591, pruned_loss=0.03228, over 16679.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2513, pruned_loss=0.03799, over 3295836.54 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:17:05,234 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-05-02 21:17:09,980 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296260.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:17:13,525 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296262.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:17:42,637 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 21:17:45,551 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296285.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:17:56,107 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 21:18:06,423 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.062e+02 2.316e+02 2.814e+02 5.874e+02, threshold=4.633e+02, percent-clipped=2.0 2023-05-02 21:18:10,613 INFO [train.py:904] (5/8) Epoch 30, batch 1950, loss[loss=0.1611, simple_loss=0.2425, pruned_loss=0.03983, over 16801.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2512, pruned_loss=0.03772, over 3303515.56 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:18:27,055 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:18:52,566 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296333.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:19:02,342 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296341.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:19:10,957 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8929, 2.2767, 2.6897, 2.9555, 2.8592, 3.4838, 2.5217, 3.4594], device='cuda:5'), covar=tensor([0.0360, 0.0643, 0.0393, 0.0429, 0.0430, 0.0266, 0.0579, 0.0240], device='cuda:5'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0196, 0.0214, 0.0171, 0.0207, 0.0171], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-05-02 21:19:21,058 INFO [train.py:904] (5/8) Epoch 30, batch 2000, loss[loss=0.1747, simple_loss=0.2686, pruned_loss=0.04037, over 15460.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2512, pruned_loss=0.03722, over 3306863.05 frames. ], batch size: 190, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:19:37,535 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 21:19:51,152 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:01,465 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4112, 3.4008, 2.2040, 3.6789, 2.8074, 3.6321, 2.1942, 2.8317], device='cuda:5'), covar=tensor([0.0314, 0.0622, 0.1562, 0.0386, 0.0764, 0.0853, 0.1611, 0.0765], device='cuda:5'), in_proj_covar=tensor([0.0180, 0.0186, 0.0198, 0.0179, 0.0183, 0.0226, 0.0209, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:20:09,551 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296389.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:23,082 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3841, 5.3667, 5.1334, 4.6176, 5.1764, 2.0502, 4.9615, 5.0213], device='cuda:5'), covar=tensor([0.0094, 0.0090, 0.0224, 0.0422, 0.0121, 0.2833, 0.0147, 0.0243], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0179, 0.0217, 0.0188, 0.0196, 0.0223, 0.0207, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:20:25,431 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.330e+02 2.133e+02 2.447e+02 2.850e+02 4.858e+02, threshold=4.894e+02, percent-clipped=1.0 2023-05-02 21:20:29,574 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9161, 2.4520, 1.8400, 2.0347, 2.7755, 2.5085, 2.8886, 2.8879], device='cuda:5'), covar=tensor([0.0285, 0.0557, 0.0838, 0.0709, 0.0362, 0.0522, 0.0298, 0.0418], device='cuda:5'), in_proj_covar=tensor([0.0241, 0.0252, 0.0240, 0.0242, 0.0252, 0.0250, 0.0249, 0.0252], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:20:30,263 INFO [train.py:904] (5/8) Epoch 30, batch 2050, loss[loss=0.168, simple_loss=0.2523, pruned_loss=0.04183, over 16921.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2511, pruned_loss=0.03809, over 3313109.20 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:20:32,304 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:12,429 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:39,729 INFO [train.py:904] (5/8) Epoch 30, batch 2100, loss[loss=0.1762, simple_loss=0.2601, pruned_loss=0.04614, over 16745.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2527, pruned_loss=0.03835, over 3321711.10 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:21:43,818 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2139, 5.1829, 4.9465, 4.3295, 5.0100, 1.8233, 4.7577, 4.7812], device='cuda:5'), covar=tensor([0.0114, 0.0099, 0.0238, 0.0490, 0.0134, 0.3192, 0.0165, 0.0276], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0179, 0.0217, 0.0187, 0.0195, 0.0222, 0.0207, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:22:09,628 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296476.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:33,201 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:36,826 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:39,134 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296497.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:46,677 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.431e+02 2.088e+02 2.487e+02 2.838e+02 5.578e+02, threshold=4.974e+02, percent-clipped=1.0 2023-05-02 21:22:48,903 INFO [train.py:904] (5/8) Epoch 30, batch 2150, loss[loss=0.1645, simple_loss=0.2618, pruned_loss=0.03357, over 16489.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2523, pruned_loss=0.03806, over 3332238.39 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:23:03,700 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:16,224 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296524.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:23,523 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296529.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:23:35,303 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9943, 4.0613, 4.2801, 4.2748, 4.3314, 4.0843, 4.1131, 4.0042], device='cuda:5'), covar=tensor([0.0415, 0.0646, 0.0488, 0.0456, 0.0573, 0.0487, 0.0812, 0.0674], device='cuda:5'), in_proj_covar=tensor([0.0448, 0.0510, 0.0487, 0.0451, 0.0534, 0.0517, 0.0593, 0.0415], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 21:23:43,109 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8733, 4.9909, 5.1825, 4.9574, 4.9993, 5.6389, 5.1334, 4.8169], device='cuda:5'), covar=tensor([0.1350, 0.2211, 0.2348, 0.2222, 0.2413, 0.0953, 0.1745, 0.2400], device='cuda:5'), in_proj_covar=tensor([0.0438, 0.0654, 0.0731, 0.0532, 0.0710, 0.0749, 0.0562, 0.0706], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:23:56,825 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296553.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:23:57,540 INFO [train.py:904] (5/8) Epoch 30, batch 2200, loss[loss=0.1977, simple_loss=0.2803, pruned_loss=0.05755, over 16775.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2526, pruned_loss=0.0387, over 3329127.72 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:24:04,175 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:13,108 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2380, 2.4648, 2.4140, 4.0094, 2.3597, 2.7590, 2.4885, 2.5873], device='cuda:5'), covar=tensor([0.1589, 0.3662, 0.3276, 0.0715, 0.4063, 0.2650, 0.3703, 0.3674], device='cuda:5'), in_proj_covar=tensor([0.0431, 0.0484, 0.0395, 0.0346, 0.0451, 0.0555, 0.0457, 0.0568], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:24:27,027 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296575.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:30,755 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:25:04,059 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.151e+02 2.512e+02 3.134e+02 5.584e+02, threshold=5.024e+02, percent-clipped=2.0 2023-05-02 21:25:06,332 INFO [train.py:904] (5/8) Epoch 30, batch 2250, loss[loss=0.1789, simple_loss=0.267, pruned_loss=0.0454, over 15587.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2531, pruned_loss=0.0387, over 3331858.13 frames. ], batch size: 191, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:25:20,055 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:25:31,200 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-05-02 21:26:17,734 INFO [train.py:904] (5/8) Epoch 30, batch 2300, loss[loss=0.1722, simple_loss=0.2651, pruned_loss=0.0397, over 16461.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2541, pruned_loss=0.03903, over 3332599.62 frames. ], batch size: 68, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:26:41,409 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:26:45,169 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:27:05,640 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:27:24,078 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 2.179e+02 2.509e+02 3.030e+02 4.955e+02, threshold=5.019e+02, percent-clipped=0.0 2023-05-02 21:27:27,335 INFO [train.py:904] (5/8) Epoch 30, batch 2350, loss[loss=0.1604, simple_loss=0.244, pruned_loss=0.03843, over 16800.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2545, pruned_loss=0.03913, over 3328950.62 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:27:28,863 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:30,549 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0176, 4.5841, 4.5282, 3.2587, 3.6386, 4.4916, 3.9047, 2.7390], device='cuda:5'), covar=tensor([0.0487, 0.0069, 0.0057, 0.0376, 0.0177, 0.0103, 0.0112, 0.0486], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0105, 0.0117, 0.0100, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 21:28:30,587 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:36,423 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296753.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:37,288 INFO [train.py:904] (5/8) Epoch 30, batch 2400, loss[loss=0.1515, simple_loss=0.2358, pruned_loss=0.03353, over 16960.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2541, pruned_loss=0.0384, over 3334150.47 frames. ], batch size: 41, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:28:54,529 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 21:29:27,156 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:29:43,968 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.201e+02 2.707e+02 3.099e+02 6.549e+02, threshold=5.413e+02, percent-clipped=3.0 2023-05-02 21:29:46,349 INFO [train.py:904] (5/8) Epoch 30, batch 2450, loss[loss=0.1587, simple_loss=0.2409, pruned_loss=0.03825, over 16778.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2546, pruned_loss=0.03857, over 3323436.49 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:30:10,977 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0413, 4.9653, 4.8940, 4.4290, 4.5827, 4.9326, 4.8341, 4.5652], device='cuda:5'), covar=tensor([0.0662, 0.0753, 0.0421, 0.0414, 0.1197, 0.0566, 0.0432, 0.0866], device='cuda:5'), in_proj_covar=tensor([0.0329, 0.0499, 0.0385, 0.0388, 0.0382, 0.0445, 0.0265, 0.0461], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:30:48,343 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296848.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:30:51,475 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9351, 4.4014, 3.2267, 2.4237, 2.8325, 2.7559, 4.7449, 3.7448], device='cuda:5'), covar=tensor([0.3034, 0.0572, 0.1855, 0.3413, 0.3099, 0.2256, 0.0386, 0.1441], device='cuda:5'), in_proj_covar=tensor([0.0339, 0.0278, 0.0317, 0.0330, 0.0309, 0.0282, 0.0307, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:30:55,455 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296853.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:30:56,382 INFO [train.py:904] (5/8) Epoch 30, batch 2500, loss[loss=0.1724, simple_loss=0.256, pruned_loss=0.04438, over 16442.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2553, pruned_loss=0.03893, over 3305601.34 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:31:19,669 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:31:23,969 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:31:59,230 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0158, 3.6527, 4.1668, 2.0551, 4.2698, 4.3997, 3.1744, 3.5216], device='cuda:5'), covar=tensor([0.0700, 0.0304, 0.0281, 0.1290, 0.0106, 0.0223, 0.0523, 0.0416], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:32:04,917 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.483e+02 2.203e+02 2.647e+02 3.147e+02 6.132e+02, threshold=5.294e+02, percent-clipped=2.0 2023-05-02 21:32:07,797 INFO [train.py:904] (5/8) Epoch 30, batch 2550, loss[loss=0.1722, simple_loss=0.2565, pruned_loss=0.04396, over 15580.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2557, pruned_loss=0.03899, over 3308089.14 frames. ], batch size: 191, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:32:49,101 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296934.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:32:49,206 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7919, 1.9712, 2.2051, 3.2530, 2.0198, 2.1480, 2.1267, 2.1206], device='cuda:5'), covar=tensor([0.1988, 0.4446, 0.3476, 0.1096, 0.5139, 0.3382, 0.4244, 0.4138], device='cuda:5'), in_proj_covar=tensor([0.0432, 0.0485, 0.0396, 0.0347, 0.0453, 0.0557, 0.0458, 0.0569], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:32:54,159 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296938.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:33:15,499 INFO [train.py:904] (5/8) Epoch 30, batch 2600, loss[loss=0.142, simple_loss=0.2346, pruned_loss=0.02474, over 16862.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2555, pruned_loss=0.03866, over 3309911.26 frames. ], batch size: 42, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:33:36,747 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:33:39,734 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:18,412 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:22,260 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 2.007e+02 2.417e+02 2.925e+02 5.072e+02, threshold=4.833e+02, percent-clipped=0.0 2023-05-02 21:34:24,415 INFO [train.py:904] (5/8) Epoch 30, batch 2650, loss[loss=0.1594, simple_loss=0.2493, pruned_loss=0.03472, over 16770.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2562, pruned_loss=0.03836, over 3318903.97 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:34:46,004 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297019.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:57,924 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 21:35:13,109 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5167, 3.8299, 3.9781, 2.6377, 3.6035, 4.0420, 3.6556, 2.4436], device='cuda:5'), covar=tensor([0.0578, 0.0202, 0.0069, 0.0479, 0.0132, 0.0114, 0.0117, 0.0517], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0137, 0.0105, 0.0117, 0.0100, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 21:35:20,545 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297044.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:35:34,429 INFO [train.py:904] (5/8) Epoch 30, batch 2700, loss[loss=0.1483, simple_loss=0.239, pruned_loss=0.02882, over 16824.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2554, pruned_loss=0.03715, over 3327416.14 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:36:20,075 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.9243, 4.1045, 2.7236, 4.7273, 3.3421, 4.6727, 2.8436, 3.4683], device='cuda:5'), covar=tensor([0.0326, 0.0407, 0.1605, 0.0349, 0.0842, 0.0495, 0.1537, 0.0730], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0186, 0.0200, 0.0180, 0.0185, 0.0227, 0.0210, 0.0187], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:36:24,030 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297090.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:36:42,011 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.120e+02 2.454e+02 2.893e+02 5.534e+02, threshold=4.909e+02, percent-clipped=1.0 2023-05-02 21:36:43,154 INFO [train.py:904] (5/8) Epoch 30, batch 2750, loss[loss=0.1558, simple_loss=0.258, pruned_loss=0.02674, over 17043.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2554, pruned_loss=0.03679, over 3336456.67 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:37:30,682 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:44,383 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297148.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:50,750 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:51,592 INFO [train.py:904] (5/8) Epoch 30, batch 2800, loss[loss=0.1692, simple_loss=0.2567, pruned_loss=0.04082, over 16776.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2552, pruned_loss=0.037, over 3328926.16 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:38:14,244 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:26,468 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 21:38:49,782 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:57,321 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:00,137 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.070e+02 2.512e+02 3.097e+02 5.652e+02, threshold=5.024e+02, percent-clipped=1.0 2023-05-02 21:39:01,288 INFO [train.py:904] (5/8) Epoch 30, batch 2850, loss[loss=0.1568, simple_loss=0.2471, pruned_loss=0.03323, over 16837.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2549, pruned_loss=0.03672, over 3331603.75 frames. ], batch size: 42, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:39:15,431 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1811, 3.3216, 3.4587, 2.2835, 3.1988, 3.5679, 3.2977, 2.1224], device='cuda:5'), covar=tensor([0.0527, 0.0151, 0.0070, 0.0453, 0.0140, 0.0101, 0.0113, 0.0487], device='cuda:5'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0135, 0.0104, 0.0116, 0.0099, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 21:39:21,722 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:26,930 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:37,169 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297229.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:40:10,547 INFO [train.py:904] (5/8) Epoch 30, batch 2900, loss[loss=0.1656, simple_loss=0.2657, pruned_loss=0.03278, over 16707.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2538, pruned_loss=0.03746, over 3326732.57 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:40:33,105 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:40:51,741 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=297283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:08,407 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297294.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:09,555 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2976, 5.8827, 6.0164, 5.6118, 5.8689, 6.3473, 5.8566, 5.5258], device='cuda:5'), covar=tensor([0.0883, 0.2051, 0.2353, 0.2233, 0.2410, 0.0967, 0.1632, 0.2387], device='cuda:5'), in_proj_covar=tensor([0.0440, 0.0657, 0.0732, 0.0532, 0.0712, 0.0750, 0.0563, 0.0710], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:41:22,344 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.045e+02 2.480e+02 2.886e+02 5.342e+02, threshold=4.960e+02, percent-clipped=3.0 2023-05-02 21:41:22,359 INFO [train.py:904] (5/8) Epoch 30, batch 2950, loss[loss=0.1839, simple_loss=0.2756, pruned_loss=0.04607, over 16602.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2525, pruned_loss=0.03753, over 3330016.79 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:41:41,547 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:18,999 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:33,299 INFO [train.py:904] (5/8) Epoch 30, batch 3000, loss[loss=0.1453, simple_loss=0.2314, pruned_loss=0.02962, over 16866.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2525, pruned_loss=0.03833, over 3328407.19 frames. ], batch size: 42, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:42:33,299 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 21:42:42,093 INFO [train.py:938] (5/8) Epoch 30, validation: loss=0.1331, simple_loss=0.238, pruned_loss=0.01415, over 944034.00 frames. 2023-05-02 21:42:42,094 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 21:43:35,256 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297392.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:43:45,093 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6970, 2.6096, 2.5144, 4.1196, 3.3985, 4.0605, 1.5819, 2.8800], device='cuda:5'), covar=tensor([0.1655, 0.0821, 0.1383, 0.0203, 0.0204, 0.0379, 0.1947, 0.0928], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0185, 0.0204, 0.0210, 0.0209, 0.0222, 0.0214, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:43:51,743 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.205e+02 2.625e+02 3.080e+02 5.870e+02, threshold=5.251e+02, percent-clipped=1.0 2023-05-02 21:43:51,759 INFO [train.py:904] (5/8) Epoch 30, batch 3050, loss[loss=0.1667, simple_loss=0.2643, pruned_loss=0.03458, over 17016.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2522, pruned_loss=0.03837, over 3333004.78 frames. ], batch size: 55, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:45:01,434 INFO [train.py:904] (5/8) Epoch 30, batch 3100, loss[loss=0.1649, simple_loss=0.2535, pruned_loss=0.0382, over 16447.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2519, pruned_loss=0.03854, over 3316862.55 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:45:55,838 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0788, 5.5727, 5.6940, 5.3513, 5.4324, 6.0867, 5.5479, 5.2109], device='cuda:5'), covar=tensor([0.1084, 0.2411, 0.2573, 0.2280, 0.2936, 0.1165, 0.1608, 0.2609], device='cuda:5'), in_proj_covar=tensor([0.0442, 0.0659, 0.0734, 0.0533, 0.0714, 0.0751, 0.0563, 0.0712], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:46:07,377 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.135e+02 2.482e+02 3.007e+02 9.801e+02, threshold=4.963e+02, percent-clipped=2.0 2023-05-02 21:46:07,400 INFO [train.py:904] (5/8) Epoch 30, batch 3150, loss[loss=0.1791, simple_loss=0.253, pruned_loss=0.05262, over 16659.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2509, pruned_loss=0.0383, over 3323222.75 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:46:42,709 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297529.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:47:17,211 INFO [train.py:904] (5/8) Epoch 30, batch 3200, loss[loss=0.1588, simple_loss=0.2536, pruned_loss=0.03197, over 17011.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2496, pruned_loss=0.03797, over 3327588.78 frames. ], batch size: 55, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:47:49,872 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:47:51,727 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:48:00,691 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1378, 2.2346, 2.3389, 3.8466, 2.1345, 2.4505, 2.3319, 2.3538], device='cuda:5'), covar=tensor([0.1893, 0.4178, 0.3436, 0.0877, 0.5145, 0.3146, 0.4065, 0.4014], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0486, 0.0396, 0.0347, 0.0453, 0.0558, 0.0458, 0.0569], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:48:13,669 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297594.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:48:27,241 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.037e+02 2.359e+02 2.785e+02 6.437e+02, threshold=4.718e+02, percent-clipped=1.0 2023-05-02 21:48:27,257 INFO [train.py:904] (5/8) Epoch 30, batch 3250, loss[loss=0.1404, simple_loss=0.2312, pruned_loss=0.02486, over 16839.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2503, pruned_loss=0.03793, over 3326372.30 frames. ], batch size: 42, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:49:20,240 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297642.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:49:36,048 INFO [train.py:904] (5/8) Epoch 30, batch 3300, loss[loss=0.2002, simple_loss=0.2834, pruned_loss=0.05852, over 12253.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2522, pruned_loss=0.03868, over 3320531.90 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:50:46,302 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.242e+02 2.585e+02 3.281e+02 4.783e+02, threshold=5.170e+02, percent-clipped=1.0 2023-05-02 21:50:46,330 INFO [train.py:904] (5/8) Epoch 30, batch 3350, loss[loss=0.1905, simple_loss=0.2747, pruned_loss=0.05315, over 16750.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2537, pruned_loss=0.03875, over 3322872.79 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:50:52,775 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 21:50:58,117 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4776, 2.4660, 2.4137, 4.2160, 2.3914, 2.8221, 2.5237, 2.6019], device='cuda:5'), covar=tensor([0.1531, 0.3819, 0.3516, 0.0672, 0.4537, 0.2976, 0.3805, 0.3752], device='cuda:5'), in_proj_covar=tensor([0.0434, 0.0486, 0.0396, 0.0347, 0.0454, 0.0558, 0.0459, 0.0570], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:51:56,065 INFO [train.py:904] (5/8) Epoch 30, batch 3400, loss[loss=0.1642, simple_loss=0.2565, pruned_loss=0.03592, over 16822.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2521, pruned_loss=0.03779, over 3332376.00 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:52:57,907 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0401, 3.6066, 4.1938, 2.1974, 4.3501, 4.4737, 3.3339, 3.6288], device='cuda:5'), covar=tensor([0.0722, 0.0343, 0.0299, 0.1147, 0.0117, 0.0218, 0.0483, 0.0358], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0139, 0.0089, 0.0135, 0.0132, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:53:07,132 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 1.981e+02 2.351e+02 2.739e+02 4.469e+02, threshold=4.702e+02, percent-clipped=0.0 2023-05-02 21:53:07,148 INFO [train.py:904] (5/8) Epoch 30, batch 3450, loss[loss=0.1563, simple_loss=0.2325, pruned_loss=0.04007, over 16430.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2512, pruned_loss=0.03751, over 3323318.45 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:53:23,366 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1794, 2.2823, 2.4479, 3.8182, 2.3156, 2.6192, 2.4046, 2.4412], device='cuda:5'), covar=tensor([0.1654, 0.3835, 0.3179, 0.0751, 0.4036, 0.2709, 0.3866, 0.3337], device='cuda:5'), in_proj_covar=tensor([0.0434, 0.0487, 0.0396, 0.0348, 0.0454, 0.0559, 0.0459, 0.0571], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 21:54:09,058 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8562, 2.7491, 2.5964, 4.6485, 3.5410, 4.2226, 1.7134, 3.1395], device='cuda:5'), covar=tensor([0.1385, 0.0851, 0.1287, 0.0258, 0.0244, 0.0443, 0.1669, 0.0841], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0185, 0.0204, 0.0211, 0.0210, 0.0222, 0.0214, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:54:17,206 INFO [train.py:904] (5/8) Epoch 30, batch 3500, loss[loss=0.1335, simple_loss=0.2203, pruned_loss=0.02339, over 16835.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2494, pruned_loss=0.03676, over 3318029.09 frames. ], batch size: 42, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:54:27,827 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-02 21:54:36,725 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8954, 4.1485, 2.7481, 4.7670, 3.4222, 4.6957, 2.8525, 3.4840], device='cuda:5'), covar=tensor([0.0350, 0.0355, 0.1490, 0.0264, 0.0762, 0.0457, 0.1517, 0.0718], device='cuda:5'), in_proj_covar=tensor([0.0181, 0.0186, 0.0199, 0.0180, 0.0183, 0.0226, 0.0208, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:54:52,020 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:55:09,873 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8596, 2.1518, 2.5373, 2.8703, 2.8046, 3.3097, 2.3687, 3.3196], device='cuda:5'), covar=tensor([0.0310, 0.0606, 0.0461, 0.0386, 0.0408, 0.0249, 0.0595, 0.0177], device='cuda:5'), in_proj_covar=tensor([0.0206, 0.0204, 0.0194, 0.0199, 0.0218, 0.0174, 0.0210, 0.0174], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-05-02 21:55:28,095 INFO [train.py:904] (5/8) Epoch 30, batch 3550, loss[loss=0.1417, simple_loss=0.2383, pruned_loss=0.02253, over 16989.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2482, pruned_loss=0.03628, over 3326969.65 frames. ], batch size: 50, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:55:29,308 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 1.998e+02 2.319e+02 2.785e+02 6.034e+02, threshold=4.637e+02, percent-clipped=1.0 2023-05-02 21:55:59,068 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:56:38,336 INFO [train.py:904] (5/8) Epoch 30, batch 3600, loss[loss=0.1342, simple_loss=0.22, pruned_loss=0.02422, over 16759.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.2478, pruned_loss=0.03593, over 3331977.92 frames. ], batch size: 39, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:56:42,415 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0455, 5.0745, 5.4579, 5.4522, 5.4877, 5.1418, 5.0669, 4.9123], device='cuda:5'), covar=tensor([0.0375, 0.0602, 0.0429, 0.0423, 0.0437, 0.0412, 0.0980, 0.0459], device='cuda:5'), in_proj_covar=tensor([0.0458, 0.0523, 0.0498, 0.0459, 0.0544, 0.0525, 0.0608, 0.0423], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 21:56:42,478 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:56:54,715 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2600, 2.0916, 1.8212, 1.9267, 2.3652, 2.0721, 2.1737, 2.3976], device='cuda:5'), covar=tensor([0.0321, 0.0458, 0.0573, 0.0462, 0.0301, 0.0396, 0.0216, 0.0335], device='cuda:5'), in_proj_covar=tensor([0.0242, 0.0252, 0.0240, 0.0242, 0.0252, 0.0250, 0.0251, 0.0253], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 21:57:24,194 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7056, 2.5303, 2.6552, 4.1799, 3.5505, 4.0514, 1.4841, 3.0910], device='cuda:5'), covar=tensor([0.1487, 0.0779, 0.1163, 0.0176, 0.0161, 0.0334, 0.1710, 0.0817], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0210, 0.0209, 0.0221, 0.0213, 0.0201], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:57:50,526 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8914, 2.5427, 2.5213, 4.1265, 3.2919, 4.1114, 1.6308, 3.0188], device='cuda:5'), covar=tensor([0.1452, 0.0896, 0.1320, 0.0220, 0.0190, 0.0406, 0.1843, 0.0879], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0210, 0.0209, 0.0220, 0.0213, 0.0201], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 21:57:53,335 INFO [train.py:904] (5/8) Epoch 30, batch 3650, loss[loss=0.1873, simple_loss=0.262, pruned_loss=0.05627, over 16916.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2475, pruned_loss=0.03673, over 3330669.95 frames. ], batch size: 116, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:57:55,115 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.298e+02 2.680e+02 3.365e+02 6.432e+02, threshold=5.360e+02, percent-clipped=4.0 2023-05-02 21:58:14,212 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:59:06,237 INFO [train.py:904] (5/8) Epoch 30, batch 3700, loss[loss=0.1574, simple_loss=0.2391, pruned_loss=0.0379, over 16161.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2464, pruned_loss=0.03845, over 3303776.86 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:00:17,706 INFO [train.py:904] (5/8) Epoch 30, batch 3750, loss[loss=0.1569, simple_loss=0.236, pruned_loss=0.03887, over 16848.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2471, pruned_loss=0.03995, over 3282548.28 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:00:19,158 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8538, 4.0361, 3.0455, 2.4149, 2.5791, 2.6363, 4.1907, 3.4961], device='cuda:5'), covar=tensor([0.2742, 0.0540, 0.1916, 0.3271, 0.3120, 0.2197, 0.0505, 0.1511], device='cuda:5'), in_proj_covar=tensor([0.0339, 0.0278, 0.0316, 0.0331, 0.0311, 0.0283, 0.0308, 0.0358], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 22:00:19,705 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.048e+02 2.384e+02 2.845e+02 4.773e+02, threshold=4.768e+02, percent-clipped=0.0 2023-05-02 22:01:30,456 INFO [train.py:904] (5/8) Epoch 30, batch 3800, loss[loss=0.1702, simple_loss=0.2429, pruned_loss=0.04873, over 16727.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2481, pruned_loss=0.04104, over 3260525.26 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:02:22,057 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5653, 4.2845, 4.2952, 2.9048, 3.7394, 4.3130, 3.7742, 2.3667], device='cuda:5'), covar=tensor([0.0582, 0.0105, 0.0052, 0.0413, 0.0119, 0.0106, 0.0117, 0.0511], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0138, 0.0106, 0.0119, 0.0101, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 22:02:43,876 INFO [train.py:904] (5/8) Epoch 30, batch 3850, loss[loss=0.1523, simple_loss=0.2322, pruned_loss=0.03621, over 16864.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2476, pruned_loss=0.04146, over 3268547.83 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:02:44,953 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.341e+02 2.538e+02 2.759e+02 9.437e+02, threshold=5.075e+02, percent-clipped=1.0 2023-05-02 22:03:16,677 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-05-02 22:03:53,280 INFO [train.py:904] (5/8) Epoch 30, batch 3900, loss[loss=0.1512, simple_loss=0.2353, pruned_loss=0.03354, over 16768.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2475, pruned_loss=0.04194, over 3277803.41 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:03:54,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9058, 4.0602, 3.0686, 2.5109, 2.6780, 2.7485, 4.2915, 3.5547], device='cuda:5'), covar=tensor([0.2760, 0.0526, 0.1883, 0.2842, 0.2597, 0.1947, 0.0439, 0.1342], device='cuda:5'), in_proj_covar=tensor([0.0340, 0.0279, 0.0317, 0.0333, 0.0312, 0.0283, 0.0309, 0.0359], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 22:04:59,538 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0774, 4.0303, 3.9831, 3.2477, 4.0053, 1.7404, 3.7873, 3.3852], device='cuda:5'), covar=tensor([0.0164, 0.0138, 0.0226, 0.0312, 0.0109, 0.3150, 0.0167, 0.0387], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0180, 0.0219, 0.0190, 0.0197, 0.0223, 0.0209, 0.0186], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:05:04,310 INFO [train.py:904] (5/8) Epoch 30, batch 3950, loss[loss=0.1613, simple_loss=0.2392, pruned_loss=0.04164, over 16716.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2481, pruned_loss=0.04299, over 3271875.64 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:05:05,533 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.189e+02 2.590e+02 3.423e+02 7.666e+02, threshold=5.180e+02, percent-clipped=4.0 2023-05-02 22:05:16,669 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 22:05:17,530 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:06:02,149 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.0384, 5.1272, 5.3796, 5.3516, 5.4250, 5.1012, 5.0493, 4.8033], device='cuda:5'), covar=tensor([0.0328, 0.0560, 0.0394, 0.0428, 0.0412, 0.0367, 0.0836, 0.0514], device='cuda:5'), in_proj_covar=tensor([0.0456, 0.0520, 0.0495, 0.0458, 0.0541, 0.0523, 0.0603, 0.0422], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-02 22:06:15,413 INFO [train.py:904] (5/8) Epoch 30, batch 4000, loss[loss=0.1743, simple_loss=0.2594, pruned_loss=0.0446, over 15687.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2482, pruned_loss=0.04319, over 3279954.71 frames. ], batch size: 191, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:06:28,638 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 22:07:12,900 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8333, 2.0078, 2.5299, 2.7596, 2.7998, 3.1676, 2.2217, 3.1803], device='cuda:5'), covar=tensor([0.0281, 0.0600, 0.0385, 0.0395, 0.0373, 0.0240, 0.0604, 0.0177], device='cuda:5'), in_proj_covar=tensor([0.0204, 0.0202, 0.0192, 0.0198, 0.0216, 0.0173, 0.0208, 0.0173], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:5') 2023-05-02 22:07:25,658 INFO [train.py:904] (5/8) Epoch 30, batch 4050, loss[loss=0.1846, simple_loss=0.2626, pruned_loss=0.0533, over 12028.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.249, pruned_loss=0.04268, over 3273506.12 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:07:27,599 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 1.991e+02 2.231e+02 2.617e+02 4.473e+02, threshold=4.462e+02, percent-clipped=0.0 2023-05-02 22:07:59,081 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.78 vs. limit=5.0 2023-05-02 22:08:37,718 INFO [train.py:904] (5/8) Epoch 30, batch 4100, loss[loss=0.1858, simple_loss=0.2838, pruned_loss=0.04388, over 16203.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2502, pruned_loss=0.0421, over 3260809.98 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:08:38,395 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.9310, 2.1648, 2.2501, 2.8527, 2.1027, 3.0796, 1.7361, 2.6161], device='cuda:5'), covar=tensor([0.1324, 0.0835, 0.1283, 0.0180, 0.0135, 0.0315, 0.1801, 0.0865], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0209, 0.0208, 0.0220, 0.0212, 0.0201], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 22:09:20,298 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:09:45,820 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4302, 4.6646, 4.5236, 4.5380, 4.2418, 4.1698, 4.2310, 4.7438], device='cuda:5'), covar=tensor([0.1198, 0.0948, 0.0984, 0.0790, 0.0806, 0.1791, 0.1065, 0.0899], device='cuda:5'), in_proj_covar=tensor([0.0745, 0.0898, 0.0736, 0.0698, 0.0575, 0.0568, 0.0755, 0.0706], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:09:53,966 INFO [train.py:904] (5/8) Epoch 30, batch 4150, loss[loss=0.1765, simple_loss=0.2764, pruned_loss=0.03836, over 16704.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2572, pruned_loss=0.0445, over 3216730.64 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:09:56,056 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.028e+02 2.326e+02 2.808e+02 4.631e+02, threshold=4.653e+02, percent-clipped=1.0 2023-05-02 22:10:35,466 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:10:55,719 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298543.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:11:12,254 INFO [train.py:904] (5/8) Epoch 30, batch 4200, loss[loss=0.1849, simple_loss=0.2835, pruned_loss=0.04315, over 16743.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2641, pruned_loss=0.04583, over 3204773.29 frames. ], batch size: 89, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:08,248 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:12:13,013 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8416, 3.9971, 2.5789, 4.6971, 3.1753, 4.6007, 2.8279, 3.2355], device='cuda:5'), covar=tensor([0.0321, 0.0384, 0.1694, 0.0326, 0.0826, 0.0593, 0.1461, 0.0848], device='cuda:5'), in_proj_covar=tensor([0.0179, 0.0184, 0.0198, 0.0177, 0.0182, 0.0224, 0.0206, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 22:12:26,993 INFO [train.py:904] (5/8) Epoch 30, batch 4250, loss[loss=0.1602, simple_loss=0.2638, pruned_loss=0.02827, over 16833.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2674, pruned_loss=0.04528, over 3195783.86 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:28,280 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.166e+02 2.498e+02 2.846e+02 5.761e+02, threshold=4.997e+02, percent-clipped=3.0 2023-05-02 22:12:40,810 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=298613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:12:49,874 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 22:13:04,652 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298629.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:13:24,366 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5564, 3.7191, 2.7608, 2.2733, 2.4501, 2.5118, 3.8932, 3.2213], device='cuda:5'), covar=tensor([0.2995, 0.0624, 0.1900, 0.2753, 0.2676, 0.2145, 0.0453, 0.1335], device='cuda:5'), in_proj_covar=tensor([0.0338, 0.0278, 0.0316, 0.0331, 0.0311, 0.0282, 0.0307, 0.0357], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 22:13:41,921 INFO [train.py:904] (5/8) Epoch 30, batch 4300, loss[loss=0.1862, simple_loss=0.2812, pruned_loss=0.04561, over 16542.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2684, pruned_loss=0.04449, over 3198406.20 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:13:52,509 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=298661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:13:57,500 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9432, 4.7933, 4.9856, 5.1714, 5.2892, 4.7486, 5.3187, 5.3216], device='cuda:5'), covar=tensor([0.1826, 0.1320, 0.1476, 0.0613, 0.0490, 0.0998, 0.0546, 0.0598], device='cuda:5'), in_proj_covar=tensor([0.0705, 0.0855, 0.0987, 0.0872, 0.0663, 0.0691, 0.0725, 0.0844], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:14:14,693 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-05-02 22:14:35,088 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:36,390 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298690.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:14:56,163 INFO [train.py:904] (5/8) Epoch 30, batch 4350, loss[loss=0.2123, simple_loss=0.2972, pruned_loss=0.06366, over 16919.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2713, pruned_loss=0.04516, over 3212833.88 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:14:57,383 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.186e+02 2.538e+02 2.866e+02 4.288e+02, threshold=5.076e+02, percent-clipped=0.0 2023-05-02 22:15:10,174 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.2263, 3.7989, 3.6766, 2.3972, 3.3607, 3.7650, 3.3832, 2.0303], device='cuda:5'), covar=tensor([0.0606, 0.0053, 0.0077, 0.0461, 0.0127, 0.0100, 0.0121, 0.0519], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0137, 0.0106, 0.0118, 0.0101, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 22:16:04,774 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298750.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:16:09,622 INFO [train.py:904] (5/8) Epoch 30, batch 4400, loss[loss=0.1875, simple_loss=0.2815, pruned_loss=0.04671, over 16380.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2732, pruned_loss=0.04619, over 3210509.96 frames. ], batch size: 35, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:17:21,883 INFO [train.py:904] (5/8) Epoch 30, batch 4450, loss[loss=0.22, simple_loss=0.2948, pruned_loss=0.07263, over 11728.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2768, pruned_loss=0.04798, over 3202822.46 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:17:23,565 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.049e+02 2.382e+02 3.028e+02 5.771e+02, threshold=4.764e+02, percent-clipped=2.0 2023-05-02 22:18:10,730 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298838.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:18:32,516 INFO [train.py:904] (5/8) Epoch 30, batch 4500, loss[loss=0.2049, simple_loss=0.2798, pruned_loss=0.06499, over 11779.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2775, pruned_loss=0.04904, over 3202639.13 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:18:34,435 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.8149, 2.8698, 2.4217, 2.6352, 3.2085, 2.8023, 3.2771, 3.3727], device='cuda:5'), covar=tensor([0.0098, 0.0394, 0.0544, 0.0459, 0.0266, 0.0396, 0.0252, 0.0268], device='cuda:5'), in_proj_covar=tensor([0.0237, 0.0247, 0.0236, 0.0238, 0.0247, 0.0245, 0.0246, 0.0248], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:18:48,173 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 22:19:06,681 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 22:19:19,856 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:19:44,896 INFO [train.py:904] (5/8) Epoch 30, batch 4550, loss[loss=0.2002, simple_loss=0.287, pruned_loss=0.05665, over 17210.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2785, pruned_loss=0.05017, over 3218156.26 frames. ], batch size: 45, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:19:45,380 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4300, 4.2088, 4.0981, 2.7935, 3.7443, 4.1751, 3.6672, 2.5564], device='cuda:5'), covar=tensor([0.0605, 0.0039, 0.0053, 0.0423, 0.0101, 0.0091, 0.0109, 0.0437], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0091, 0.0093, 0.0136, 0.0105, 0.0117, 0.0100, 0.0131], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 22:19:46,102 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 1.782e+02 2.097e+02 2.367e+02 4.731e+02, threshold=4.194e+02, percent-clipped=0.0 2023-05-02 22:20:57,296 INFO [train.py:904] (5/8) Epoch 30, batch 4600, loss[loss=0.191, simple_loss=0.2774, pruned_loss=0.05227, over 16839.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2797, pruned_loss=0.05044, over 3221850.81 frames. ], batch size: 39, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:21:14,574 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.1092, 5.1232, 4.8658, 4.1674, 5.0546, 2.1309, 4.7663, 4.3939], device='cuda:5'), covar=tensor([0.0066, 0.0059, 0.0150, 0.0336, 0.0058, 0.2676, 0.0092, 0.0238], device='cuda:5'), in_proj_covar=tensor([0.0188, 0.0179, 0.0218, 0.0190, 0.0196, 0.0222, 0.0207, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:21:43,600 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298985.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:21:50,680 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 22:21:59,724 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 22:22:09,593 INFO [train.py:904] (5/8) Epoch 30, batch 4650, loss[loss=0.1881, simple_loss=0.2799, pruned_loss=0.0482, over 16705.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2793, pruned_loss=0.05052, over 3226586.93 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:22:10,066 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8685, 4.7094, 4.9246, 5.0601, 5.1933, 4.7004, 5.2091, 5.2308], device='cuda:5'), covar=tensor([0.1535, 0.1135, 0.1312, 0.0612, 0.0476, 0.0893, 0.0558, 0.0505], device='cuda:5'), in_proj_covar=tensor([0.0696, 0.0843, 0.0977, 0.0862, 0.0654, 0.0681, 0.0715, 0.0832], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:22:10,879 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 1.831e+02 2.100e+02 2.427e+02 4.549e+02, threshold=4.199e+02, percent-clipped=1.0 2023-05-02 22:23:09,809 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299045.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:23:22,504 INFO [train.py:904] (5/8) Epoch 30, batch 4700, loss[loss=0.1932, simple_loss=0.2752, pruned_loss=0.05564, over 16834.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2767, pruned_loss=0.04962, over 3201491.45 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:23:47,767 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0223, 4.2461, 4.1107, 4.1409, 3.8061, 3.8181, 3.8654, 4.2469], device='cuda:5'), covar=tensor([0.1057, 0.0812, 0.0842, 0.0698, 0.0767, 0.1830, 0.0890, 0.0906], device='cuda:5'), in_proj_covar=tensor([0.0728, 0.0877, 0.0721, 0.0683, 0.0563, 0.0558, 0.0736, 0.0691], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:24:36,512 INFO [train.py:904] (5/8) Epoch 30, batch 4750, loss[loss=0.21, simple_loss=0.2924, pruned_loss=0.0638, over 12065.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2729, pruned_loss=0.04754, over 3194180.29 frames. ], batch size: 248, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:24:37,719 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 1.904e+02 2.113e+02 2.472e+02 4.285e+02, threshold=4.226e+02, percent-clipped=1.0 2023-05-02 22:25:08,338 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:24,794 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6319, 2.4731, 2.3450, 3.7531, 2.3251, 3.6795, 1.5309, 2.7435], device='cuda:5'), covar=tensor([0.1460, 0.0889, 0.1362, 0.0198, 0.0148, 0.0404, 0.1760, 0.0897], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0208, 0.0208, 0.0219, 0.0212, 0.0201], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 22:25:25,865 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:49,339 INFO [train.py:904] (5/8) Epoch 30, batch 4800, loss[loss=0.1689, simple_loss=0.261, pruned_loss=0.03838, over 16478.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2701, pruned_loss=0.0459, over 3186733.69 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:26:37,336 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:37,393 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:38,556 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299187.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:46,079 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299191.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:27:04,373 INFO [train.py:904] (5/8) Epoch 30, batch 4850, loss[loss=0.1818, simple_loss=0.2794, pruned_loss=0.04208, over 16322.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2699, pruned_loss=0.04428, over 3198596.02 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:27:06,459 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.292e+02 1.894e+02 2.171e+02 2.574e+02 8.129e+02, threshold=4.343e+02, percent-clipped=1.0 2023-05-02 22:27:49,434 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:02,958 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299243.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:16,586 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299252.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:19,003 INFO [train.py:904] (5/8) Epoch 30, batch 4900, loss[loss=0.168, simple_loss=0.2659, pruned_loss=0.03507, over 15528.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2687, pruned_loss=0.04309, over 3190329.47 frames. ], batch size: 191, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:28:23,299 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 22:28:24,269 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5684, 3.5104, 3.4773, 2.6317, 3.3250, 2.0591, 3.1029, 2.7771], device='cuda:5'), covar=tensor([0.0174, 0.0183, 0.0197, 0.0324, 0.0141, 0.2686, 0.0162, 0.0336], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0178, 0.0217, 0.0188, 0.0195, 0.0221, 0.0206, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:28:49,732 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6909, 4.7700, 4.5849, 4.2310, 4.2400, 4.6735, 4.5136, 4.3973], device='cuda:5'), covar=tensor([0.0683, 0.0500, 0.0354, 0.0320, 0.1023, 0.0574, 0.0424, 0.0630], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0481, 0.0374, 0.0375, 0.0369, 0.0431, 0.0255, 0.0442], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:29:06,027 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299285.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:29:33,777 INFO [train.py:904] (5/8) Epoch 30, batch 4950, loss[loss=0.1689, simple_loss=0.2724, pruned_loss=0.03272, over 16852.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2685, pruned_loss=0.0428, over 3194521.30 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:29:34,223 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299304.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:29:34,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 1.940e+02 2.221e+02 2.664e+02 5.987e+02, threshold=4.443e+02, percent-clipped=2.0 2023-05-02 22:30:02,835 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:30:17,911 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299333.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:30:31,357 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1144, 2.4782, 2.4282, 3.9051, 2.2594, 2.7547, 2.4813, 2.5533], device='cuda:5'), covar=tensor([0.1641, 0.3380, 0.3065, 0.0598, 0.3928, 0.2567, 0.3527, 0.3191], device='cuda:5'), in_proj_covar=tensor([0.0429, 0.0483, 0.0391, 0.0344, 0.0449, 0.0554, 0.0455, 0.0565], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:30:36,206 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:30:47,909 INFO [train.py:904] (5/8) Epoch 30, batch 5000, loss[loss=0.1732, simple_loss=0.2767, pruned_loss=0.03488, over 16731.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2697, pruned_loss=0.04254, over 3202155.52 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:30:56,536 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8538, 2.7805, 2.6663, 1.9540, 2.5926, 2.7465, 2.6310, 1.8309], device='cuda:5'), covar=tensor([0.0534, 0.0101, 0.0110, 0.0416, 0.0156, 0.0155, 0.0160, 0.0486], device='cuda:5'), in_proj_covar=tensor([0.0140, 0.0091, 0.0093, 0.0136, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 22:31:26,701 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 22:31:33,619 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:31:46,454 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:32:01,707 INFO [train.py:904] (5/8) Epoch 30, batch 5050, loss[loss=0.1698, simple_loss=0.2591, pruned_loss=0.04026, over 16518.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2705, pruned_loss=0.0426, over 3211058.13 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:32:02,881 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.023e+02 2.400e+02 2.818e+02 3.992e+02, threshold=4.801e+02, percent-clipped=0.0 2023-05-02 22:32:38,163 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3091, 5.2836, 5.1750, 4.3043, 5.2172, 1.7745, 4.8878, 4.6413], device='cuda:5'), covar=tensor([0.0108, 0.0109, 0.0179, 0.0590, 0.0126, 0.3161, 0.0134, 0.0330], device='cuda:5'), in_proj_covar=tensor([0.0187, 0.0178, 0.0217, 0.0189, 0.0195, 0.0221, 0.0206, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:33:13,775 INFO [train.py:904] (5/8) Epoch 30, batch 5100, loss[loss=0.1585, simple_loss=0.2498, pruned_loss=0.0336, over 16663.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2687, pruned_loss=0.042, over 3215000.66 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:33:19,891 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.4943, 3.4481, 2.6857, 2.2036, 2.2557, 2.3577, 3.5748, 3.0279], device='cuda:5'), covar=tensor([0.3054, 0.0658, 0.1937, 0.2831, 0.2738, 0.2288, 0.0511, 0.1499], device='cuda:5'), in_proj_covar=tensor([0.0339, 0.0277, 0.0316, 0.0330, 0.0309, 0.0282, 0.0307, 0.0356], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 22:33:47,048 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4424, 3.6597, 3.3923, 3.1278, 3.0422, 3.5289, 3.2545, 3.3509], device='cuda:5'), covar=tensor([0.0839, 0.0643, 0.0459, 0.0416, 0.0939, 0.0544, 0.1812, 0.0577], device='cuda:5'), in_proj_covar=tensor([0.0318, 0.0481, 0.0374, 0.0376, 0.0370, 0.0433, 0.0255, 0.0443], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:33:57,986 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:34:00,985 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.7301, 1.7447, 1.4925, 1.3659, 1.7856, 1.5104, 1.6389, 1.9450], device='cuda:5'), covar=tensor([0.0263, 0.0389, 0.0563, 0.0480, 0.0263, 0.0384, 0.0223, 0.0319], device='cuda:5'), in_proj_covar=tensor([0.0235, 0.0246, 0.0236, 0.0236, 0.0247, 0.0244, 0.0244, 0.0247], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:34:30,238 INFO [train.py:904] (5/8) Epoch 30, batch 5150, loss[loss=0.1652, simple_loss=0.2633, pruned_loss=0.0335, over 16821.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2688, pruned_loss=0.04154, over 3203842.83 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:34:31,342 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 1.896e+02 2.234e+02 2.667e+02 3.487e+02, threshold=4.469e+02, percent-clipped=0.0 2023-05-02 22:35:14,597 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6176, 3.3675, 3.8414, 1.8119, 4.0007, 4.0268, 3.0531, 3.0503], device='cuda:5'), covar=tensor([0.0828, 0.0304, 0.0191, 0.1335, 0.0085, 0.0160, 0.0444, 0.0493], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0139, 0.0088, 0.0133, 0.0132, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 22:35:33,166 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:35:43,069 INFO [train.py:904] (5/8) Epoch 30, batch 5200, loss[loss=0.1771, simple_loss=0.2644, pruned_loss=0.04493, over 16778.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2669, pruned_loss=0.04085, over 3204560.25 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:35:51,508 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.3087, 5.5852, 5.2974, 5.3798, 5.1302, 5.0412, 4.9178, 5.6886], device='cuda:5'), covar=tensor([0.1204, 0.0765, 0.1043, 0.0805, 0.0740, 0.0767, 0.1165, 0.0841], device='cuda:5'), in_proj_covar=tensor([0.0724, 0.0876, 0.0718, 0.0680, 0.0560, 0.0555, 0.0734, 0.0689], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:36:28,579 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-02 22:36:48,436 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299599.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:36:55,734 INFO [train.py:904] (5/8) Epoch 30, batch 5250, loss[loss=0.1855, simple_loss=0.272, pruned_loss=0.0495, over 16911.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2647, pruned_loss=0.04042, over 3219979.05 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:36:56,957 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.002e+02 2.220e+02 2.747e+02 4.551e+02, threshold=4.439e+02, percent-clipped=1.0 2023-05-02 22:37:55,021 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5811, 3.5180, 3.5080, 2.7017, 3.3252, 2.0979, 3.1723, 2.7974], device='cuda:5'), covar=tensor([0.0164, 0.0187, 0.0186, 0.0286, 0.0121, 0.2583, 0.0172, 0.0286], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0177, 0.0216, 0.0188, 0.0194, 0.0220, 0.0205, 0.0183], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:38:06,896 INFO [train.py:904] (5/8) Epoch 30, batch 5300, loss[loss=0.1469, simple_loss=0.2386, pruned_loss=0.02762, over 16443.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2607, pruned_loss=0.03905, over 3231074.19 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:38:36,223 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 22:38:42,897 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:39:18,188 INFO [train.py:904] (5/8) Epoch 30, batch 5350, loss[loss=0.1713, simple_loss=0.2589, pruned_loss=0.04183, over 12374.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2597, pruned_loss=0.03853, over 3233543.86 frames. ], batch size: 248, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:39:19,352 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.376e+02 1.953e+02 2.222e+02 2.592e+02 5.137e+02, threshold=4.444e+02, percent-clipped=1.0 2023-05-02 22:40:29,717 INFO [train.py:904] (5/8) Epoch 30, batch 5400, loss[loss=0.1719, simple_loss=0.271, pruned_loss=0.03639, over 16707.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2614, pruned_loss=0.0391, over 3222122.67 frames. ], batch size: 89, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:40:38,773 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-02 22:40:51,533 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299769.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:06,402 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 22:41:10,049 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:34,633 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299798.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:41:46,649 INFO [train.py:904] (5/8) Epoch 30, batch 5450, loss[loss=0.1849, simple_loss=0.2758, pruned_loss=0.04701, over 16760.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2641, pruned_loss=0.04049, over 3207373.29 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:41:47,799 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.110e+02 2.402e+02 2.864e+02 4.092e+02, threshold=4.805e+02, percent-clipped=0.0 2023-05-02 22:42:07,721 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 22:42:27,031 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:27,185 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:35,126 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 22:42:52,028 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299847.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:43:02,914 INFO [train.py:904] (5/8) Epoch 30, batch 5500, loss[loss=0.1925, simple_loss=0.2909, pruned_loss=0.04708, over 16701.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2706, pruned_loss=0.04393, over 3194580.21 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:43:11,738 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299859.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:44:04,080 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.6711, 2.5480, 2.4459, 3.8301, 2.5874, 3.8673, 1.6137, 2.8485], device='cuda:5'), covar=tensor([0.1432, 0.0868, 0.1319, 0.0188, 0.0215, 0.0418, 0.1798, 0.0854], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0184, 0.0203, 0.0208, 0.0209, 0.0219, 0.0213, 0.0202], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 22:44:09,167 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299895.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:14,814 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:21,849 INFO [train.py:904] (5/8) Epoch 30, batch 5550, loss[loss=0.1874, simple_loss=0.2819, pruned_loss=0.04645, over 16757.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2779, pruned_loss=0.04877, over 3163135.39 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:44:23,800 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.226e+02 2.904e+02 3.261e+02 4.114e+02 7.155e+02, threshold=6.523e+02, percent-clipped=13.0 2023-05-02 22:44:40,972 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2019, 4.2490, 4.1067, 3.8270, 3.8270, 4.1831, 3.9067, 3.9882], device='cuda:5'), covar=tensor([0.0642, 0.0625, 0.0345, 0.0324, 0.0799, 0.0519, 0.0943, 0.0632], device='cuda:5'), in_proj_covar=tensor([0.0319, 0.0486, 0.0377, 0.0377, 0.0372, 0.0436, 0.0256, 0.0446], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:44:46,966 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 22:45:32,581 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299947.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:45:43,014 INFO [train.py:904] (5/8) Epoch 30, batch 5600, loss[loss=0.2132, simple_loss=0.3029, pruned_loss=0.06174, over 16717.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.283, pruned_loss=0.0532, over 3108426.94 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:46:18,990 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3571, 4.4618, 4.2752, 4.0030, 3.8455, 4.4034, 4.1822, 4.0636], device='cuda:5'), covar=tensor([0.0799, 0.0682, 0.0430, 0.0446, 0.1115, 0.0596, 0.0776, 0.0815], device='cuda:5'), in_proj_covar=tensor([0.0319, 0.0484, 0.0376, 0.0377, 0.0371, 0.0435, 0.0256, 0.0446], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:46:27,276 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:47:11,378 INFO [train.py:904] (5/8) Epoch 30, batch 5650, loss[loss=0.246, simple_loss=0.3164, pruned_loss=0.08776, over 11470.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2878, pruned_loss=0.05724, over 3073902.07 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:47:13,276 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 3.189e+02 3.868e+02 4.538e+02 6.994e+02, threshold=7.735e+02, percent-clipped=1.0 2023-05-02 22:47:48,226 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:48:23,964 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2270, 2.4510, 2.4855, 3.9556, 2.2245, 2.7711, 2.4530, 2.5706], device='cuda:5'), covar=tensor([0.1459, 0.3355, 0.2869, 0.0558, 0.4101, 0.2335, 0.3419, 0.3152], device='cuda:5'), in_proj_covar=tensor([0.0426, 0.0479, 0.0389, 0.0340, 0.0446, 0.0550, 0.0452, 0.0561], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:48:27,605 INFO [train.py:904] (5/8) Epoch 30, batch 5700, loss[loss=0.2079, simple_loss=0.2961, pruned_loss=0.05989, over 16189.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2895, pruned_loss=0.05884, over 3073721.23 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:48:57,397 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:49:45,356 INFO [train.py:904] (5/8) Epoch 30, batch 5750, loss[loss=0.2095, simple_loss=0.2981, pruned_loss=0.06049, over 16656.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2929, pruned_loss=0.06081, over 3050712.72 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:49:49,227 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.972e+02 3.119e+02 3.894e+02 5.149e+02 1.112e+03, threshold=7.788e+02, percent-clipped=3.0 2023-05-02 22:50:18,761 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300125.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:50:35,577 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300134.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:51:06,741 INFO [train.py:904] (5/8) Epoch 30, batch 5800, loss[loss=0.1847, simple_loss=0.2772, pruned_loss=0.04607, over 15296.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.293, pruned_loss=0.06009, over 3038801.98 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:51:07,778 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300154.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:51:15,454 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6087, 4.7984, 4.9316, 4.7441, 4.7921, 5.3085, 4.7994, 4.5488], device='cuda:5'), covar=tensor([0.1263, 0.1847, 0.2655, 0.1766, 0.2180, 0.0840, 0.1544, 0.2372], device='cuda:5'), in_proj_covar=tensor([0.0432, 0.0640, 0.0715, 0.0519, 0.0692, 0.0730, 0.0550, 0.0691], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 22:51:34,641 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5875, 3.6584, 2.7522, 2.3111, 2.4184, 2.4596, 3.9228, 3.2320], device='cuda:5'), covar=tensor([0.3061, 0.0574, 0.1909, 0.2856, 0.2772, 0.2211, 0.0436, 0.1351], device='cuda:5'), in_proj_covar=tensor([0.0338, 0.0276, 0.0316, 0.0331, 0.0308, 0.0281, 0.0306, 0.0354], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 22:52:25,623 INFO [train.py:904] (5/8) Epoch 30, batch 5850, loss[loss=0.226, simple_loss=0.3092, pruned_loss=0.07143, over 15460.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2908, pruned_loss=0.05846, over 3046853.94 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:52:28,950 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.843e+02 2.638e+02 2.992e+02 3.617e+02 7.830e+02, threshold=5.985e+02, percent-clipped=1.0 2023-05-02 22:53:48,213 INFO [train.py:904] (5/8) Epoch 30, batch 5900, loss[loss=0.1853, simple_loss=0.2884, pruned_loss=0.04107, over 16481.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2907, pruned_loss=0.05803, over 3070510.06 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:54:06,668 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2057, 3.9901, 4.6660, 2.1558, 4.8623, 4.8616, 3.6903, 3.7018], device='cuda:5'), covar=tensor([0.0725, 0.0282, 0.0155, 0.1303, 0.0070, 0.0150, 0.0336, 0.0442], device='cuda:5'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0139, 0.0088, 0.0134, 0.0131, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 22:55:09,237 INFO [train.py:904] (5/8) Epoch 30, batch 5950, loss[loss=0.1872, simple_loss=0.2806, pruned_loss=0.04694, over 17219.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2915, pruned_loss=0.05709, over 3075357.41 frames. ], batch size: 45, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:55:12,898 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.782e+02 3.249e+02 4.004e+02 6.648e+02, threshold=6.499e+02, percent-clipped=2.0 2023-05-02 22:56:29,063 INFO [train.py:904] (5/8) Epoch 30, batch 6000, loss[loss=0.2122, simple_loss=0.2947, pruned_loss=0.06488, over 11637.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2895, pruned_loss=0.05607, over 3083036.70 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:56:29,063 INFO [train.py:929] (5/8) Computing validation loss 2023-05-02 22:56:39,789 INFO [train.py:938] (5/8) Epoch 30, validation: loss=0.1471, simple_loss=0.2591, pruned_loss=0.01755, over 944034.00 frames. 2023-05-02 22:56:39,790 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-02 22:56:55,596 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0331, 4.0583, 4.3280, 4.2903, 4.3197, 4.0712, 4.0788, 4.0739], device='cuda:5'), covar=tensor([0.0364, 0.0664, 0.0443, 0.0450, 0.0488, 0.0477, 0.0852, 0.0522], device='cuda:5'), in_proj_covar=tensor([0.0441, 0.0505, 0.0483, 0.0448, 0.0529, 0.0512, 0.0587, 0.0412], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 22:57:43,366 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-05-02 22:57:57,804 INFO [train.py:904] (5/8) Epoch 30, batch 6050, loss[loss=0.2146, simple_loss=0.2852, pruned_loss=0.07202, over 11958.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2883, pruned_loss=0.0557, over 3086097.92 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:58:01,297 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.541e+02 2.904e+02 3.584e+02 6.768e+02, threshold=5.808e+02, percent-clipped=1.0 2023-05-02 22:58:22,772 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.7787, 4.8659, 4.6764, 4.3351, 4.3400, 4.7669, 4.6242, 4.5051], device='cuda:5'), covar=tensor([0.0764, 0.0761, 0.0352, 0.0387, 0.1034, 0.0518, 0.0506, 0.0764], device='cuda:5'), in_proj_covar=tensor([0.0316, 0.0480, 0.0371, 0.0372, 0.0368, 0.0429, 0.0254, 0.0441], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:5') 2023-05-02 22:58:29,328 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:58:35,450 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300429.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:59:15,659 INFO [train.py:904] (5/8) Epoch 30, batch 6100, loss[loss=0.2238, simple_loss=0.3009, pruned_loss=0.07332, over 11587.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.288, pruned_loss=0.05496, over 3093909.66 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:59:17,353 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300454.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:59:47,469 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:00:16,866 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.46 vs. limit=2.0 2023-05-02 23:00:32,834 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300502.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 23:00:34,170 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.2673, 5.5341, 5.3053, 5.3353, 5.0550, 4.9656, 4.9684, 5.6720], device='cuda:5'), covar=tensor([0.1369, 0.0854, 0.0998, 0.0819, 0.0831, 0.0930, 0.1185, 0.0820], device='cuda:5'), in_proj_covar=tensor([0.0720, 0.0870, 0.0715, 0.0677, 0.0557, 0.0553, 0.0727, 0.0683], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 23:00:34,981 INFO [train.py:904] (5/8) Epoch 30, batch 6150, loss[loss=0.1752, simple_loss=0.2686, pruned_loss=0.04095, over 16745.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2863, pruned_loss=0.05445, over 3109234.56 frames. ], batch size: 76, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:00:38,234 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.879e+02 2.817e+02 3.397e+02 3.962e+02 7.100e+02, threshold=6.795e+02, percent-clipped=3.0 2023-05-02 23:01:05,092 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 23:01:53,695 INFO [train.py:904] (5/8) Epoch 30, batch 6200, loss[loss=0.1989, simple_loss=0.2749, pruned_loss=0.06139, over 11102.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2835, pruned_loss=0.05349, over 3117403.46 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:03:06,741 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2382, 2.3604, 2.3956, 4.0597, 2.2110, 2.6298, 2.3678, 2.4824], device='cuda:5'), covar=tensor([0.1551, 0.3634, 0.3116, 0.0581, 0.4290, 0.2687, 0.3907, 0.3424], device='cuda:5'), in_proj_covar=tensor([0.0427, 0.0479, 0.0390, 0.0341, 0.0447, 0.0551, 0.0453, 0.0563], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 23:03:13,168 INFO [train.py:904] (5/8) Epoch 30, batch 6250, loss[loss=0.1876, simple_loss=0.2754, pruned_loss=0.04985, over 17084.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.283, pruned_loss=0.0528, over 3138505.61 frames. ], batch size: 53, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:03:16,360 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.845e+02 3.157e+02 4.016e+02 6.897e+02, threshold=6.314e+02, percent-clipped=1.0 2023-05-02 23:03:29,148 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0523, 3.1897, 3.5118, 1.9531, 3.0148, 2.4000, 3.5245, 3.5974], device='cuda:5'), covar=tensor([0.0229, 0.0900, 0.0618, 0.2267, 0.0873, 0.1030, 0.0559, 0.0829], device='cuda:5'), in_proj_covar=tensor([0.0161, 0.0172, 0.0172, 0.0157, 0.0149, 0.0133, 0.0146, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 23:03:34,792 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300617.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:04:31,371 INFO [train.py:904] (5/8) Epoch 30, batch 6300, loss[loss=0.1808, simple_loss=0.2815, pruned_loss=0.04009, over 16765.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2826, pruned_loss=0.05208, over 3134144.57 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:04:37,005 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.8888, 2.7025, 2.8421, 2.1753, 2.6804, 2.1819, 2.7474, 2.9032], device='cuda:5'), covar=tensor([0.0282, 0.0872, 0.0547, 0.1807, 0.0856, 0.0918, 0.0580, 0.0766], device='cuda:5'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0134, 0.0146, 0.0185], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:5') 2023-05-02 23:04:37,280 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 23:04:56,313 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8140, 2.3697, 2.2481, 3.1875, 1.9005, 3.4891, 1.5928, 2.6997], device='cuda:5'), covar=tensor([0.1462, 0.0932, 0.1429, 0.0234, 0.0209, 0.0537, 0.1932, 0.0947], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0182, 0.0202, 0.0207, 0.0208, 0.0219, 0.0212, 0.0201], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 23:04:56,670 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-02 23:05:11,442 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300678.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:05:50,450 INFO [train.py:904] (5/8) Epoch 30, batch 6350, loss[loss=0.2365, simple_loss=0.3041, pruned_loss=0.08448, over 11506.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.283, pruned_loss=0.05281, over 3129070.24 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:53,980 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.820e+02 3.153e+02 4.098e+02 7.645e+02, threshold=6.307e+02, percent-clipped=4.0 2023-05-02 23:06:29,229 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:07:07,479 INFO [train.py:904] (5/8) Epoch 30, batch 6400, loss[loss=0.1783, simple_loss=0.2689, pruned_loss=0.04386, over 16775.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2834, pruned_loss=0.0541, over 3126358.64 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:07:42,794 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300777.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:08:19,386 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300802.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:08:21,341 INFO [train.py:904] (5/8) Epoch 30, batch 6450, loss[loss=0.1948, simple_loss=0.2889, pruned_loss=0.05035, over 16901.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2831, pruned_loss=0.0534, over 3111004.54 frames. ], batch size: 116, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:08:24,284 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.836e+02 3.339e+02 4.246e+02 8.060e+02, threshold=6.678e+02, percent-clipped=2.0 2023-05-02 23:09:38,134 INFO [train.py:904] (5/8) Epoch 30, batch 6500, loss[loss=0.1732, simple_loss=0.2696, pruned_loss=0.03839, over 16789.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2811, pruned_loss=0.05248, over 3126524.88 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:09:52,558 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300863.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:10:10,051 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.0168, 3.3721, 3.3424, 2.0821, 3.1411, 3.4075, 3.1748, 1.9490], device='cuda:5'), covar=tensor([0.0686, 0.0090, 0.0101, 0.0546, 0.0143, 0.0153, 0.0138, 0.0551], device='cuda:5'), in_proj_covar=tensor([0.0141, 0.0093, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:5') 2023-05-02 23:10:58,890 INFO [train.py:904] (5/8) Epoch 30, batch 6550, loss[loss=0.181, simple_loss=0.2875, pruned_loss=0.03729, over 16900.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2837, pruned_loss=0.05295, over 3136602.70 frames. ], batch size: 96, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:11:01,654 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.987e+02 2.626e+02 3.208e+02 3.708e+02 1.015e+03, threshold=6.415e+02, percent-clipped=2.0 2023-05-02 23:12:19,427 INFO [train.py:904] (5/8) Epoch 30, batch 6600, loss[loss=0.1842, simple_loss=0.2895, pruned_loss=0.03944, over 16878.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2858, pruned_loss=0.05316, over 3145362.57 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:12:37,546 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 23:12:48,898 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:13:38,992 INFO [train.py:904] (5/8) Epoch 30, batch 6650, loss[loss=0.2056, simple_loss=0.2915, pruned_loss=0.05978, over 15303.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2869, pruned_loss=0.0546, over 3123759.72 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:13:43,935 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.674e+02 3.483e+02 3.945e+02 7.489e+02, threshold=6.965e+02, percent-clipped=2.0 2023-05-02 23:14:06,133 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8159, 4.6273, 4.8152, 4.9867, 5.1612, 4.6638, 5.1618, 5.1577], device='cuda:5'), covar=tensor([0.1943, 0.1318, 0.1684, 0.0767, 0.0628, 0.0964, 0.0683, 0.0723], device='cuda:5'), in_proj_covar=tensor([0.0678, 0.0822, 0.0949, 0.0842, 0.0637, 0.0661, 0.0702, 0.0812], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 23:14:48,444 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.8974, 5.2559, 5.4232, 5.2018, 5.2394, 5.7704, 5.1741, 4.9720], device='cuda:5'), covar=tensor([0.1069, 0.1599, 0.2111, 0.1621, 0.2080, 0.0774, 0.1698, 0.2343], device='cuda:5'), in_proj_covar=tensor([0.0433, 0.0641, 0.0716, 0.0522, 0.0693, 0.0731, 0.0553, 0.0698], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-02 23:14:55,118 INFO [train.py:904] (5/8) Epoch 30, batch 6700, loss[loss=0.2438, simple_loss=0.3072, pruned_loss=0.09018, over 11418.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2858, pruned_loss=0.0554, over 3092895.28 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:16:01,695 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5207, 3.4104, 3.8547, 1.9465, 4.0074, 4.0069, 3.0317, 2.9152], device='cuda:5'), covar=tensor([0.0896, 0.0341, 0.0213, 0.1338, 0.0085, 0.0196, 0.0482, 0.0565], device='cuda:5'), in_proj_covar=tensor([0.0152, 0.0113, 0.0105, 0.0141, 0.0089, 0.0136, 0.0133, 0.0133], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 23:16:11,407 INFO [train.py:904] (5/8) Epoch 30, batch 6750, loss[loss=0.184, simple_loss=0.2718, pruned_loss=0.04809, over 16857.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2848, pruned_loss=0.05551, over 3085561.14 frames. ], batch size: 116, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:16:15,731 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.866e+02 3.351e+02 4.005e+02 5.751e+02, threshold=6.702e+02, percent-clipped=0.0 2023-05-02 23:17:07,691 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:17:29,013 INFO [train.py:904] (5/8) Epoch 30, batch 6800, loss[loss=0.1945, simple_loss=0.2788, pruned_loss=0.05514, over 16667.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2848, pruned_loss=0.0556, over 3089718.22 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:17:35,480 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301158.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:17:41,773 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.3715, 3.4788, 3.5960, 3.5767, 3.5918, 3.4294, 3.4417, 3.4686], device='cuda:5'), covar=tensor([0.0458, 0.0694, 0.0546, 0.0515, 0.0582, 0.0617, 0.0934, 0.0640], device='cuda:5'), in_proj_covar=tensor([0.0445, 0.0510, 0.0486, 0.0450, 0.0533, 0.0516, 0.0592, 0.0415], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 23:18:43,922 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:18:47,474 INFO [train.py:904] (5/8) Epoch 30, batch 6850, loss[loss=0.1715, simple_loss=0.2839, pruned_loss=0.0296, over 16738.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2859, pruned_loss=0.05609, over 3079702.77 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:18:51,708 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.806e+02 3.426e+02 4.107e+02 8.215e+02, threshold=6.851e+02, percent-clipped=6.0 2023-05-02 23:18:54,491 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 23:20:04,232 INFO [train.py:904] (5/8) Epoch 30, batch 6900, loss[loss=0.2063, simple_loss=0.2927, pruned_loss=0.05989, over 15263.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2882, pruned_loss=0.05592, over 3080777.86 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:20:35,495 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:20:47,296 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6888, 4.7339, 5.0254, 4.9944, 5.0488, 4.7321, 4.7002, 4.5822], device='cuda:5'), covar=tensor([0.0383, 0.0602, 0.0407, 0.0432, 0.0535, 0.0411, 0.0958, 0.0512], device='cuda:5'), in_proj_covar=tensor([0.0445, 0.0509, 0.0485, 0.0450, 0.0531, 0.0515, 0.0591, 0.0413], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-02 23:21:22,702 INFO [train.py:904] (5/8) Epoch 30, batch 6950, loss[loss=0.252, simple_loss=0.3172, pruned_loss=0.09337, over 11403.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2896, pruned_loss=0.05714, over 3082473.04 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:21:26,936 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.877e+02 3.278e+02 4.034e+02 6.348e+02, threshold=6.555e+02, percent-clipped=0.0 2023-05-02 23:21:50,651 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:22:38,933 INFO [train.py:904] (5/8) Epoch 30, batch 7000, loss[loss=0.1961, simple_loss=0.2944, pruned_loss=0.04893, over 16945.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2895, pruned_loss=0.0562, over 3085775.76 frames. ], batch size: 109, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:23:28,692 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 23:23:55,579 INFO [train.py:904] (5/8) Epoch 30, batch 7050, loss[loss=0.1943, simple_loss=0.2893, pruned_loss=0.04968, over 16728.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2899, pruned_loss=0.05558, over 3108259.87 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:24:00,753 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.564e+02 3.175e+02 3.880e+02 7.852e+02, threshold=6.351e+02, percent-clipped=2.0 2023-05-02 23:24:38,926 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:24:50,439 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 23:25:14,085 INFO [train.py:904] (5/8) Epoch 30, batch 7100, loss[loss=0.1945, simple_loss=0.2849, pruned_loss=0.0521, over 17016.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2886, pruned_loss=0.0556, over 3085024.78 frames. ], batch size: 41, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:25:20,971 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301458.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:25:45,818 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8932, 2.8695, 2.9964, 1.6039, 3.1282, 3.2644, 2.6717, 2.3733], device='cuda:5'), covar=tensor([0.1192, 0.0287, 0.0265, 0.1414, 0.0143, 0.0232, 0.0533, 0.0657], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0139, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 23:26:08,196 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4660, 4.7007, 4.4887, 4.4857, 4.2683, 4.1818, 4.2612, 4.7285], device='cuda:5'), covar=tensor([0.1175, 0.0864, 0.1108, 0.0904, 0.0845, 0.1624, 0.1069, 0.0978], device='cuda:5'), in_proj_covar=tensor([0.0724, 0.0874, 0.0720, 0.0682, 0.0558, 0.0558, 0.0733, 0.0686], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-02 23:26:14,373 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301491.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:21,593 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:21,994 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 23:26:33,766 INFO [train.py:904] (5/8) Epoch 30, batch 7150, loss[loss=0.1819, simple_loss=0.2699, pruned_loss=0.04694, over 16714.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2876, pruned_loss=0.05567, over 3078430.57 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:26:36,535 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301506.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:37,393 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.724e+02 3.109e+02 4.036e+02 9.536e+02, threshold=6.218e+02, percent-clipped=1.0 2023-05-02 23:27:04,776 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9812, 4.9826, 4.8159, 4.0552, 4.9009, 1.9589, 4.6290, 4.4898], device='cuda:5'), covar=tensor([0.0108, 0.0101, 0.0221, 0.0479, 0.0107, 0.2920, 0.0139, 0.0296], device='cuda:5'), in_proj_covar=tensor([0.0186, 0.0177, 0.0217, 0.0188, 0.0194, 0.0221, 0.0205, 0.0181], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 23:27:37,678 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 23:27:49,049 INFO [train.py:904] (5/8) Epoch 30, batch 7200, loss[loss=0.1866, simple_loss=0.2774, pruned_loss=0.04787, over 15455.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2855, pruned_loss=0.05433, over 3066725.93 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:28:40,366 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3490, 2.0744, 1.7299, 1.8454, 2.3449, 2.0266, 2.0462, 2.5090], device='cuda:5'), covar=tensor([0.0328, 0.0573, 0.0760, 0.0647, 0.0356, 0.0512, 0.0287, 0.0351], device='cuda:5'), in_proj_covar=tensor([0.0234, 0.0244, 0.0235, 0.0236, 0.0247, 0.0242, 0.0242, 0.0245], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 23:29:10,981 INFO [train.py:904] (5/8) Epoch 30, batch 7250, loss[loss=0.1749, simple_loss=0.2656, pruned_loss=0.04207, over 16929.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2828, pruned_loss=0.05304, over 3067481.29 frames. ], batch size: 109, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:29:15,148 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.328e+02 2.684e+02 3.448e+02 5.554e+02, threshold=5.368e+02, percent-clipped=0.0 2023-05-02 23:30:27,542 INFO [train.py:904] (5/8) Epoch 30, batch 7300, loss[loss=0.2245, simple_loss=0.2994, pruned_loss=0.07485, over 11589.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2822, pruned_loss=0.05288, over 3076266.80 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:31:22,706 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5984, 2.6987, 2.3308, 2.5872, 3.0120, 2.6495, 3.0709, 3.1987], device='cuda:5'), covar=tensor([0.0136, 0.0445, 0.0550, 0.0419, 0.0281, 0.0414, 0.0265, 0.0283], device='cuda:5'), in_proj_covar=tensor([0.0233, 0.0243, 0.0234, 0.0235, 0.0245, 0.0241, 0.0241, 0.0244], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 23:31:44,858 INFO [train.py:904] (5/8) Epoch 30, batch 7350, loss[loss=0.1901, simple_loss=0.2785, pruned_loss=0.05088, over 16784.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2842, pruned_loss=0.05444, over 3041309.03 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:31:50,965 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 2.739e+02 3.075e+02 3.816e+02 5.602e+02, threshold=6.150e+02, percent-clipped=1.0 2023-05-02 23:32:44,290 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301743.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:33:00,864 INFO [train.py:904] (5/8) Epoch 30, batch 7400, loss[loss=0.1769, simple_loss=0.2757, pruned_loss=0.03903, over 16846.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2846, pruned_loss=0.0543, over 3065709.76 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:33:15,212 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3883, 3.1945, 3.5429, 1.8088, 3.6510, 3.7089, 2.9220, 2.7734], device='cuda:5'), covar=tensor([0.0832, 0.0295, 0.0211, 0.1305, 0.0110, 0.0214, 0.0461, 0.0510], device='cuda:5'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0139, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 23:33:54,218 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:33:56,651 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 23:34:10,809 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301796.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:23,456 INFO [train.py:904] (5/8) Epoch 30, batch 7450, loss[loss=0.2324, simple_loss=0.2991, pruned_loss=0.08287, over 11596.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2859, pruned_loss=0.05579, over 3053398.41 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:34:24,241 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301804.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:29,852 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.990e+02 2.877e+02 3.334e+02 4.200e+02 6.879e+02, threshold=6.668e+02, percent-clipped=3.0 2023-05-02 23:35:30,252 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:35:45,205 INFO [train.py:904] (5/8) Epoch 30, batch 7500, loss[loss=0.2349, simple_loss=0.3043, pruned_loss=0.08269, over 11581.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2851, pruned_loss=0.05422, over 3066976.08 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:35:54,030 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 23:36:57,912 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:05,269 INFO [train.py:904] (5/8) Epoch 30, batch 7550, loss[loss=0.2092, simple_loss=0.293, pruned_loss=0.06267, over 16420.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2846, pruned_loss=0.05476, over 3052669.06 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:37:07,795 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301905.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:11,512 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.767e+02 3.333e+02 4.100e+02 7.380e+02, threshold=6.666e+02, percent-clipped=1.0 2023-05-02 23:38:23,956 INFO [train.py:904] (5/8) Epoch 30, batch 7600, loss[loss=0.1862, simple_loss=0.2806, pruned_loss=0.04596, over 17196.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.284, pruned_loss=0.05508, over 3048653.36 frames. ], batch size: 46, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:38:34,794 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301960.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 23:38:43,772 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:39:47,162 INFO [train.py:904] (5/8) Epoch 30, batch 7650, loss[loss=0.1972, simple_loss=0.2825, pruned_loss=0.05592, over 16695.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2846, pruned_loss=0.05571, over 3047766.40 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:39:53,139 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 2.820e+02 3.410e+02 4.459e+02 8.175e+02, threshold=6.820e+02, percent-clipped=3.0 2023-05-02 23:41:02,995 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4066, 3.4905, 2.1430, 3.8271, 2.6758, 3.8377, 2.3599, 2.8501], device='cuda:5'), covar=tensor([0.0324, 0.0415, 0.1685, 0.0279, 0.0867, 0.0633, 0.1484, 0.0810], device='cuda:5'), in_proj_covar=tensor([0.0177, 0.0182, 0.0196, 0.0173, 0.0180, 0.0220, 0.0204, 0.0184], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-02 23:41:06,729 INFO [train.py:904] (5/8) Epoch 30, batch 7700, loss[loss=0.1824, simple_loss=0.2666, pruned_loss=0.04914, over 16119.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2846, pruned_loss=0.05635, over 3045474.35 frames. ], batch size: 35, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:41:57,593 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:18,125 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:26,402 INFO [train.py:904] (5/8) Epoch 30, batch 7750, loss[loss=0.2036, simple_loss=0.2952, pruned_loss=0.05598, over 16729.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2847, pruned_loss=0.05593, over 3060477.42 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:42:33,839 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.636e+02 3.091e+02 3.789e+02 8.720e+02, threshold=6.183e+02, percent-clipped=2.0 2023-05-02 23:43:12,802 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302134.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:43:41,536 INFO [train.py:904] (5/8) Epoch 30, batch 7800, loss[loss=0.2053, simple_loss=0.294, pruned_loss=0.05823, over 16285.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2857, pruned_loss=0.05673, over 3067570.75 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:44:55,025 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302200.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:45:00,020 INFO [train.py:904] (5/8) Epoch 30, batch 7850, loss[loss=0.215, simple_loss=0.2897, pruned_loss=0.07014, over 11344.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2859, pruned_loss=0.05588, over 3066842.63 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:45:06,746 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.702e+02 3.253e+02 3.935e+02 9.747e+02, threshold=6.506e+02, percent-clipped=3.0 2023-05-02 23:46:14,777 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 23:46:15,242 INFO [train.py:904] (5/8) Epoch 30, batch 7900, loss[loss=0.2265, simple_loss=0.3013, pruned_loss=0.07585, over 11700.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2851, pruned_loss=0.05533, over 3077470.22 frames. ], batch size: 250, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:46:16,929 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302255.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:46:25,392 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:46:25,490 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:46:45,815 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 23:47:22,551 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302296.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:47:32,980 INFO [train.py:904] (5/8) Epoch 30, batch 7950, loss[loss=0.1901, simple_loss=0.2768, pruned_loss=0.05176, over 17045.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2857, pruned_loss=0.05574, over 3086719.45 frames. ], batch size: 55, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:47:41,024 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.902e+02 2.712e+02 3.186e+02 3.569e+02 8.017e+02, threshold=6.372e+02, percent-clipped=2.0 2023-05-02 23:48:48,639 INFO [train.py:904] (5/8) Epoch 30, batch 8000, loss[loss=0.2567, simple_loss=0.3208, pruned_loss=0.09633, over 11635.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2863, pruned_loss=0.05621, over 3087325.99 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:48:54,576 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:49:57,556 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302399.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:50:03,901 INFO [train.py:904] (5/8) Epoch 30, batch 8050, loss[loss=0.2061, simple_loss=0.2966, pruned_loss=0.05778, over 15418.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2863, pruned_loss=0.05614, over 3075555.15 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:50:11,664 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.982e+02 3.316e+02 3.990e+02 1.007e+03, threshold=6.633e+02, percent-clipped=4.0 2023-05-02 23:50:31,548 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.4559, 4.3256, 4.5111, 4.6590, 4.8170, 4.4218, 4.7981, 4.8474], device='cuda:5'), covar=tensor([0.2196, 0.1378, 0.1823, 0.0864, 0.0735, 0.1018, 0.0750, 0.0831], device='cuda:5'), in_proj_covar=tensor([0.0672, 0.0817, 0.0945, 0.0838, 0.0635, 0.0656, 0.0696, 0.0805], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-02 23:51:10,644 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302447.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:51:21,037 INFO [train.py:904] (5/8) Epoch 30, batch 8100, loss[loss=0.1879, simple_loss=0.276, pruned_loss=0.04989, over 15476.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2848, pruned_loss=0.055, over 3089656.26 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:51:56,200 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:52:36,005 INFO [train.py:904] (5/8) Epoch 30, batch 8150, loss[loss=0.1736, simple_loss=0.2597, pruned_loss=0.04381, over 16415.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2816, pruned_loss=0.05318, over 3110199.63 frames. ], batch size: 75, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:52:43,486 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.950e+02 2.749e+02 3.387e+02 4.131e+02 6.245e+02, threshold=6.774e+02, percent-clipped=0.0 2023-05-02 23:53:14,966 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6454, 3.8985, 2.9660, 2.3481, 2.6261, 2.5868, 4.1539, 3.4127], device='cuda:5'), covar=tensor([0.3283, 0.0616, 0.1873, 0.3004, 0.2915, 0.2130, 0.0462, 0.1474], device='cuda:5'), in_proj_covar=tensor([0.0339, 0.0276, 0.0316, 0.0330, 0.0308, 0.0281, 0.0305, 0.0353], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:5') 2023-05-02 23:53:28,250 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302538.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:53:52,245 INFO [train.py:904] (5/8) Epoch 30, batch 8200, loss[loss=0.1801, simple_loss=0.2745, pruned_loss=0.04288, over 16248.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2798, pruned_loss=0.05317, over 3098687.77 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:53:53,888 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:53:55,137 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:54:03,365 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302561.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:13,018 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:13,853 INFO [train.py:904] (5/8) Epoch 30, batch 8250, loss[loss=0.1691, simple_loss=0.2735, pruned_loss=0.03237, over 16822.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2791, pruned_loss=0.05086, over 3096561.59 frames. ], batch size: 96, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:55:22,071 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.554e+02 2.948e+02 3.684e+02 6.863e+02, threshold=5.896e+02, percent-clipped=1.0 2023-05-02 23:55:23,079 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:57,344 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 23:56:02,641 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:35,855 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:38,711 INFO [train.py:904] (5/8) Epoch 30, batch 8300, loss[loss=0.1592, simple_loss=0.2548, pruned_loss=0.0318, over 15440.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2768, pruned_loss=0.04817, over 3087605.73 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:57:43,506 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302693.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 23:58:01,698 INFO [train.py:904] (5/8) Epoch 30, batch 8350, loss[loss=0.1822, simple_loss=0.2655, pruned_loss=0.04947, over 12095.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.276, pruned_loss=0.04625, over 3078449.47 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:58:09,446 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.283e+02 2.623e+02 3.174e+02 4.916e+02, threshold=5.246e+02, percent-clipped=0.0 2023-05-02 23:58:23,056 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 23:58:38,850 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-05-02 23:59:22,773 INFO [train.py:904] (5/8) Epoch 30, batch 8400, loss[loss=0.1743, simple_loss=0.2707, pruned_loss=0.03892, over 16724.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2733, pruned_loss=0.04445, over 3065484.41 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:00:44,414 INFO [train.py:904] (5/8) Epoch 30, batch 8450, loss[loss=0.1748, simple_loss=0.2587, pruned_loss=0.04543, over 12386.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2712, pruned_loss=0.04304, over 3045118.85 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:00:52,080 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.019e+02 2.324e+02 2.808e+02 4.179e+02, threshold=4.647e+02, percent-clipped=0.0 2023-05-03 00:00:54,609 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([1.8704, 2.9020, 2.5904, 4.2265, 2.7138, 4.0590, 1.5967, 3.1121], device='cuda:5'), covar=tensor([0.1301, 0.0698, 0.1091, 0.0187, 0.0125, 0.0330, 0.1716, 0.0644], device='cuda:5'), in_proj_covar=tensor([0.0175, 0.0182, 0.0202, 0.0206, 0.0207, 0.0219, 0.0211, 0.0200], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-03 00:01:31,167 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302833.0, num_to_drop=1, layers_to_drop={3} 2023-05-03 00:01:57,025 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-03 00:02:04,577 INFO [train.py:904] (5/8) Epoch 30, batch 8500, loss[loss=0.1474, simple_loss=0.2419, pruned_loss=0.02646, over 15223.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.268, pruned_loss=0.04125, over 3040805.38 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:02:08,494 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302856.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:02:48,229 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302881.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:03:01,312 INFO [scaling.py:679] (5/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-05-03 00:03:26,611 INFO [train.py:904] (5/8) Epoch 30, batch 8550, loss[loss=0.1748, simple_loss=0.2759, pruned_loss=0.03685, over 16879.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2663, pruned_loss=0.04017, over 3048263.78 frames. ], batch size: 116, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:03:28,402 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302904.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:03:38,836 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.520e+02 2.175e+02 2.493e+02 2.958e+02 5.835e+02, threshold=4.986e+02, percent-clipped=2.0 2023-05-03 00:04:45,106 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302942.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:05:04,987 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302952.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:05:08,035 INFO [train.py:904] (5/8) Epoch 30, batch 8600, loss[loss=0.1698, simple_loss=0.2668, pruned_loss=0.03637, over 16817.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2659, pruned_loss=0.03905, over 3032034.83 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:06:16,113 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302988.0, num_to_drop=1, layers_to_drop={3} 2023-05-03 00:06:39,316 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303000.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:06:46,615 INFO [train.py:904] (5/8) Epoch 30, batch 8650, loss[loss=0.1488, simple_loss=0.254, pruned_loss=0.02175, over 16843.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.264, pruned_loss=0.03779, over 3006091.59 frames. ], batch size: 102, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:06:58,837 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 2.161e+02 2.643e+02 3.389e+02 5.700e+02, threshold=5.286e+02, percent-clipped=3.0 2023-05-03 00:08:31,347 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303052.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:08:34,588 INFO [train.py:904] (5/8) Epoch 30, batch 8700, loss[loss=0.1518, simple_loss=0.2447, pruned_loss=0.02949, over 16604.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2614, pruned_loss=0.03655, over 3022283.23 frames. ], batch size: 62, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:10:14,439 INFO [train.py:904] (5/8) Epoch 30, batch 8750, loss[loss=0.1693, simple_loss=0.2722, pruned_loss=0.03323, over 16466.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2612, pruned_loss=0.03606, over 3015145.51 frames. ], batch size: 75, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:10:25,192 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 2.109e+02 2.507e+02 3.042e+02 7.094e+02, threshold=5.015e+02, percent-clipped=1.0 2023-05-03 00:10:38,657 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303113.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:11:15,605 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-03 00:11:23,659 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303133.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:12:06,346 INFO [train.py:904] (5/8) Epoch 30, batch 8800, loss[loss=0.1663, simple_loss=0.2613, pruned_loss=0.03569, over 16239.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2607, pruned_loss=0.03566, over 3025729.93 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:13:03,255 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303181.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:13:50,930 INFO [train.py:904] (5/8) Epoch 30, batch 8850, loss[loss=0.1509, simple_loss=0.2435, pruned_loss=0.02913, over 12693.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2631, pruned_loss=0.03487, over 3037610.39 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:14:00,695 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.249e+02 2.609e+02 3.163e+02 6.486e+02, threshold=5.217e+02, percent-clipped=3.0 2023-05-03 00:14:15,061 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-03 00:15:03,450 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303237.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:15:38,693 INFO [train.py:904] (5/8) Epoch 30, batch 8900, loss[loss=0.1711, simple_loss=0.2668, pruned_loss=0.03775, over 16624.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2639, pruned_loss=0.03417, over 3045707.47 frames. ], batch size: 62, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:15:47,401 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3479, 4.3916, 4.6826, 4.6736, 4.6803, 4.4179, 4.3776, 4.3569], device='cuda:5'), covar=tensor([0.0344, 0.0661, 0.0393, 0.0382, 0.0472, 0.0463, 0.0929, 0.0448], device='cuda:5'), in_proj_covar=tensor([0.0431, 0.0489, 0.0470, 0.0435, 0.0514, 0.0496, 0.0569, 0.0400], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-03 00:15:49,495 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303258.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:17:03,890 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303288.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:17:24,953 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.3694, 4.6986, 4.4838, 4.5332, 4.2697, 4.2303, 4.1450, 4.7426], device='cuda:5'), covar=tensor([0.1262, 0.0927, 0.1119, 0.0866, 0.0818, 0.1500, 0.1265, 0.0973], device='cuda:5'), in_proj_covar=tensor([0.0713, 0.0857, 0.0705, 0.0668, 0.0544, 0.0547, 0.0717, 0.0673], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:17:43,161 INFO [train.py:904] (5/8) Epoch 30, batch 8950, loss[loss=0.1476, simple_loss=0.2461, pruned_loss=0.02457, over 16605.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2635, pruned_loss=0.03435, over 3048264.52 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:17:53,311 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.074e+02 2.389e+02 2.805e+02 4.954e+02, threshold=4.779e+02, percent-clipped=0.0 2023-05-03 00:18:16,141 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303319.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:18:53,963 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303336.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:19:32,199 INFO [train.py:904] (5/8) Epoch 30, batch 9000, loss[loss=0.1481, simple_loss=0.2421, pruned_loss=0.02702, over 12291.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2598, pruned_loss=0.03306, over 3059787.14 frames. ], batch size: 250, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:19:32,199 INFO [train.py:929] (5/8) Computing validation loss 2023-05-03 00:19:42,084 INFO [train.py:938] (5/8) Epoch 30, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01984, over 944034.00 frames. 2023-05-03 00:19:42,086 INFO [train.py:939] (5/8) Maximum memory allocated so far is 17913MB 2023-05-03 00:21:28,466 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303403.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:29,182 INFO [train.py:904] (5/8) Epoch 30, batch 9050, loss[loss=0.1838, simple_loss=0.2675, pruned_loss=0.05008, over 12940.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2608, pruned_loss=0.03356, over 3062975.83 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:21:38,014 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303408.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:39,556 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 2.159e+02 2.466e+02 3.038e+02 8.070e+02, threshold=4.932e+02, percent-clipped=3.0 2023-05-03 00:22:30,376 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.2721, 4.2531, 4.0932, 3.3846, 4.1888, 1.7422, 3.9725, 3.7576], device='cuda:5'), covar=tensor([0.0147, 0.0130, 0.0237, 0.0290, 0.0140, 0.2983, 0.0173, 0.0335], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0172, 0.0211, 0.0181, 0.0188, 0.0216, 0.0199, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:23:00,915 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4392, 2.7547, 3.1468, 1.9687, 2.7944, 2.0751, 3.0399, 3.0122], device='cuda:5'), covar=tensor([0.0296, 0.1060, 0.0553, 0.2239, 0.0850, 0.1085, 0.0668, 0.1039], device='cuda:5'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0155, 0.0146, 0.0131, 0.0143, 0.0180], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:5') 2023-05-03 00:23:07,839 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.6601, 2.6209, 1.9216, 2.7874, 2.1571, 2.8193, 2.1667, 2.4043], device='cuda:5'), covar=tensor([0.0350, 0.0367, 0.1300, 0.0281, 0.0711, 0.0570, 0.1255, 0.0638], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0178, 0.0192, 0.0168, 0.0176, 0.0215, 0.0201, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-03 00:23:13,949 INFO [train.py:904] (5/8) Epoch 30, batch 9100, loss[loss=0.1825, simple_loss=0.2668, pruned_loss=0.0491, over 12187.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2607, pruned_loss=0.03433, over 3061805.94 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:23:33,683 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303464.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:25:12,855 INFO [train.py:904] (5/8) Epoch 30, batch 9150, loss[loss=0.1604, simple_loss=0.2567, pruned_loss=0.03207, over 16201.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2616, pruned_loss=0.03455, over 3051897.33 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:25:25,341 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.268e+02 2.639e+02 3.152e+02 6.290e+02, threshold=5.277e+02, percent-clipped=4.0 2023-05-03 00:26:23,067 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303537.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:26:54,953 INFO [train.py:904] (5/8) Epoch 30, batch 9200, loss[loss=0.1512, simple_loss=0.2409, pruned_loss=0.03077, over 17111.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2576, pruned_loss=0.03377, over 3076218.71 frames. ], batch size: 49, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:27:08,063 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.2337, 3.2788, 2.0090, 3.6192, 2.5316, 3.5920, 2.1803, 2.7522], device='cuda:5'), covar=tensor([0.0398, 0.0468, 0.1746, 0.0357, 0.0930, 0.0626, 0.1693, 0.0834], device='cuda:5'), in_proj_covar=tensor([0.0174, 0.0178, 0.0192, 0.0168, 0.0176, 0.0215, 0.0201, 0.0179], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-03 00:27:31,990 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303574.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:27:52,353 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303585.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:28:27,172 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.9999, 2.1891, 2.2909, 3.4583, 2.1413, 2.4269, 2.3293, 2.2955], device='cuda:5'), covar=tensor([0.1446, 0.4052, 0.3428, 0.0710, 0.4659, 0.2831, 0.3845, 0.3846], device='cuda:5'), in_proj_covar=tensor([0.0418, 0.0471, 0.0386, 0.0332, 0.0441, 0.0540, 0.0445, 0.0551], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:28:29,161 INFO [train.py:904] (5/8) Epoch 30, batch 9250, loss[loss=0.166, simple_loss=0.2597, pruned_loss=0.03618, over 16812.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2573, pruned_loss=0.03376, over 3068937.53 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:28:41,900 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.148e+02 2.583e+02 3.115e+02 7.022e+02, threshold=5.165e+02, percent-clipped=1.0 2023-05-03 00:28:49,585 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303614.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:29:39,042 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.5043, 3.6420, 3.6649, 2.6122, 3.3107, 3.6863, 3.4338, 2.1648], device='cuda:5'), covar=tensor([0.0473, 0.0066, 0.0060, 0.0383, 0.0131, 0.0101, 0.0091, 0.0507], device='cuda:5'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0134, 0.0103, 0.0114, 0.0097, 0.0130], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-03 00:29:39,069 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303635.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:30:19,441 INFO [train.py:904] (5/8) Epoch 30, batch 9300, loss[loss=0.1482, simple_loss=0.2366, pruned_loss=0.02993, over 16592.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2556, pruned_loss=0.03331, over 3069348.83 frames. ], batch size: 62, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:30:25,049 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.7798, 4.0214, 4.1115, 2.9542, 3.6217, 4.0984, 3.7262, 2.4110], device='cuda:5'), covar=tensor([0.0474, 0.0060, 0.0047, 0.0367, 0.0126, 0.0098, 0.0093, 0.0480], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0134, 0.0102, 0.0114, 0.0097, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-03 00:31:05,542 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.1510, 4.1749, 4.4588, 4.4314, 4.4348, 4.2303, 4.2154, 4.2148], device='cuda:5'), covar=tensor([0.0439, 0.1011, 0.0659, 0.0806, 0.0862, 0.0755, 0.1041, 0.0599], device='cuda:5'), in_proj_covar=tensor([0.0432, 0.0491, 0.0471, 0.0437, 0.0516, 0.0498, 0.0570, 0.0402], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-03 00:31:53,275 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4466, 3.4058, 3.4960, 3.5642, 3.5962, 3.3597, 3.5845, 3.6575], device='cuda:5'), covar=tensor([0.1401, 0.0936, 0.1062, 0.0710, 0.0612, 0.2155, 0.0895, 0.0796], device='cuda:5'), in_proj_covar=tensor([0.0657, 0.0796, 0.0919, 0.0824, 0.0621, 0.0645, 0.0682, 0.0789], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:32:05,586 INFO [train.py:904] (5/8) Epoch 30, batch 9350, loss[loss=0.1539, simple_loss=0.2418, pruned_loss=0.03301, over 12092.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2553, pruned_loss=0.03327, over 3067082.45 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:32:13,874 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303708.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:32:16,692 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.024e+02 2.345e+02 2.902e+02 6.854e+02, threshold=4.690e+02, percent-clipped=2.0 2023-05-03 00:33:46,256 INFO [train.py:904] (5/8) Epoch 30, batch 9400, loss[loss=0.1713, simple_loss=0.2749, pruned_loss=0.03386, over 16870.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2549, pruned_loss=0.03257, over 3076624.04 frames. ], batch size: 116, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:33:50,775 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303756.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:33:56,173 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303759.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:35:14,464 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.9343, 5.2288, 5.4345, 5.1862, 5.2686, 5.7767, 5.2449, 4.9923], device='cuda:5'), covar=tensor([0.0938, 0.1692, 0.1837, 0.1705, 0.2210, 0.0796, 0.1382, 0.2082], device='cuda:5'), in_proj_covar=tensor([0.0419, 0.0623, 0.0699, 0.0507, 0.0673, 0.0719, 0.0540, 0.0677], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-03 00:35:25,710 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.1777, 2.5744, 2.6397, 1.9598, 2.7596, 2.8111, 2.5287, 2.5362], device='cuda:5'), covar=tensor([0.0627, 0.0283, 0.0238, 0.0984, 0.0129, 0.0272, 0.0428, 0.0427], device='cuda:5'), in_proj_covar=tensor([0.0144, 0.0107, 0.0098, 0.0134, 0.0084, 0.0128, 0.0127, 0.0127], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:5') 2023-05-03 00:35:26,329 INFO [train.py:904] (5/8) Epoch 30, batch 9450, loss[loss=0.1644, simple_loss=0.2512, pruned_loss=0.03885, over 12410.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2567, pruned_loss=0.03279, over 3061153.11 frames. ], batch size: 250, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:35:29,339 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.0581, 2.2033, 2.2538, 3.6448, 2.1595, 2.5227, 2.3359, 2.3449], device='cuda:5'), covar=tensor([0.1487, 0.3909, 0.3412, 0.0647, 0.4385, 0.2730, 0.3837, 0.3733], device='cuda:5'), in_proj_covar=tensor([0.0418, 0.0471, 0.0386, 0.0332, 0.0441, 0.0540, 0.0445, 0.0550], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:35:37,000 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 2.120e+02 2.504e+02 3.073e+02 7.133e+02, threshold=5.009e+02, percent-clipped=4.0 2023-05-03 00:35:38,600 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303810.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:35:59,104 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-03 00:37:08,355 INFO [train.py:904] (5/8) Epoch 30, batch 9500, loss[loss=0.1667, simple_loss=0.2598, pruned_loss=0.0368, over 16582.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.256, pruned_loss=0.03265, over 3063652.76 frames. ], batch size: 62, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:37:43,977 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303871.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:36,437 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303896.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:54,510 INFO [train.py:904] (5/8) Epoch 30, batch 9550, loss[loss=0.1709, simple_loss=0.2737, pruned_loss=0.03403, over 16344.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2558, pruned_loss=0.03274, over 3062098.29 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:39:08,521 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 2.050e+02 2.496e+02 3.159e+02 5.329e+02, threshold=4.992e+02, percent-clipped=3.0 2023-05-03 00:39:18,213 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303914.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:39:49,727 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303930.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:40:36,090 INFO [train.py:904] (5/8) Epoch 30, batch 9600, loss[loss=0.151, simple_loss=0.2407, pruned_loss=0.03064, over 12649.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2566, pruned_loss=0.03325, over 3062172.87 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:40:43,424 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303957.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:40:53,328 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303962.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:41:16,233 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.5399, 4.5378, 4.3524, 3.6408, 4.4254, 1.7368, 4.2099, 4.0327], device='cuda:5'), covar=tensor([0.0109, 0.0093, 0.0233, 0.0311, 0.0133, 0.3007, 0.0146, 0.0323], device='cuda:5'), in_proj_covar=tensor([0.0182, 0.0173, 0.0211, 0.0181, 0.0188, 0.0217, 0.0199, 0.0176], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:41:16,266 INFO [zipformer.py:625] (5/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303974.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:42:24,255 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([5.7887, 6.1154, 5.8545, 5.9264, 5.5913, 5.5049, 5.4163, 6.2227], device='cuda:5'), covar=tensor([0.1301, 0.0893, 0.1003, 0.0803, 0.0704, 0.0626, 0.1298, 0.0865], device='cuda:5'), in_proj_covar=tensor([0.0707, 0.0850, 0.0699, 0.0663, 0.0542, 0.0542, 0.0712, 0.0668], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:42:31,202 INFO [train.py:904] (5/8) Epoch 30, batch 9650, loss[loss=0.1732, simple_loss=0.2609, pruned_loss=0.04277, over 12285.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2585, pruned_loss=0.0338, over 3040713.47 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:42:48,102 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.176e+02 2.625e+02 3.264e+02 9.247e+02, threshold=5.250e+02, percent-clipped=2.0 2023-05-03 00:43:41,197 INFO [zipformer.py:625] (5/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304035.0, num_to_drop=1, layers_to_drop={3} 2023-05-03 00:44:20,849 INFO [train.py:904] (5/8) Epoch 30, batch 9700, loss[loss=0.1727, simple_loss=0.2676, pruned_loss=0.03891, over 16358.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2581, pruned_loss=0.03364, over 3045120.54 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:44:30,042 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304059.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:46:04,106 INFO [train.py:904] (5/8) Epoch 30, batch 9750, loss[loss=0.184, simple_loss=0.2732, pruned_loss=0.04738, over 16741.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2579, pruned_loss=0.03407, over 3047300.73 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:46:09,340 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=304107.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:46:17,047 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.052e+02 2.422e+02 2.861e+02 5.308e+02, threshold=4.844e+02, percent-clipped=2.0 2023-05-03 00:47:41,842 INFO [train.py:904] (5/8) Epoch 30, batch 9800, loss[loss=0.1798, simple_loss=0.2806, pruned_loss=0.03956, over 15346.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.258, pruned_loss=0.03322, over 3058739.09 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:48:04,644 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304166.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:48:38,214 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([2.3294, 3.6349, 3.6611, 2.5202, 3.2174, 3.6566, 3.4660, 2.0789], device='cuda:5'), covar=tensor([0.0551, 0.0064, 0.0062, 0.0408, 0.0152, 0.0099, 0.0093, 0.0521], device='cuda:5'), in_proj_covar=tensor([0.0137, 0.0090, 0.0091, 0.0134, 0.0102, 0.0114, 0.0097, 0.0129], device='cuda:5'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:5') 2023-05-03 00:48:53,967 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.4587, 3.4989, 3.6919, 3.6727, 3.7055, 3.5324, 3.5601, 3.5876], device='cuda:5'), covar=tensor([0.0425, 0.0894, 0.0568, 0.0537, 0.0541, 0.0647, 0.0813, 0.0488], device='cuda:5'), in_proj_covar=tensor([0.0429, 0.0488, 0.0469, 0.0435, 0.0513, 0.0495, 0.0566, 0.0399], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:5') 2023-05-03 00:49:24,995 INFO [train.py:904] (5/8) Epoch 30, batch 9850, loss[loss=0.1429, simple_loss=0.231, pruned_loss=0.02738, over 12328.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2586, pruned_loss=0.03271, over 3066004.96 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:49:39,350 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.072e+02 2.330e+02 2.794e+02 5.653e+02, threshold=4.660e+02, percent-clipped=2.0 2023-05-03 00:50:19,720 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304230.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:51:13,345 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304252.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:51:16,025 INFO [train.py:904] (5/8) Epoch 30, batch 9900, loss[loss=0.1619, simple_loss=0.2693, pruned_loss=0.02724, over 16706.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.259, pruned_loss=0.03282, over 3063487.48 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:52:13,325 INFO [zipformer.py:625] (5/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=304278.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:52:42,629 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.0214, 3.8867, 4.0569, 4.1652, 4.2848, 3.8439, 4.2741, 4.3162], device='cuda:5'), covar=tensor([0.1693, 0.1186, 0.1400, 0.0830, 0.0634, 0.1656, 0.0788, 0.0794], device='cuda:5'), in_proj_covar=tensor([0.0650, 0.0786, 0.0904, 0.0815, 0.0613, 0.0636, 0.0675, 0.0781], device='cuda:5'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:52:56,876 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1978, 2.3556, 2.1657, 2.1944, 2.7205, 2.4393, 2.6164, 2.8992], device='cuda:5'), covar=tensor([0.0187, 0.0529, 0.0559, 0.0574, 0.0300, 0.0467, 0.0299, 0.0306], device='cuda:5'), in_proj_covar=tensor([0.0225, 0.0239, 0.0229, 0.0230, 0.0241, 0.0238, 0.0233, 0.0236], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:52:59,559 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.1141, 2.3109, 2.1171, 2.1789, 2.6804, 2.3928, 2.5203, 2.8386], device='cuda:5'), covar=tensor([0.0211, 0.0541, 0.0629, 0.0601, 0.0343, 0.0491, 0.0318, 0.0324], device='cuda:5'), in_proj_covar=tensor([0.0225, 0.0239, 0.0229, 0.0230, 0.0241, 0.0238, 0.0233, 0.0236], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:5') 2023-05-03 00:53:14,797 INFO [train.py:904] (5/8) Epoch 30, batch 9950, loss[loss=0.1645, simple_loss=0.2625, pruned_loss=0.03327, over 16679.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2609, pruned_loss=0.03321, over 3049789.26 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:53:31,715 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 2.067e+02 2.374e+02 2.801e+02 5.953e+02, threshold=4.748e+02, percent-clipped=1.0 2023-05-03 00:54:21,398 INFO [zipformer.py:625] (5/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304330.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:54:28,005 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([4.6136, 4.7913, 4.8928, 4.7398, 4.7841, 5.3010, 4.7852, 4.5265], device='cuda:5'), covar=tensor([0.1251, 0.1784, 0.2233, 0.1815, 0.2308, 0.0950, 0.1560, 0.2236], device='cuda:5'), in_proj_covar=tensor([0.0413, 0.0617, 0.0693, 0.0502, 0.0668, 0.0712, 0.0535, 0.0671], device='cuda:5'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:5') 2023-05-03 00:55:15,646 INFO [train.py:904] (5/8) Epoch 30, batch 10000, loss[loss=0.1772, simple_loss=0.2644, pruned_loss=0.04505, over 12434.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2596, pruned_loss=0.03289, over 3056124.15 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:56:00,058 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-03 00:56:57,651 INFO [train.py:904] (5/8) Epoch 30, batch 10050, loss[loss=0.1839, simple_loss=0.2858, pruned_loss=0.04097, over 16940.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2598, pruned_loss=0.03309, over 3042188.41 frames. ], batch size: 109, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:57:10,561 INFO [optim.py:368] (5/8) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.064e+02 2.395e+02 2.910e+02 6.626e+02, threshold=4.790e+02, percent-clipped=3.0 2023-05-03 00:57:13,638 INFO [zipformer.py:1454] (5/8) attn_weights_entropy = tensor([3.5344, 3.5753, 2.2756, 4.0520, 2.6839, 3.9851, 2.3354, 2.8767], device='cuda:5'), covar=tensor([0.0363, 0.0437, 0.1625, 0.0255, 0.0946, 0.0601, 0.1667, 0.0898], device='cuda:5'), in_proj_covar=tensor([0.0172, 0.0176, 0.0190, 0.0167, 0.0175, 0.0212, 0.0199, 0.0178], device='cuda:5'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:5') 2023-05-03 00:57:17,950 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-03 00:57:54,757 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-03 00:58:34,294 INFO [train.py:904] (5/8) Epoch 30, batch 10100, loss[loss=0.1543, simple_loss=0.2493, pruned_loss=0.02965, over 16223.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.26, pruned_loss=0.03319, over 3037006.15 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:58:57,154 INFO [zipformer.py:625] (5/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304466.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:59:20,486 INFO [scaling.py:679] (5/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-03 00:59:55,820 INFO [train.py:1169] (5/8) Done!